Let the app propose a status. Keep the Airtable token out of its hands.
Set one record’s DemoStatus in a fixed disposable table.
-The app submits proof and the exact action. The general gateway owns the PATCH and token.
+The app submits proof and the exact action. The general gateway owns the PATCH and token.
+With the authenticated owner UI, separately approve bounded private acquisition and release to the local agent.
Checking the demo…
Approved authority
Fixed base
—
Fixed table
—
@@ -29,15 +30,20 @@
Let the app propose a status. Keep the Airtable token out of its hands.
-
Each experiment fixes its record, operation and signed arguments.
+
+
+
+
+
+
Each experiment fixes its record, operation and signed arguments.
The native review binds the exact action, authority and expiry.
Run the exact update before replay. Recovery resubmits the original operation to the native replay boundary;
-only another read can occur. Wait three seconds after response loss before observing again. The local simulator supplies signed responses only for this closed fixture; its UI state is never authority.
+
Run the exact update before replay. Recovery requests a separate native proof and exact approval for a bounded read.
+Replaying the original operation only inspects retained history. Wait three seconds after response loss before observing again. The local simulator supplies signed responses only for this closed fixture; its UI state is never authority.
Measured native result
No action submitted
-
Verify proof
Claim operation
Enter provider
Record response
Observe value + echo
-
Authorization, a recorded HTTP response and a matching fresh value/echo are separate facts.
+
Verify authority
Approve exact effect
Admit and grant
Enter provider or recipient
Record response or ACK
Observe separately
+
Authorization, an acquisition response, a recipient ACK and a matching fresh value/echo are separate facts.
This is a development demo, not protected production qualification. Simulation never reaches Airtable.
diff --git a/demos/airtable-gateway/web/owner.css b/demos/airtable-gateway/web/owner.css
new file mode 100644
index 000000000..a879c3d82
--- /dev/null
+++ b/demos/airtable-gateway/web/owner.css
@@ -0,0 +1 @@
+:root{font-family:ui-sans-serif,system-ui,sans-serif;color:#172c34;background:#f4f3ed}*{box-sizing:border-box}body{margin:0}main{max-width:1040px;margin:3rem auto;padding:0 1.5rem}h1{font-size:clamp(2rem,5vw,3.8rem);letter-spacing:-.045em}p{line-height:1.65;max-width:78ch}.eyebrow{text-transform:uppercase;letter-spacing:.14em;font-size:.8rem}form,section{padding:1.5rem;background:#fff;border:1px solid #c3cbc7;border-radius:.65rem;margin:1.5rem 0}label{display:block}input{display:block;width:min(100%,32rem);padding:.8rem;margin:1rem 0;font:inherit}button{background:#174e45;color:white;border:0;border-radius:.35rem;padding:.85rem 1.2rem;font:inherit;cursor:pointer}button:disabled{opacity:.45;cursor:wait}#decline{background:#594146}.actions{display:flex;gap:1rem;flex-wrap:wrap;margin-top:1rem}pre{white-space:pre-wrap;overflow-wrap:anywhere;background:#f3f5f3;padding:1rem;font: .85rem/1.6 ui-monospace,monospace}details{margin:1rem 0}.limits{font-size:.85rem;color:#4c605c}[hidden]{display:none!important}
diff --git a/demos/airtable-gateway/web/owner.html b/demos/airtable-gateway/web/owner.html
new file mode 100644
index 000000000..670134b00
--- /dev/null
+++ b/demos/airtable-gateway/web/owner.html
@@ -0,0 +1,14 @@
+
+
+Owner review · Auths
+
Private operator channel
Review the exact effect.
+
The agent proposes. Your authenticated decision permits the operator to sign the native request shown here. It does not supply missing resource authority or establish a provider outcome.
+
+
Authentication required. Private reviews are unavailable to the agent.
+
Complete native review
+
Escaped text below is data. Inspect the destination, exact request, consequences, reservations, recipient and recovery limits. Nothing is abbreviated.
+
Exact native signing request
+
Approval validity is shown as Unix seconds in not_before and expires_at. The gateway checks it again before entry.
+
+
Local development custody. Authentication maps this owner session to the installed exact approver. This UI makes no claim about human attention, production qualification or exclusive provider causation.
diff --git a/demos/airtable-gateway/web/owner.js b/demos/airtable-gateway/web/owner.js
new file mode 100644
index 000000000..836131602
--- /dev/null
+++ b/demos/airtable-gateway/web/owner.js
@@ -0,0 +1,70 @@
+'use strict';
+const element = id => document.getElementById(id);
+const renderer = 'auths.literal-complete-review/1';
+let bearer, csrf, pending, busy = false;
+const headers = () => bearer ? {'Authorization': 'Bearer ' + bearer} : {};
+async function post(path, value) {
+ const response = await fetch(path, {method: 'POST', headers: {
+ 'Content-Type': 'application/json', ...headers(), ...(csrf ? {'X-Owner-CSRF': csrf} : {})
+ }, body: JSON.stringify(value)});
+ if (!response.ok) throw new Error('The owner channel refused this request. Sign in again or inspect the current review.');
+ return response.json();
+}
+function clear() {
+ pending = null;
+ element('approve').disabled = true;
+ element('decline').disabled = true;
+ element('review').hidden = true;
+ element('literal').textContent = '';
+ element('signing').textContent = '';
+}
+async function refresh() {
+ if (!csrf || busy) return;
+ try {
+ const response = await fetch('/api/review', {cache: 'no-store', headers: headers()});
+ if (!response.ok) throw new Error('Owner session expired. Sign in again.');
+ const value = await response.json();
+ if (!value) { clear(); element('status').textContent = 'Signed in. Waiting for an exact native review.'; return; }
+ if (value.native.schema !== 'auths.private-effect-signing/2' || value.native.renderer !== renderer
+ || typeof value.native.literal_review !== 'string' || !value.native.literal_review) {
+ throw new Error('Unsupported native review. Approval is unavailable.');
+ }
+ element('literal').textContent = value.native.literal_review;
+ element('signing').textContent = JSON.stringify(value.native.approval_request, null, 2)
+ + '\nNative signing message:\n' + value.native.signing_message_hex;
+ pending = value.review_id;
+ element('review').hidden = false;
+ element('approve').disabled = false;
+ element('decline').disabled = false;
+ element('status').textContent = 'Review the full native value before deciding.';
+ } catch (error) { clear(); bearer = csrf = null; element('login').hidden = false; element('status').textContent = error.message; }
+}
+element('login').addEventListener('submit', async event => {
+ event.preventDefault();
+ const credential = element('credential').value;
+ element('credential').value = '';
+ try {
+ const session = await post('/api/login', {credential});
+ if (session.renderer !== renderer) throw new Error('Unsupported review renderer.');
+ csrf = session.csrf;
+ bearer = session.bearer;
+ element('login').hidden = true;
+ await refresh();
+ } catch (error) { element('status').textContent = error.message; }
+});
+for (const decision of ['approve', 'decline']) element(decision).addEventListener('click', async () => {
+ if (!pending || busy) return;
+ const review_id = pending;
+ busy = true;
+ element('approve').disabled = true;
+ element('decline').disabled = true;
+ try {
+ await post('/api/decision', {review_id, decision, renderer});
+ clear();
+ element('status').textContent = decision === 'approve'
+ ? 'Consent recorded for that exact native request. Inspect the gateway outcome separately.'
+ : 'Declined. No exact approval was issued.';
+ } catch (error) { clear(); element('status').textContent = error.message; }
+ finally { busy = false; }
+});
+setInterval(refresh, 1000);
diff --git a/demos/airtable-gateway/web/styles.css b/demos/airtable-gateway/web/styles.css
index cafb35838..ddd8ea6fc 100644
--- a/demos/airtable-gateway/web/styles.css
+++ b/demos/airtable-gateway/web/styles.css
@@ -1,3 +1,5 @@
:root{font-family:Inter,ui-sans-serif,system-ui,sans-serif;color:#122035;background:#f7f8fa;font-synthesis:none}*{box-sizing:border-box}body{margin:0}header,footer{padding:24px max(24px,calc((100vw - 1120px)/2));display:flex;justify-content:space-between;gap:20px;border-bottom:1px solid #dde2ea}header a{font-weight:800;letter-spacing:.08em;text-decoration:none;color:#122035}header span{font-size:11px;letter-spacing:.09em;font-weight:700;color:#25615a}main{max-width:1120px;margin:auto;padding:64px 24px}.eyebrow{font-size:12px;text-transform:uppercase;letter-spacing:.15em;color:#416489;font-weight:700}h1{font-size:clamp(30px,4vw,49px);line-height:1.16;letter-spacing:-.045em;margin:14px 0 22px}h2{font-size:21px;margin-top:0}.intro{font-size:18px;line-height:1.6;max-width:750px;color:#4a5c72}#connection{margin:30px 0;padding:14px 18px;border-left:3px solid #3574d5;background:#eef3fb;font-size:14px}.workbench{display:grid;grid-template-columns:1fr 1fr;gap:24px;margin:32px 0}article,.result{border:1px solid #dbe2ed;background:white;border-radius:12px;padding:28px}article p,.limits p{font-size:14px;line-height:1.7;color:#526074}dl{display:grid;grid-template-columns:1fr 1.4fr;gap:14px;font-size:13px}dt{color:#637389}dd{margin:0;font-weight:650;overflow-wrap:anywhere}.digest{font-family:ui-monospace,monospace;font-size:11px}label{display:block;font-size:13px;font-weight:700;margin:24px 0 10px}select{width:100%;padding:14px;background:#f7f9fd;border:1px solid #b7c6da;border-radius:7px;font:inherit;font-size:14px}button{width:100%;padding:16px;border:0;border-radius:7px;background:#2563eb;color:white;font:inherit;font-size:14px;font-weight:700;cursor:pointer}button:disabled{opacity:.5;cursor:wait}.small{font-size:12px!important}.stages{display:flex;flex-wrap:wrap;gap:8px;list-style:none;padding:0;color:#55708c;font-size:12px}.stages li{border:1px solid #dde5ef;padding:8px;border-radius:5px}.result #outcome{overflow-wrap:anywhere;font-size:26px}.result p{line-height:1.7;color:#526074}pre{white-space:pre-wrap;overflow-wrap:anywhere;background:#111f34;color:#d7e5f7;border-radius:8px;padding:22px;font:12px/1.65 ui-monospace,monospace;max-height:450px;overflow:auto}.result a{display:inline-block;padding-top:12px;font-size:14px;font-weight:700;color:#2563eb}.limits{padding:34px 2px 0;max-width:860px}footer{border-top:1px solid #dde2ea;font-size:12px;color:#637389;justify-content:center}@media(max-width:760px){main{padding-top:34px}.workbench{grid-template-columns:1fr}header{flex-direction:column}article,.result{padding:22px}dl{grid-template-columns:1fr 1fr}.result #outcome{font-size:21px}}
#provider-link{display:block;margin:14px 0;color:#2563eb;font-weight:700;font-size:14px}#provider-link[hidden]{display:none}
+
+#approval-review { white-space: pre-wrap; overflow-wrap: anywhere; }
diff --git a/demos/google-employee-delegation/broker/src/employee_read_demo/google_drive.py b/demos/google-employee-delegation/broker/src/employee_read_demo/google_drive.py
index b9aa907a3..723b3972b 100644
--- a/demos/google-employee-delegation/broker/src/employee_read_demo/google_drive.py
+++ b/demos/google-employee-delegation/broker/src/employee_read_demo/google_drive.py
@@ -4,7 +4,6 @@
import http.client
import ipaddress
-import json
import queue
import socket
import ssl
@@ -18,23 +17,16 @@
from .contracts import MAX_METADATA_BYTES, ReadCommand, TaskDescriptor
from .token_store import TokenVault
+from .google_metadata import (
+ DriveRepresentation, GoogleDenied, GoogleUnavailable, validate_metadata,
+)
HOST = "www.googleapis.com"
-DOC_MIME = "application/vnd.google-apps.document"
-TEXT_MIME = "text/plain"
METADATA_LIMIT = MAX_METADATA_BYTES
REQUEST_DEADLINE_SECONDS = 15
_DNS_SLOTS = threading.BoundedSemaphore(4)
-class GoogleDenied(PermissionError):
- """Provider refusal; no private file detail is included."""
-
-
-class GoogleUnavailable(RuntimeError):
- """Transport or malformed provider result; retry needs fresh admission."""
-
-
@dataclass
class ProviderTrace:
"""Observed Drive GETs and body bytes; capacity is reserved separately."""
@@ -217,7 +209,7 @@ def __init__(self, task: TaskDescriptor, tokens: TokenVault) -> None:
def set_attempt_observer(self, observer: ProviderAttemptObserver) -> None:
self.trace.observer = observer
- def _metadata(self, command: ReadCommand) -> tuple[dict, str]:
+ def _metadata(self, command: ReadCommand) -> tuple[DriveRepresentation, str]:
if command not in self.task.calls():
raise GoogleDenied("document is outside approved task")
if int(time.time()) >= self.task.expires_at:
@@ -232,23 +224,7 @@ def _metadata(self, command: ReadCommand) -> tuple[dict, str]:
raise GoogleDenied("Google refused document metadata")
if status != 200 or "application/json" not in content_type:
raise GoogleUnavailable("Google metadata request was unsuccessful")
- try:
- metadata = json.loads(body)
- except (ValueError, UnicodeDecodeError) as error:
- raise GoogleUnavailable("Google metadata is malformed") from error
- if not isinstance(metadata, dict) or metadata.get("id") != command.file_id or metadata.get("trashed") is True:
- raise GoogleDenied("Google document metadata did not match")
- capabilities = metadata.get("capabilities")
- if not isinstance(capabilities, dict) or capabilities.get("canDownload") is not True:
- raise GoogleDenied("Google does not permit download")
- mime = metadata.get("mimeType")
- if mime not in {DOC_MIME, TEXT_MIME}:
- raise GoogleDenied("unsupported Google document type")
- if mime == TEXT_MIME:
- size = metadata.get("size")
- if size is not None and (not str(size).isdecimal() or int(size) > command.max_bytes):
- raise GoogleDenied("document exceeds approved byte limit")
- return metadata, lease.access_token
+ return validate_metadata(body, command), lease.access_token
def inspect_exact(self, command: ReadCommand) -> None:
"""Check employee-readable metadata without fetching document content."""
@@ -259,10 +235,9 @@ def inspect_exact(self, command: ReadCommand) -> None:
def read_exact(self, command: ReadCommand) -> bytes:
self.trace = ProviderTrace(observer=self.trace.observer,
task_expires_at=self.task.expires_at)
- metadata, access_token = self._metadata(command)
+ representation, access_token = self._metadata(command)
file_id = quote(command.file_id, safe="")
- mime = metadata["mimeType"]
- if mime == DOC_MIME:
+ if representation is DriveRepresentation.GOOGLE_DOCUMENT:
path = f"/drive/v3/files/{file_id}/export?mimeType=text%2Fplain"
else:
path = f"/drive/v3/files/{file_id}?alt=media"
diff --git a/demos/google-employee-delegation/broker/src/employee_read_demo/google_metadata.py b/demos/google-employee-delegation/broker/src/employee_read_demo/google_metadata.py
new file mode 100644
index 000000000..950206f4e
--- /dev/null
+++ b/demos/google-employee-delegation/broker/src/employee_read_demo/google_metadata.py
@@ -0,0 +1,61 @@
+"""Concrete Drive metadata reference, separate from transport and admission.
+
+This application-owned domain reference grants no read, custody or disclosure
+capability. AP074's native recipe cutover must compare against its corpus and
+explicitly document stricter decoding instead of importing this code at runtime.
+"""
+
+from __future__ import annotations
+
+from enum import Enum
+import json
+
+from .contracts import MAX_METADATA_BYTES, ReadCommand
+
+
+class GoogleDenied(PermissionError):
+ """Provider refusal; no private file detail is included."""
+
+
+class GoogleUnavailable(RuntimeError):
+ """Transport or malformed result; no retry authority is created."""
+
+
+class DriveRepresentation(Enum):
+ GOOGLE_DOCUMENT = "application/vnd.google-apps.document"
+ PLAIN_TEXT = "text/plain"
+
+
+def validate_metadata(body: bytes, command: ReadCommand) -> DriveRepresentation:
+ """Validate one bounded transport response against the exact read resource.
+
+ The caller owns status/media/time checks; the finite byte cap is rechecked. This reference preserves the
+ existing optional size/trashed behavior and Python JSON duplicate handling.
+ Native gateway decoding must additionally reject duplicate members.
+ """
+ if len(body) > MAX_METADATA_BYTES:
+ raise GoogleUnavailable("Google metadata exceeds byte limit")
+ try:
+ metadata = json.loads(body)
+ except (ValueError, UnicodeDecodeError) as error:
+ raise GoogleUnavailable("Google metadata is malformed") from error
+ if (not isinstance(metadata, dict) or metadata.get("id") != command.file_id
+ or metadata.get("trashed") is True):
+ raise GoogleDenied("Google document metadata did not match")
+ capabilities = metadata.get("capabilities")
+ if not isinstance(capabilities, dict) or capabilities.get("canDownload") is not True:
+ raise GoogleDenied("Google does not permit download")
+ try:
+ representation = DriveRepresentation(metadata.get("mimeType"))
+ except (ValueError, TypeError) as error:
+ raise GoogleDenied("unsupported Google document type") from error
+ if representation is DriveRepresentation.PLAIN_TEXT:
+ size = metadata.get("size")
+ if size is not None:
+ try:
+ permitted = str(size).isdecimal() and int(size) <= command.max_bytes
+ except ValueError:
+ permitted = False
+ if not permitted:
+ raise GoogleDenied("document exceeds approved byte limit")
+ return representation
diff --git a/demos/google-employee-delegation/broker/tests/test_google_metadata_reference.py b/demos/google-employee-delegation/broker/tests/test_google_metadata_reference.py
new file mode 100644
index 000000000..5713172cf
--- /dev/null
+++ b/demos/google-employee-delegation/broker/tests/test_google_metadata_reference.py
@@ -0,0 +1,91 @@
+"""Freeze the concrete AP070 metadata reference before native capability extraction."""
+from __future__ import annotations
+
+import json
+import os
+from pathlib import Path
+import unittest
+
+from employee_read_demo.contracts import MAX_METADATA_BYTES, ReadCommand
+from employee_read_demo.google_metadata import (
+ DriveRepresentation, GoogleDenied, GoogleUnavailable, validate_metadata,
+)
+
+
+class GoogleMetadataReferenceTests(unittest.TestCase):
+ def test_metadata_reference_fixture_is_current(self) -> None:
+ command = ReadCommand("google-drive-document-read/v1", "atlas", "instance", "task",
+ "a" * 64, "connection", 1, "file-A", "text", 1024)
+ command.validate()
+ base = {"id": "file-A", "mimeType": DriveRepresentation.PLAIN_TEXT.value,
+ "capabilities": {"canDownload": True}}
+ cases = []
+ declarations = [
+ ("plain-no-size", {}, "plain-text"),
+ ("size-at-limit", {"size": "1024"}, "plain-text"),
+ ("size-integer", {"size": 1024}, "plain-text"),
+ ("size-zero", {"size": "0"}, "plain-text"),
+ ("size-leading-zero", {"size": "01024"}, "plain-text"),
+ ("size-unicode-decimal", {"size": "١٠٢٤"}, "plain-text"),
+ ("size-null", {"size": None}, "plain-text"),
+ ("size-over-limit", {"size": "1025"}, "denied"),
+ ("size-negative", {"size": -1}, "denied"),
+ ("size-float", {"size": 1.0}, "denied"),
+ ("size-boolean", {"size": True}, "denied"),
+ ("size-plus", {"size": "+1"}, "denied"),
+ ("size-decimal-overflow", {"size": "9" * 5000}, "denied"),
+ ("trashed", {"trashed": True}, "denied"),
+ ("trashed-false", {"trashed": False}, "plain-text"),
+ ("trashed-null", {"trashed": None}, "plain-text"),
+ ("trashed-nonboolean", {"trashed": "true"}, "plain-text"),
+ ("different-resource", {"id": "file-B"}, "denied"),
+ ("numeric-resource", {"id": 1}, "denied"),
+ ("download-denied", {"capabilities": {"canDownload": False}}, "denied"),
+ ("download-number", {"capabilities": {"canDownload": 1}}, "denied"),
+ ("download-absent", {"capabilities": {}}, "denied"),
+ ("capabilities-null", {"capabilities": None}, "denied"),
+ ("document", {"mimeType": DriveRepresentation.GOOGLE_DOCUMENT.value}, "google-document"),
+ ("document-size-not-content-bound", {"mimeType": DriveRepresentation.GOOGLE_DOCUMENT.value, "size": "1025"}, "google-document"),
+ ("unsupported-mime", {"mimeType": "application/pdf"}, "denied"),
+ ("mime-null", {"mimeType": None}, "denied"),
+ ("mime-array", {"mimeType": []}, "denied"),
+ ]
+ encoded = [(name, json.dumps({**base, **change}, ensure_ascii=False).encode(), expected)
+ for name, change, expected in declarations]
+ encoded.extend([
+ ("nonobject", b"[]", "denied"),
+ ("missing-identity", b'{}', "denied"),
+ ("invalid-json", b"{", "unavailable"),
+ ("invalid-utf8", b"\xff", "unavailable"),
+ ("utf16-reference", json.dumps(base).encode("utf-16-le"), "plain-text"),
+ ("metadata-limit", json.dumps(base).encode().ljust(MAX_METADATA_BYTES, b" "), "plain-text"),
+ ("metadata-limit-plus-one", json.dumps(base).encode().ljust(MAX_METADATA_BYTES+1, b" "), "unavailable"),
+ ("duplicate-last-wins-reference", b'{"id":"file-B","id":"file-A","mimeType":"text/plain","capabilities":{"canDownload":true}}', "plain-text"),
+ ])
+ for name, body, expected in encoded:
+ with self.subTest(name=name):
+ try:
+ result = validate_metadata(body, command)
+ actual = "google-document" if result is DriveRepresentation.GOOGLE_DOCUMENT else "plain-text"
+ except GoogleDenied:
+ actual = "denied"
+ except GoogleUnavailable:
+ actual = "unavailable"
+ self.assertEqual(actual, expected)
+ cases.append({"case": name, "metadata_hex": body.hex(), "reference_result": actual,
+ "native_stricter_decode_required": name in {"duplicate-last-wins-reference", "utf16-reference", "trashed-nonboolean", "size-unicode-decimal"}})
+ artifact = json.dumps({"schema": "auths.google-drive-metadata-reference/1",
+ "scope": "pure concrete AP070 metadata reference; no transport, authority, state, native equivalence or live-provider claim",
+ "file_id": command.file_id, "maximum_content_bytes": command.max_bytes,
+ "hardening": ["Native bounded UTF-8 JSON must reject duplicate members, alternate encodings, nonboolean trash flags and non-ASCII decimal sizes even when this reference accepts them.",
+ "Non-string MIME values and oversized decimal conversions produce typed denial instead of Python runtime errors.",
+ "Status/media/transport limits and post-metadata freshness remain separate prerequisites."],
+ "cases": cases}, indent=2, ensure_ascii=False).encode()+b"\n"
+ destination = Path(__file__).resolve().parents[2]/"contracts"/"examples"/"google-metadata-reference.json"
+ if os.environ.get("AUTHS_UPDATE_FIXTURES"):
+ destination.write_bytes(artifact)
+ self.assertEqual(destination.read_bytes(), artifact)
+
+
+if __name__ == "__main__":
+ unittest.main()
diff --git a/demos/google-employee-delegation/contracts/examples/google-metadata-reference.json b/demos/google-employee-delegation/contracts/examples/google-metadata-reference.json
new file mode 100644
index 000000000..73af20f34
--- /dev/null
+++ b/demos/google-employee-delegation/contracts/examples/google-metadata-reference.json
@@ -0,0 +1,229 @@
+{
+ "schema": "auths.google-drive-metadata-reference/1",
+ "scope": "pure concrete AP070 metadata reference; no transport, authority, state, native equivalence or live-provider claim",
+ "file_id": "file-A",
+ "maximum_content_bytes": 1024,
+ "hardening": [
+ "Native bounded UTF-8 JSON must reject duplicate members, alternate encodings, nonboolean trash flags and non-ASCII decimal sizes even when this reference accepts them.",
+ "Non-string MIME values and oversized decimal conversions produce typed denial instead of Python runtime errors.",
+ "Status/media/transport limits and post-metadata freshness remain separate prerequisites."
+ ],
+ "cases": [
+ {
+ "case": "plain-no-size",
+ "metadata_hex": "7b226964223a202266696c652d41222c20226d696d6554797065223a2022746578742f706c61696e222c20226361706162696c6974696573223a207b2263616e446f776e6c6f6164223a20747275657d7d",
+ "reference_result": "plain-text",
+ "native_stricter_decode_required": false
+ },
+ {
+ "case": "size-at-limit",
+ "metadata_hex": "7b226964223a202266696c652d41222c20226d696d6554797065223a2022746578742f706c61696e222c20226361706162696c6974696573223a207b2263616e446f776e6c6f6164223a20747275657d2c202273697a65223a202231303234227d",
+ "reference_result": "plain-text",
+ "native_stricter_decode_required": false
+ },
+ {
+ "case": "size-integer",
+ "metadata_hex": "7b226964223a202266696c652d41222c20226d696d6554797065223a2022746578742f706c61696e222c20226361706162696c6974696573223a207b2263616e446f776e6c6f6164223a20747275657d2c202273697a65223a20313032347d",
+ "reference_result": "plain-text",
+ "native_stricter_decode_required": false
+ },
+ {
+ "case": "size-zero",
+ "metadata_hex": "7b226964223a202266696c652d41222c20226d696d6554797065223a2022746578742f706c61696e222c20226361706162696c6974696573223a207b2263616e446f776e6c6f6164223a20747275657d2c202273697a65223a202230227d",
+ "reference_result": "plain-text",
+ "native_stricter_decode_required": false
+ },
+ {
+ "case": "size-leading-zero",
+ "metadata_hex": "7b226964223a202266696c652d41222c20226d696d6554797065223a2022746578742f706c61696e222c20226361706162696c6974696573223a207b2263616e446f776e6c6f6164223a20747275657d2c202273697a65223a20223031303234227d",
+ "reference_result": "plain-text",
+ "native_stricter_decode_required": false
+ },
+ {
+ "case": "size-unicode-decimal",
+ "metadata_hex": "7b226964223a202266696c652d41222c20226d696d6554797065223a2022746578742f706c61696e222c20226361706162696c6974696573223a207b2263616e446f776e6c6f6164223a20747275657d2c202273697a65223a2022d9a1d9a0d9a2d9a4227d",
+ "reference_result": "plain-text",
+ "native_stricter_decode_required": true
+ },
+ {
+ "case": "size-null",
+ "metadata_hex": "7b226964223a202266696c652d41222c20226d696d6554797065223a2022746578742f706c61696e222c20226361706162696c6974696573223a207b2263616e446f776e6c6f6164223a20747275657d2c202273697a65223a206e756c6c7d",
+ "reference_result": "plain-text",
+ "native_stricter_decode_required": false
+ },
+ {
+ "case": "size-over-limit",
+ "metadata_hex": "7b226964223a202266696c652d41222c20226d696d6554797065223a2022746578742f706c61696e222c20226361706162696c6974696573223a207b2263616e446f776e6c6f6164223a20747275657d2c202273697a65223a202231303235227d",
+ "reference_result": "denied",
+ "native_stricter_decode_required": false
+ },
+ {
+ "case": "size-negative",
+ "metadata_hex": "7b226964223a202266696c652d41222c20226d696d6554797065223a2022746578742f706c61696e222c20226361706162696c6974696573223a207b2263616e446f776e6c6f6164223a20747275657d2c202273697a65223a202d317d",
+ "reference_result": "denied",
+ "native_stricter_decode_required": false
+ },
+ {
+ "case": "size-float",
+ "metadata_hex": "7b226964223a202266696c652d41222c20226d696d6554797065223a2022746578742f706c61696e222c20226361706162696c6974696573223a207b2263616e446f776e6c6f6164223a20747275657d2c202273697a65223a20312e307d",
+ "reference_result": "denied",
+ "native_stricter_decode_required": false
+ },
+ {
+ "case": "size-boolean",
+ "metadata_hex": "7b226964223a202266696c652d41222c20226d696d6554797065223a2022746578742f706c61696e222c20226361706162696c6974696573223a207b2263616e446f776e6c6f6164223a20747275657d2c202273697a65223a20747275657d",
+ "reference_result": "denied",
+ "native_stricter_decode_required": false
+ },
+ {
+ "case": "size-plus",
+ "metadata_hex": "7b226964223a202266696c652d41222c20226d696d6554797065223a2022746578742f706c61696e222c20226361706162696c6974696573223a207b2263616e446f776e6c6f6164223a20747275657d2c202273697a65223a20222b31227d",
+ "reference_result": "denied",
+ "native_stricter_decode_required": false
+ },
+ {
+ "case": "size-decimal-overflow",
+ "metadata_hex": "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",
+ "reference_result": "denied",
+ "native_stricter_decode_required": false
+ },
+ {
+ "case": "trashed",
+ "metadata_hex": "7b226964223a202266696c652d41222c20226d696d6554797065223a2022746578742f706c61696e222c20226361706162696c6974696573223a207b2263616e446f776e6c6f6164223a20747275657d2c202274726173686564223a20747275657d",
+ "reference_result": "denied",
+ "native_stricter_decode_required": false
+ },
+ {
+ "case": "trashed-false",
+ "metadata_hex": "7b226964223a202266696c652d41222c20226d696d6554797065223a2022746578742f706c61696e222c20226361706162696c6974696573223a207b2263616e446f776e6c6f6164223a20747275657d2c202274726173686564223a2066616c73657d",
+ "reference_result": "plain-text",
+ "native_stricter_decode_required": false
+ },
+ {
+ "case": "trashed-null",
+ "metadata_hex": "7b226964223a202266696c652d41222c20226d696d6554797065223a2022746578742f706c61696e222c20226361706162696c6974696573223a207b2263616e446f776e6c6f6164223a20747275657d2c202274726173686564223a206e756c6c7d",
+ "reference_result": "plain-text",
+ "native_stricter_decode_required": false
+ },
+ {
+ "case": "trashed-nonboolean",
+ "metadata_hex": "7b226964223a202266696c652d41222c20226d696d6554797065223a2022746578742f706c61696e222c20226361706162696c6974696573223a207b2263616e446f776e6c6f6164223a20747275657d2c202274726173686564223a202274727565227d",
+ "reference_result": "plain-text",
+ "native_stricter_decode_required": true
+ },
+ {
+ "case": "different-resource",
+ "metadata_hex": "7b226964223a202266696c652d42222c20226d696d6554797065223a2022746578742f706c61696e222c20226361706162696c6974696573223a207b2263616e446f776e6c6f6164223a20747275657d7d",
+ "reference_result": "denied",
+ "native_stricter_decode_required": false
+ },
+ {
+ "case": "numeric-resource",
+ "metadata_hex": "7b226964223a20312c20226d696d6554797065223a2022746578742f706c61696e222c20226361706162696c6974696573223a207b2263616e446f776e6c6f6164223a20747275657d7d",
+ "reference_result": "denied",
+ "native_stricter_decode_required": false
+ },
+ {
+ "case": "download-denied",
+ "metadata_hex": "7b226964223a202266696c652d41222c20226d696d6554797065223a2022746578742f706c61696e222c20226361706162696c6974696573223a207b2263616e446f776e6c6f6164223a2066616c73657d7d",
+ "reference_result": "denied",
+ "native_stricter_decode_required": false
+ },
+ {
+ "case": "download-number",
+ "metadata_hex": "7b226964223a202266696c652d41222c20226d696d6554797065223a2022746578742f706c61696e222c20226361706162696c6974696573223a207b2263616e446f776e6c6f6164223a20317d7d",
+ "reference_result": "denied",
+ "native_stricter_decode_required": false
+ },
+ {
+ "case": "download-absent",
+ "metadata_hex": "7b226964223a202266696c652d41222c20226d696d6554797065223a2022746578742f706c61696e222c20226361706162696c6974696573223a207b7d7d",
+ "reference_result": "denied",
+ "native_stricter_decode_required": false
+ },
+ {
+ "case": "capabilities-null",
+ "metadata_hex": "7b226964223a202266696c652d41222c20226d696d6554797065223a2022746578742f706c61696e222c20226361706162696c6974696573223a206e756c6c7d",
+ "reference_result": "denied",
+ "native_stricter_decode_required": false
+ },
+ {
+ "case": "document",
+ "metadata_hex": "7b226964223a202266696c652d41222c20226d696d6554797065223a20226170706c69636174696f6e2f766e642e676f6f676c652d617070732e646f63756d656e74222c20226361706162696c6974696573223a207b2263616e446f776e6c6f6164223a20747275657d7d",
+ "reference_result": "google-document",
+ "native_stricter_decode_required": false
+ },
+ {
+ "case": "document-size-not-content-bound",
+ "metadata_hex": "7b226964223a202266696c652d41222c20226d696d6554797065223a20226170706c69636174696f6e2f766e642e676f6f676c652d617070732e646f63756d656e74222c20226361706162696c6974696573223a207b2263616e446f776e6c6f6164223a20747275657d2c202273697a65223a202231303235227d",
+ "reference_result": "google-document",
+ "native_stricter_decode_required": false
+ },
+ {
+ "case": "unsupported-mime",
+ "metadata_hex": "7b226964223a202266696c652d41222c20226d696d6554797065223a20226170706c69636174696f6e2f706466222c20226361706162696c6974696573223a207b2263616e446f776e6c6f6164223a20747275657d7d",
+ "reference_result": "denied",
+ "native_stricter_decode_required": false
+ },
+ {
+ "case": "mime-null",
+ "metadata_hex": "7b226964223a202266696c652d41222c20226d696d6554797065223a206e756c6c2c20226361706162696c6974696573223a207b2263616e446f776e6c6f6164223a20747275657d7d",
+ "reference_result": "denied",
+ "native_stricter_decode_required": false
+ },
+ {
+ "case": "mime-array",
+ "metadata_hex": "7b226964223a202266696c652d41222c20226d696d6554797065223a205b5d2c20226361706162696c6974696573223a207b2263616e446f776e6c6f6164223a20747275657d7d",
+ "reference_result": "denied",
+ "native_stricter_decode_required": false
+ },
+ {
+ "case": "nonobject",
+ "metadata_hex": "5b5d",
+ "reference_result": "denied",
+ "native_stricter_decode_required": false
+ },
+ {
+ "case": "missing-identity",
+ "metadata_hex": "7b7d",
+ "reference_result": "denied",
+ "native_stricter_decode_required": false
+ },
+ {
+ "case": "invalid-json",
+ "metadata_hex": "7b",
+ "reference_result": "unavailable",
+ "native_stricter_decode_required": false
+ },
+ {
+ "case": "invalid-utf8",
+ "metadata_hex": "ff",
+ "reference_result": "unavailable",
+ "native_stricter_decode_required": false
+ },
+ {
+ "case": "utf16-reference",
+ "metadata_hex": "7b0022006900640022003a0020002200660069006c0065002d00410022002c00200022006d0069006d006500540079007000650022003a002000220074006500780074002f0070006c00610069006e0022002c00200022006300610070006100620069006c006900740069006500730022003a0020007b002200630061006e0044006f0077006e006c006f006100640022003a00200074007200750065007d007d00",
+ "reference_result": "plain-text",
+ "native_stricter_decode_required": true
+ },
+ {
+ "case": "metadata-limit",
+ "metadata_hex": "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",
+ "reference_result": "plain-text",
+ "native_stricter_decode_required": false
+ },
+ {
+ "case": "metadata-limit-plus-one",
+ "metadata_hex": "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",
+ "reference_result": "unavailable",
+ "native_stricter_decode_required": false
+ },
+ {
+ "case": "duplicate-last-wins-reference",
+ "metadata_hex": "7b226964223a2266696c652d42222c226964223a2266696c652d41222c226d696d6554797065223a22746578742f706c61696e222c226361706162696c6974696573223a7b2263616e446f776e6c6f6164223a747275657d7d",
+ "reference_result": "plain-text",
+ "native_stricter_decode_required": true
+ }
+ ]
+}
diff --git a/demos/stripe-gateway/README.md b/demos/stripe-gateway/README.md
index a9a65727c..d69441350 100644
--- a/demos/stripe-gateway/README.md
+++ b/demos/stripe-gateway/README.md
@@ -1,14 +1,15 @@
# Stripe refunds through the general gateway
The maintained Stripe demo uses a data-only platform-refund recipe and installed
-Auths SDKs. The app holds its own scoped signing identity, submits proof/action
+Auths SDKs. The app holds its own scoped signing identity, submits proof/action/exact-approval
bytes over the gateway application socket, and has no Stripe key or private
gateway state. No coupled Stripe vertical service executes this demo.
The web workbench and CLI show the actual gateway result. Verification,
admission, HTTP response and linked provider read-back are separate stages.
-The evidence page exports the native audit-bundle format with available signed
-outcome/pre-entry observations; its display and history are explicitly unsigned.
+The evidence page exports a submission inventory with redacted signed public
+outcomes; its display and history are explicitly unsigned. Full audit facts and
+protected pre-entry observations are available only through the operator export.
Every closed fixture action now traverses the public native approval protocol:
the app constructs an exact typed proposal, two of three operator-held manager
@@ -19,6 +20,32 @@ The workbench shows the native review fields and includes insufficient-quorum
and duplicate-response cases. This local simulator is integration evidence, not
a human approval inbox: simulator state or a UI click never authorizes an action.
+The operator additionally requests the complete native exact-effect review and
+signs its canonical request. Simulation supplies this signature from a separate
+operator-only ephemeral development key. The app forwards the approval unchanged;
+it holds neither that key nor the private actor used for candidate/recovery reads.
+Credential-checking installs stage a candidate, acquire separately authorized
+checks, and publish from retained evidence before peers join. Adding the software
+observer requires explicit native configuration review and CAS activation.
+
+A write returns `response-recorded` when its HTTP response is retained. Choose
+**Separately approve and observe the exact operation** to authorize a distinct
+read. Its native report separates original-write facts, read charges and evidence
+disposition. Replaying a mutation returns retained facts without a read or write;
+unknown outcomes never cause an automatic read, resend or refund.
+
+Live mode never uses the scripted exact-effect reviewer. Startup and each new
+recovery read pause for at most ten minutes on an owner-only consent FIFO. In a
+separate trusted operator shell, use `docker exec CONTAINER cat REVIEW_PATH` to
+read the **entire** `literal-review.txt` and its `literal-review.json` signing
+request at the private paths printed by the operator. After
+checking its exact destination, units, consequences and signing request, use
+`docker exec CONTAINER sh -c 'printf "approve OPERATION\n" > CONSENT_PATH'` with
+the exact operation and consent path printed there. The application UID cannot
+read or write this private channel. Missing or different consent fails closed.
+This is local Docker-operator development custody, not a remote human approval
+inbox, independent-human trial or production signing-custody claim.
+
## Source-free Docker quick start
Download `stripe-gateway-demo.tgz` and its checksum from the **Stripe gateway demo**
@@ -40,8 +67,9 @@ Open . Start with **Exact refund**, then replay, approval
boundary, altered proof/action, two-gateway race and budget experiments. Each choice has
one fixed operation and payment. Run the count/sum positive before its overflow.
The separate operator starts two native gateway processes for the race and a
-third installation for the second challenge. Each installation checks its own
-credential and account; all three share the operation/budget store. The web app runs
+third joined installation. All three pin the same connection, trust and
+configuration and share the operation/budget store. The fresh-challenge
+experiment deliberately substitutes an uninstalled challenge and is refused. The web app runs
as UID 62004. Gateway state and root/observer keys remain root-private.
Default mode is clearly labeled **simulation**. A counting Stripe double measures
@@ -167,7 +195,7 @@ production claim. Never rename its evidence as AWS/PostgreSQL evidence.
The workflow builds a digest-listed handoff, installs its actual SDK artifacts
outside the checkout and runs the native simulation plus a real browser.
`selftest.py` verifies independent wire mapping/read-back, pre-custody refusals,
-post-lease guards, count/sum boundaries, two-process race, fresh-challenge replay,
+post-lease guards, count/sum boundaries, two-process race, fresh-challenge refusal,
process restart, actual response drop after effect, UID isolation, doctor and
offline audit. Provider traffic is counted independently of the app display.
Reports and desktop/mobile screenshots are retained as artifacts.
@@ -179,3 +207,18 @@ not disable that gate or provision shared environment credentials. The Stripe
family's current report contributes only one of the two required live providers.
Human review and actual GitHub-release installation are separate acceptance gates.
See [AP-SPEC-067](../../docs/specs/0067-stripe-general-gateway-demo.md).
+
+The application export contains public history, not a completed private audit. In
+a trusted operator shell, while the gateway is running, complete it with native
+journal evidence in a new owner-private file:
+
+```sh
+docker exec auths-stripe-gateway-demo /opt/demo/bin/auths-gateway-simulation \
+ audit-export --state-dir /run/demo/gateway0 \
+ --bundle /run/demo/app/audit.json --out /run/demo/private/audit-bundle.json
+```
+
+Run the offline auditor against that private bundle with independently pinned
+trust and observer values. Keep the bundle and detailed report in the operator
+boundary; public output does not contain protected facts. The export is read-only,
+never overwrites an existing file, and grants no provider entry or retry.
diff --git a/demos/stripe-gateway/app.py b/demos/stripe-gateway/app.py
index bf02e990b..40ed6c20d 100644
--- a/demos/stripe-gateway/app.py
+++ b/demos/stripe-gateway/app.py
@@ -1,5 +1,5 @@
#!/usr/bin/env python3
-"""Application: public SDK authoring and proof/action-only gateway submission."""
+"""Application: public SDK authoring and proof/action/exact-approval gateway submission."""
import argparse
import asyncio
from dataclasses import asdict
@@ -7,6 +7,8 @@
import os
from pathlib import Path
import secrets
+import socket
+import struct
import threading
import time
from http.server import BaseHTTPRequestHandler, ThreadingHTTPServer
@@ -16,7 +18,7 @@
from auths.authoring import (GrantEvidence, collect_approvals, propose_mcp_approval,
sign_approval_action)
from auths.gateway import (GatewayClient, GatewayEndpoint, GatewayProtocolError,
- GatewaySignedObservation, GatewayPreEntryObservations)
+ GatewaySignedObservation)
from auths.self_hosted import load_signer_file
from demo_common import Refusal, b64, canonical, decode, read, require, unb64, write
@@ -27,7 +29,7 @@
'altered-action', 'forged-proof', 'expired-action', 'race', 'lost-response',
'count-positive', 'count-overflow', 'sum-positive', 'sum-overflow',
'approval-insufficient')
-EXPERIMENTS = LABELS + ('replay', 'fresh-replay', 'observe', 'typescript',
+EXPERIMENTS = tuple('recover-' + label for label in LABELS) + LABELS + ('replay', 'fresh-replay', 'observe', 'typescript',
'approval-duplicate')
@@ -73,7 +75,7 @@ def summary(self):
'platform': self.plan['platform'], 'expires_at': self.plan['expires_at'],
'relative_limit': '50% of independently read amount_received',
'grant_limits': self.plan['limits'], 'experiments': list(EXPERIMENTS),
- 'approval': self.plan['approval'],
+ 'approval': self.plan['approval'], 'effect_review': self.plan['effect_review'],
'sdk_version': importlib.metadata.version('auths'),
'proposals': {label: case['arguments'] for label, case in self.plan['cases'].items()},
'records': self.records[-64:]}
@@ -133,6 +135,7 @@ async def packet(self, label, *, fresh=False):
async def exchange(self, packet, host=0, *, typescript=False):
proof, action, operation = packet
+ approval = unb64(self.plan['effect_approvals'][operation])
if typescript:
completed = await asyncio.create_subprocess_exec(
'/usr/local/bin/node', str(HERE / 'submit.mjs'),
@@ -141,12 +144,13 @@ async def exchange(self, packet, host=0, *, typescript=False):
stderr=asyncio.subprocess.DEVNULL,
env={'PATH': '/usr/local/bin:/usr/bin:/bin'})
output, _ = await asyncio.wait_for(completed.communicate(
- canonical({'proof_b64': b64(proof), 'action_b64': b64(action)})), 60)
+ canonical({'proof_b64': b64(proof), 'action_b64': b64(action),
+ 'approval_b64': b64(approval)})), 60)
require(completed.returncode == 0 and len(output) <= 65536, 'demo.typescript-unavailable')
result = decode(output)
else:
try:
- result = asdict(await self.clients[host].submit(proof=proof, action=action))
+ result = asdict(await self.clients[host].submit(proof=proof, action=action, approval=approval))
except GatewayProtocolError:
# The IPC request may have entered. Never retry automatically.
result = {'outcome': 'transport-unconfirmed', 'code': 'demo.gateway-unavailable'}
@@ -164,9 +168,6 @@ async def remember(self, packet, record):
observed = await self.clients[0].observe_outcome(operation)
if isinstance(observed, GatewaySignedObservation):
entry['outcome_b64'] = b64(observed.observation)
- pre = await self.clients[0].observe_pre_entry(operation)
- if isinstance(pre, GatewayPreEntryObservations):
- entry['pre_entry_b64'] = [b64(item) for item in pre.observations]
except GatewayProtocolError:
pass # Missing signed evidence stays missing; audit must report it.
self.entries[operation] = entry
@@ -176,8 +177,56 @@ async def remember(self, packet, record):
self.approval_responses.append({'operation_id': operation, 'response': response})
recorded.add((operation, response))
+ async def recover(self, source):
+ """Request independent operator consent, then submit its native read proof."""
+ require(source in self.plan['cases'], 'demo.unknown-recovery-source')
+ def request_review():
+ with socket.socket(socket.AF_UNIX, socket.SOCK_STREAM) as channel:
+ channel.settimeout(660 if self.plan['mode'] == 'live' else 90)
+ channel.connect(str(self.sockets[0].parent / 'review.sock'))
+ payload = canonical({'source': source})
+ channel.sendall(struct.pack('>I', len(payload)) + payload)
+ def receive(length):
+ output = b''
+ while len(output) < length:
+ block = channel.recv(length - len(output))
+ require(bool(block), 'demo.review-response-incomplete')
+ output += block
+ return output
+ length = struct.unpack('>I', receive(4))[0]
+ require(0 < length <= 4 * 1024 * 1024, 'demo.review-response-bound')
+ return decode(receive(length))
+ approved = await asyncio.to_thread(request_review)
+ if set(approved) == {'code'}:
+ return {'operation_id': source, 'language': 'python', 'result': {
+ 'outcome': 'not-submitted', 'code': 'demo.operator-recovery-review-refused'}}
+ require(set(approved) == {'proof', 'action', 'approval', 'literal_review'},
+ 'demo.operator-review-unavailable')
+ raw = await self.clients[0].recover(source, proof=unb64(approved['proof']),
+ action=unb64(approved['action']), approval=unb64(approved['approval']))
+ report = decode(raw)
+ require(report.get('outcome') in ('recovery-report', 'refused'), 'demo.recovery-refused')
+ # The native report is retained unchanged. The UI label does not infer
+ # an observation or causation from a provider response.
+ return {'operation_id': source, 'language': 'python', 'result': {
+ **report, 'literal_review': approved['literal_review']}}
+
async def execute(self, experiment):
require(experiment in EXPERIMENTS, 'demo.unknown-experiment')
+ if experiment.startswith('recover-'):
+ source = experiment[len('recover-'):]
+ source = {'loss': 'lost-response', 'competing': 'competing-writer'}.get(source, source)
+ record = await self.recover(source)
+ result = {'schema': 'auths.stripe-gateway-demo-result/1',
+ 'experiment': experiment, 'action': self.plan['cases'][source]['arguments'],
+ 'records': [record], 'decided_by': ('operator-review'
+ if record['result']['outcome'] == 'not-submitted' else 'gateway'),
+ 'production_qualified': False,
+ 'display_signature': 'unsigned', 'recorded_at': int(time.time())}
+ require(len(self.records) < 256, 'demo.session-full')
+ self.records.append(result)
+ write(self.directory / 'results.json', canonical(self.records))
+ return result
if experiment == 'approval-duplicate':
records = [self.approval_duplicate()]
result = {'schema': 'auths.stripe-gateway-demo-result/1', 'experiment': experiment,
@@ -222,7 +271,7 @@ async def execute(self, experiment):
return result
def audit(self):
- return {'schema': 'auths.gateway-audit-bundle/2',
+ return {'schema': 'auths.gateway-audit-bundle/3',
'recipe_b64': b64(read(HERE / 'recipe.json')),
'profile_lock_b64': b64(read(HERE / 'profile.lock.json')),
'trusted_context_b64': self.plan['contexts'][0],
diff --git a/demos/stripe-gateway/contract.json b/demos/stripe-gateway/contract.json
index c0d37f8d5..6f1d39898 100644
--- a/demos/stripe-gateway/contract.json
+++ b/demos/stripe-gateway/contract.json
@@ -2,7 +2,7 @@
"schema": "auths.gateway-demo-contract/1",
"platform": "stripe",
"family": "stripe-platform-refund-v1",
- "application_boundary": "installed public GatewayClient; proof and action only",
+ "application_boundary": "installed public GatewayClient; proof, action and native exact-effect approval",
"recipe_source": "qualification/simulation/live/stripe-platform/recipe.json",
"profile_source": "qualification/simulation/live/stripe-platform/profile.toml",
"qualification_workflow": ".github/workflows/recipe-qualification.yml",
diff --git a/demos/stripe-gateway/operate.py b/demos/stripe-gateway/operate.py
index 0341dbbb5..ff0f13d96 100644
--- a/demos/stripe-gateway/operate.py
+++ b/demos/stripe-gateway/operate.py
@@ -2,22 +2,31 @@
"""Separate operator: disposable resources, independent trust, gateway installation."""
import argparse
import asyncio
+from dataclasses import dataclass
import os
import re
from pathlib import Path
import secrets
import signal
+import socketserver
+import struct
+import threading
import subprocess
import sys
import time
+from typing import Optional
-from auths.authoring import approval_requests, approve, open_approval_request, propose_mcp_approval
-from auths.self_hosted import (ApprovalRequirement, ApproverAnchor, AssurancePolicy,
+from auths.adapters.custody import PublicControlEvidence
+from auths.authoring import (GrantEvidence, approval_requests, approve, collect_approvals,
+ open_approval_request, propose_mcp_approval, sign_approval_action)
+from auths.testkit import development_ed25519_identity_key
+from auths.self_hosted import (ExactMcpTool, StringField, IntegerField, OptionalField,
+ ApprovalRequirement, ApproverAnchor, AssurancePolicy,
AssuranceRequirement, Permission, Profile, TrustAnchor,
TrustedContextRequest, author_root_grant, compile_trusted_context,
load_signer_file)
from demo_common import Refusal, b64, canonical, decode, digest, read, require, unb64, write
-from app import LABELS, approval_target
+from app import Application, LABELS, approval_target
HERE = Path(__file__).resolve().parent
HOME = Path('/run/demo')
@@ -27,6 +36,7 @@
APP_UID, APP_GID = 62004, 62005
ENVIRONMENT = {'PATH': '/opt/consumer/bin:/usr/local/bin:/usr/bin:/bin', 'PYTHONNOUSERSITE': '1'}
MANAGERS = ('manager-a', 'manager-b', 'manager-c')
+READ_TOOLS = ('auths_gateway_candidate_check_v1', 'auths_gateway_recovery_read_v1')
APPROVALS_REQUIRED = 2
ACTION_VALIDITY_SECONDS = 1800
@@ -73,7 +83,7 @@ async def authority(binary, resources, directory, mode):
app.mkdir(mode=0o700)
require(private.is_dir() and not private.is_symlink() and private.stat().st_uid == 0
and private.stat().st_mode & 0o777 == 0o700, 'demo.operator-private-directory-required')
- for name, parent in [('root', private), ('actor', app),
+ for name, parent in [('root', private), ('reader', private), ('actor', app),
*((name, private) for name in MANAGERS)]:
write(parent / (name + '.seed'), secrets.token_bytes(32), new=True)
write(parent / (name + '.json'), canonical({'schema': 'auths.approval-signer/1',
@@ -86,6 +96,8 @@ async def authority(binary, resources, directory, mode):
now, expires = int(time.time()), int(time.time()) + 3600
audience = 'mcp://stripe-platform-refunds'
permission = Permission('tools/call', audience + '/tools/create_refund_v1')
+ permissions = (permission,) + tuple(Permission('tools/call', audience + '/tools/' + tool)
+ for tool in READ_TOOLS)
policy = AssurancePolicy('raw-key-baseline', tuple(AssuranceRequirement(*row) for row in [
('root', 'every', 'self-certifying-identifier'),
('actor', 'every', 'self-certifying-identifier'),
@@ -108,7 +120,7 @@ async def authority(binary, resources, directory, mode):
for item in grant.evidence]}
limits[name] = {'count': count, 'sum': capacity, 'ceiling': 2000, 'window_seconds': window}
anchors = [TrustAnchor('operator-root', root.descriptor.principal, ('raw-key-v1',),
- (Profile('auths.mcp', 2),), (permission,), (audience,), (audience,),
+ (Profile('auths.mcp', 2),), permissions, (audience,), (audience,),
now - 3600, expires, 1, 'raw-key-baseline')]
approver_principals = tuple(managers[name].descriptor.principal for name in MANAGERS)
options = {'anchors': anchors, 'assurance': policy,
@@ -118,6 +130,17 @@ async def authority(binary, resources, directory, mode):
now - 3600, expires) for principal in approver_principals),
'approval_requirements': (ApprovalRequirement(approver_principals,
APPROVALS_REQUIRED),)}
+ reader = load_signer_file(private / 'reader.json').signer
+ try:
+ read_grant = await author_root_grant(signer=root, subject=reader.descriptor.principal,
+ profile=Profile('auths.mcp', 2), permissions=permissions[1:], audiences=[audience],
+ not_before=now - 3600, expires_at=expires, remaining_depth=0,
+ assurance_floor='raw-key-baseline', requested_at=now)
+ write(private / 'read-grant.json', canonical({'bytes': b64(read_grant.signed_grant),
+ 'evidence': [{'type': item.evidence_type, 'media_type': item.media_type,
+ 'bytes': b64(item.bytes)} for item in read_grant.evidence]}), new=True)
+ finally:
+ await reader.aclose()
template = compile_trusted_context(**options)
challenges = [secrets.token_bytes(32), secrets.token_bytes(32)]
contexts = [compile_trusted_context(**options,
@@ -192,6 +215,76 @@ async def approval_bundle(label, *, fresh=False):
await signer.aclose()
+@dataclass(frozen=True)
+class CandidateCheck:
+ operation_id: str
+ operator_namespace: str
+ recipe_digest: str
+ candidate_connection: str
+ expected_previous: Optional[str]
+ read_contract: str
+ maximum_response_bytes: int
+
+
+@dataclass(frozen=True)
+class RecoveryRead:
+ operation_id: str
+ operator_namespace: str
+ recipe_digest: str
+ source_effect: str
+ source_operation: str
+ read_contract: str
+ maximum_response_bytes: int
+
+
+def read_contract(call):
+ """Closed transport shape only; the native gateway owns every read binding."""
+ name = call['name']
+ require(name in READ_TOOLS, 'demo.unlisted-read-tool')
+ command_type = CandidateCheck if name == READ_TOOLS[0] else RecoveryRead
+ fields = {key: StringField(min_length=1, max_length=128)
+ for key in command_type.__annotations__}
+ fields['maximum_response_bytes'] = IntegerField(1, 1048576)
+ if command_type is CandidateCheck:
+ fields['expected_previous'] = OptionalField(StringField(min_length=64, max_length=64))
+ contract = ExactMcpTool(service=call['service'], name=name,
+ command_type=command_type, fields=fields)
+ return contract, contract.validate_arguments(call['arguments'])
+
+
+def operator_consent(directory, name):
+ """Bounded owner-only consent for the already saved exact native review.
+
+ The Docker operator reads the complete literal review and signing request
+ and writes one exact token to this private FIFO. The app cannot access it.
+ This is development custody, not an authenticated remote approval inbox.
+ """
+ consent = directory / 'consent'
+ os.mkfifo(consent, mode=0o600)
+ descriptor = os.open(consent, os.O_RDONLY | os.O_NONBLOCK | os.O_NOFOLLOW)
+ try:
+ print('Operator review required: read ' + str(directory / 'literal-review.txt')
+ + ' and its literal-review.json signing request through the trusted container operator shell; then write "approve '
+ + name + '" followed by a newline to ' + str(consent), flush=True)
+ deadline, response = time.monotonic() + 600, b''
+ while time.monotonic() < deadline:
+ require(len(response) < 256, 'demo.operator-review-response-bound')
+ try:
+ part = os.read(descriptor, 256 - len(response))
+ except BlockingIOError:
+ part = b''
+ response += part
+ if b'\n' in response:
+ require(response == ('approve ' + name + '\n').encode(),
+ 'demo.operator-review-declined')
+ return
+ time.sleep(0.05)
+ raise Refusal('demo.operator-review-expired')
+ finally:
+ os.close(descriptor)
+ consent.unlink(missing_ok=True)
+
+
class Operator:
def __init__(self, mode):
require(os.getuid() == 0, 'demo.operator-root-required')
@@ -204,6 +297,7 @@ def __init__(self, mode):
self.sockets.mkdir(mode=0o750)
os.chown(self.sockets, 0, APP_GID)
self.port = None
+ self.effect_reviewer = development_ed25519_identity_key()
def start_gateway(self, host):
socket_path = self.sockets / ('app' + str(host) + '.sock')
@@ -260,35 +354,206 @@ def prepare(self):
plan = asyncio.run(authority(self.binary, resources, HOME, self.mode))
attempts = HOME / 'attempts'
attempts.mkdir(mode=0o700)
+ effect_policy = PRIVATE / 'effect-policy.json'
+ write(effect_policy, gateway(self.binary, 'effect-policy', '--partition', 'stripe-demo',
+ '--epoch', '1', '--approver-public-key', self.effect_reviewer.public_key.hex()), new=True)
for host in range(3):
state = HOME / ('gateway' + str(host))
args = [self.binary, 'install', '--state-dir', state, '--recipe', HERE / 'recipe.json',
'--profile-lock', HERE / 'profile.lock.json', '--trusted-context',
- HOME / 'private' / ('context-' + str(1 if host == 2 else 0) + '.cbor'), '--approve-digest', plan['recipe_digest'],
- '--provider', 'stripe', '--alias', 'demo' + str(host), '--credential-stdin',
- '--attempt-store', attempts, '--account-label', resources['platform']]
- # Each installation checks the account and restricted credential.
- # All three share the operation/budget store; no simulation join
- # may accidentally probe the live provider with a synthetic key.
- if self.port is not None:
- args.extend(['--loopback-provider', self.port])
+ PRIVATE / 'context-0.cbor', '--effect-policy', effect_policy,
+ '--approve-digest', plan['recipe_digest'], '--provider', 'stripe',
+ '--alias', 'demo', '--credential-stdin', '--attempt-store', attempts]
+ args.extend(['--account-label', resources['platform']] if host == 0 else ['--join'])
child(args, stdin=(self.keys['runtime'] + '\n').encode())
+ proposal = state / 'candidate-installation.json'
+ if proposal.exists():
+ call = PRIVATE / 'candidate-call.json'
+ gateway(self.binary, 'effect-candidate-call', '--state-dir', state,
+ '--candidate', proposal, '--operation', 'install-check', '--call-out', call)
+ proof, action = asyncio.run(self.read_packet(decode(read(call)), plan))
+ approved = self.exact_approval(state, 'install-check', proof, action,
+ '--candidate', proposal)
+ require(approved is not None, 'demo.candidate-effect-approval-unavailable')
+ arguments = ['--state-dir', state, '--candidate', proposal, '--proof', approved[0],
+ '--action', approved[1], '--approval', approved[2]]
+ local = ['--loopback-provider', self.port] if self.port is not None else []
+ gateway(self.binary, 'candidate-acquire', *arguments, *local)
+ gateway(self.binary, 'candidate-publish', *arguments)
# A shared software observer exists only in this labeled development
# deployment. It is never a production signing-custody claim.
gateway(self.binary, 'observer-init', '--state-dir', HOME / 'gateway0')
for host in [1, 2]:
write(HOME / ('gateway' + str(host)) / 'observer.seed', read(HOME / 'gateway0' / 'observer.seed'), new=True)
self.observer = decode(gateway(self.binary, 'observer-show', '--state-dir', HOME / 'gateway0'))
+ configuration = gateway(self.binary, 'effect-configuration-review',
+ '--state-dir', HOME / 'gateway0')
+ write(PRIVATE / 'configuration-review.json', configuration, new=True)
+ reviewed = decode(configuration)
+ gateway(self.binary, 'effect-configuration-activate', '--state-dir', HOME / 'gateway0',
+ '--expected-revision', reviewed['previous_revision_hex'],
+ '--expected-candidate', reviewed['candidate_hex'])
write(HOME / 'private' / 'pins.json', canonical({'trusted_context_sha256':
digest(read(HOME / 'private' / 'context-0.cbor')), 'observer': self.observer}), new=True)
for host in range(3):
self.start_gateway(host)
+ self.prepare_exact_approvals(plan)
+ self.start_recovery_reviews(plan)
return plan
+ async def read_packet(self, call, plan):
+ """Managers simulate separately authorizing a native-authored read call."""
+ contract, command = read_contract(call)
+ grant = decode(read(PRIVATE / 'read-grant.json'))
+ reader = load_signer_file(PRIVATE / 'reader.json').signer
+ managers = [load_signer_file(PRIVATE / (name + '.json')).signer for name in MANAGERS]
+ try:
+ proposal = propose_mcp_approval(contract=contract, command=command,
+ required=APPROVALS_REQUIRED, approvers=plan['approval']['principals'],
+ actor=reader.descriptor.principal, actor_grant=unb64(grant['bytes']),
+ challenge=unb64(plan['challenges'][0]), evaluation_time=int(time.time()),
+ validity_seconds=ACTION_VALIDITY_SECONDS)
+ responses = []
+ for signer in managers[:APPROVALS_REQUIRED]:
+ request = next(item for item in approval_requests(proposal)
+ if item.approver == signer.descriptor.principal)
+ responses.append((await approve(open_approval_request(request.text,
+ now=int(time.time())), signer)).text)
+ signed = await sign_approval_action(proposal, reader, grants=[
+ GrantEvidence(unb64(grant['bytes']), tuple(PublicControlEvidence(
+ item['type'], item['media_type'], unb64(item['bytes']))
+ for item in grant['evidence']))])
+ return collect_approvals(proposal, responses).assemble(signed), proposal.action
+ finally:
+ await reader.aclose()
+ for signer in managers:
+ await signer.aclose()
+
+ def exact_approval(self, state, name, proof, action, *mode):
+ """Scripted reviewer over complete native literal output; no human-trial claim."""
+ directory = PRIVATE / name
+ directory.mkdir(mode=0o700)
+ proof_path, action_path = directory / 'proof', directory / 'action'
+ request, approval = directory / 'request', directory / 'approval'
+ write(proof_path, proof, new=True)
+ write(action_path, action, new=True)
+ now = int(time.time())
+ output = child([self.binary, 'effect-review', '--state-dir', state,
+ '--proof', proof_path, '--action', action_path, '--not-before', str(now - 1),
+ '--expires-at', str(now + ACTION_VALIDITY_SECONDS), '--request-out', request,
+ '--binding-out', directory / 'binding',
+ *mode], check=False)
+ if output.returncode != 0:
+ return None
+ review = decode(output.stdout)
+ require(review['schema'] == 'auths.private-effect-signing/2'
+ and review['renderer'] == 'auths.literal-complete-review/1'
+ and type(review['literal_review']) is str and review['literal_review'],
+ 'demo.native-effect-review-required')
+ write(directory / 'literal-review.json', output.stdout, new=True)
+ write(directory / 'literal-review.txt', review['literal_review'].encode('utf-8'), new=True)
+ if self.mode == 'live':
+ operator_consent(directory, name)
+ signature = self.effect_reviewer.sign(bytes.fromhex(review['signing_message_hex']))
+ write(approval, gateway(self.binary, 'effect-approval', '--request', request,
+ '--signature-hex', signature.hex()), new=True)
+ return proof_path, action_path, approval
+
+ def prepare_exact_approvals(self, plan):
+ """Preapprove finite scripted mutations; refusals use an explicitly substituted approval."""
+ application = Application(HOME / 'app', [self.sockets / ('app' + str(host) + '.sock')
+ for host in range(3)])
+ values, reviews = {}, {}
+ for label in LABELS:
+ proof, action, _ = asyncio.run(application.packet(label))
+ result = self.exact_approval(HOME / 'gateway0', 'effect-' + label, proof, action)
+ if result is not None:
+ values[label] = b64(read(result[2]))
+ review = decode(read(result[2].parent / 'literal-review.json'))
+ reviews[label] = review['literal_review']
+ require('exact' in values, 'demo.exact-effect-approval-unavailable')
+ for label in LABELS:
+ # A malformed/unauthorized action cannot be reviewed. Sending an
+ # approval for another effect is an adversarial binding substitution,
+ # never an application-created authority or a bypass.
+ values.setdefault(label, values['exact'])
+ values['exact:fresh'] = values['exact']
+ plan['effect_approvals'] = values
+ plan['effect_review'] = {'mode': ('scripted operator reviewer; not an independent human trial'
+ if self.mode == 'simulation' else 'separate private operator consent; development signing custody'),
+ 'renderer': 'auths.literal-complete-review/1', 'reviews': reviews}
+ write(HOME / 'app/application.json', canonical(plan))
+ for path in (HOME / 'app').iterdir():
+ os.chown(path, APP_UID, APP_GID)
+
+ def prepare_recovery(self, source, sequence, plan):
+ """Produce separate read authority; deriving a call never grants entry."""
+ operation = 'read-' + source + '-' + str(sequence)
+ call = decode(gateway(self.binary, 'effect-recovery-call', '--state-dir', HOME / 'gateway0',
+ '--source-operation', source, '--recovery-operation', operation))
+ proof, action = asyncio.run(self.read_packet(call, plan))
+ approved = self.exact_approval(HOME / 'gateway0', operation, proof, action,
+ '--recovery-of', source)
+ require(approved is not None, 'demo.recovery-review-refused')
+ return {'proof': b64(proof), 'action': b64(action), 'approval': b64(read(approved[2])),
+ 'literal_review': decode(read(approved[2].parent / 'literal-review.json'))['literal_review']}
+
+ def start_recovery_reviews(self, plan):
+ """Finite development reviewer, isolated from the application's UID.
+
+ This is an explicit operator signing channel, not native recovery
+ semantics. Live mode requires private operator consent in exact_approval.
+ """
+ operator = self
+ counts = {}
+
+ class Handler(socketserver.BaseRequestHandler):
+ def handle(self):
+ self.request.settimeout(660 if operator.mode == 'live' else 30)
+ def receive(length):
+ output = b''
+ while len(output) < length:
+ block = self.request.recv(length - len(output))
+ require(bool(block), 'demo.review-request-incomplete')
+ output += block
+ return output
+ try:
+ length = struct.unpack('>I', receive(4))[0]
+ require(0 < length <= 1024, 'demo.review-request-bound')
+ request = decode(receive(length))
+ require(type(request) is dict and set(request) == {'source'}
+ and type(request['source']) is str
+ and request['source'] in plan['cases'], 'demo.review-source-refused')
+ source = request['source']
+ sequence = counts.get(source, 0) + 1
+ require(sequence <= 4, 'demo.review-attempt-limit')
+ counts[source] = sequence
+ response = operator.prepare_recovery(source, sequence, plan)
+ except (Refusal, ValueError, OSError, TimeoutError):
+ response = {'code': 'demo.operator-recovery-review-refused'}
+ raw = canonical(response)
+ require(len(raw) <= 4 * 1024 * 1024, 'demo.review-response-bound')
+ self.request.sendall(struct.pack('>I', len(raw)) + raw)
+
+ path = self.sockets / 'review.sock'
+ server = socketserver.UnixStreamServer(str(path), Handler)
+ os.chown(path, 0, APP_GID)
+ os.chmod(path, 0o660)
+ thread = threading.Thread(target=server.serve_forever, daemon=True)
+ thread.start()
+ self.review_server, self.review_thread = server, thread
+
def application(self, *arguments):
return subprocess.run(['/opt/consumer/bin/python', str(HERE / 'app.py'), *arguments],
env=ENVIRONMENT, capture_output=True, user=APP_UID, group=APP_GID, extra_groups=[], timeout=90)
+ def export_audit(self, destination):
+ # Only the operator can reach native private audit facts. The application
+ # bundle is an untrusted submission inventory with redacted public history.
+ gateway(self.binary, 'audit-export', '--state-dir', HOME / 'gateway0',
+ '--bundle', HOME / 'app/audit.json', '--out', destination)
+ return destination
+
def witnesses(self):
values = []
for host in range(3):
@@ -300,6 +565,10 @@ def witnesses(self):
return values
def stop(self):
+ if hasattr(self, 'review_server'):
+ self.review_server.shutdown()
+ self.review_server.server_close()
+ self.review_thread.join(timeout=5)
for process, log, _ in self.processes:
if process.poll() is None:
process.terminate()
diff --git a/demos/stripe-gateway/recipe.json b/demos/stripe-gateway/recipe.json
index 9f9f90ce1..ba4893b10 100644
--- a/demos/stripe-gateway/recipe.json
+++ b/demos/stripe-gateway/recipe.json
@@ -1,5 +1,5 @@
{
- "schema": "auths.gateway-recipe-source/2",
+ "schema": "auths.gateway-recipe-source/3",
"profile_schema_digest": "d2f5b926a3b6989d13b3a4614372e298442de697d92d5e06330288f7de4e257a",
"service": "stripe-platform-refunds",
"tool": "create_refund_v1",
@@ -167,5 +167,27 @@
"name": "metadata[auths_echo]"
},
"observe": "/metadata/auths_echo"
+ },
+ "review": {
+ "schema": "auths.effect-review-contract/1",
+ "consequences": [
+ "Issue one refund against the identified payment intent for the displayed amount. This is not a recurring charge.",
+ "The provider determines refund settlement timing. A response or receipt does not itself establish settlement."
+ ],
+ "fields": {
+ "amount": {
+ "kind": "integer",
+ "label": "Amount",
+ "unit": "minor units of the declared currency"
+ },
+ "currency": {
+ "kind": "identifier",
+ "label": "Currency"
+ },
+ "payment_intent": {
+ "kind": "identifier",
+ "label": "Payment intent"
+ }
+ }
}
}
diff --git a/demos/stripe-gateway/selftest.py b/demos/stripe-gateway/selftest.py
index d6b8ae374..ab5c0b162 100644
--- a/demos/stripe-gateway/selftest.py
+++ b/demos/stripe-gateway/selftest.py
@@ -1,6 +1,10 @@
"""Installed-client/native-process evidence; expected facts do not come from the UI."""
import base64
+import os
+import tempfile
+import time
from pathlib import Path
+from concurrent.futures import ThreadPoolExecutor
import subprocess
import urllib.request
from urllib.parse import urlencode
@@ -8,6 +12,22 @@
from demo_common import canonical, decode, digest, read, require, write
+def private_action_echo(label, action):
+ """Test-only expected link from the operator's exact native approval input."""
+ from operate import PRIVATE
+ binding = decode(read(PRIVATE / ('effect-' + label) / 'binding'))
+ commitment = binding['action']
+ require(binding['schema'] == 'auths.bound-effect/3'
+ and binding['namespace'] == action['operator_namespace']
+ and binding['operation'] == action['operation_id']
+ and type(commitment) is list and len(commitment) == 32
+ and all(type(value) is int and 0 <= value <= 255 for value in commitment),
+ 'demo.selftest-private-action-binding')
+ return 'auths-e1-' + digest(b'auths.gateway-echo/1\0'
+ + action['operator_namespace'].encode() + b'\0'
+ + action['operation_id'].encode() + b'\0' + bytes(commitment))
+
+
def test_native_refusals_replay_race_and_loss(operator):
"""No double replaces gateway verification, admission, persistence or observation."""
from operate import HOME
@@ -60,12 +80,44 @@ def execute(label):
'demo.selftest-read-guard:' + label)
for label in ['exact', 'boundary']:
result, records, traffic = execute(label)
- require(records[0]['outcome'] == 'observed-by-provider', 'demo.selftest-observation:' + label)
+ require(records[0]['outcome'] == 'response-recorded' and records[0]['status'] == 200,
+ 'demo.selftest-response-is-not-observation:' + label)
+ _, recovery, recovery_traffic = execute('recover-' + label)
+ report = recovery[0]['report']
+ require(recovery[0]['outcome'] == 'recovery-report'
+ and report['source']['stage'] == 'observed-by-provider'
+ and report['source']['provider_linked'] is True
+ and report['acquisition']['stage'] == 'response-recorded'
+ and report['disposition'] == 'recorded'
+ and not any(item['kind'] == 'write' for item in recovery_traffic)
+ and cases[-1]['native_counters']['write_transport_entries'] == 0,
+ 'demo.selftest-independent-native-observation:' + label)
+ inspect_before, inspect_traffic = operator.witnesses(), len(entries())
+ inspected = operator.application('submit', '--experiment', label)
+ require(inspect_before == operator.witnesses() and inspect_traffic == len(entries()),
+ 'demo.selftest-source-inspection-must-not-enter-provider')
+ require(inspected.returncode == 0, 'demo.selftest-source-inspection-refused')
+ records = [item['result'] for item in decode(inspected.stdout)['records']]
+ require(records[0]['outcome'] == 'observed-by-provider', 'demo.selftest-source-observation')
writes = [item for item in traffic if item['kind'] == 'write']
- require(len(writes) == 1 and cases[-1]['native_counters']['write_transport_entries'] == 1,
+ require(len(writes) == 1 and cases[-2]['native_counters']['write_transport_entries'] == 1,
'demo.selftest-write-count:' + label)
action = result['action']
- echo = records[0]['evidence']['echo']
+ evidence = records[0]['evidence']
+ language = decode(inspected.stdout)['records'][0]['language']
+ require(language in ('python', 'typescript'), 'demo.selftest-client-language')
+ commitment_field, time_field = (('evidenceCommitment', 'observedAt')
+ if language == 'typescript' else ('evidence_commitment', 'observed_at'))
+ require(set(evidence) == {'schema', 'channel', commitment_field, time_field}
+ and evidence['schema'] == 'auths.gateway-public-evidence/1'
+ and evidence['channel'] == 'read-back'
+ and type(evidence[commitment_field]) is str
+ and len(evidence[commitment_field]) == 64
+ and all(value in '0123456789abcdef' for value in evidence[commitment_field])
+ and type(evidence[time_field]) is int and 0 <= evidence[time_field] <= 2**53 - 1,
+ 'demo.selftest-public-hiding-evidence')
+ # Private test expectation never returns through the application result.
+ echo = private_action_echo(label, action)
key = 'auths-i1-' + digest(b'auths.gateway-idempotency-key/1\0'
+ action['operator_namespace'].encode() + b'\0' + action['operation_id'].encode())
expected = urlencode([('amount', str(action['amount'])), ('metadata[auths_echo]', echo),
@@ -84,23 +136,42 @@ def execute(label):
refund = decode(refund_bytes)
require(refund['amount'] == action['amount'] and refund['payment_intent'] == action['payment_intent']
and refund['metadata']['auths_echo'] == echo, 'demo.selftest-independent-read-back')
- cases[-1]['independent_read_back'] = {'fixture_api_entries': 1, 'response_sha256': digest(refund_bytes)}
- for label in ['replay', 'fresh-replay', 'typescript']:
+ require(evidence[commitment_field] != digest(refund_bytes),
+ 'demo.selftest-public-raw-evidence-digest')
+ from operate import PRIVATE
+ write(PRIVATE / ('independent-readback-' + label + '.json'), refund_bytes, new=True)
+ cases[-1]['independent_read_back'] = {'fixture_api_entries': 1, 'exact_fields_and_native_link_checked': True}
+ for label in ['replay', 'typescript']:
_, records, traffic = execute(label)
- require(records[0]['outcome'] == 'not-entered' and records[0]['code'] == 'gateway.attempt.replay'
- and not traffic and cases[-1]['native_counters']['credential_lease_calls'] == 0,
- 'demo.selftest-replay:' + label)
+ require(records[0]['outcome'] == 'observed-by-provider'
+ and not traffic and not any(cases[-1]['native_counters'].values()),
+ 'demo.selftest-replay-retains-observation-without-new-entry:' + label)
+ _, records, traffic = execute('fresh-replay')
+ require(records[0]['outcome'] == 'denied' and records[0]['code'] == 'challenge-mismatch'
+ and not traffic and not any(cases[-1]['native_counters'].values()),
+ 'demo.selftest-substituted-challenge-under-installed-trust')
_, records, traffic = execute('race')
- require(sorted(item['outcome'] for item in records) == ['not-entered', 'observed-by-provider']
+ require(any(item['outcome'] == 'response-recorded' for item in records)
+ and all(item['outcome'] in ('reserved', 'unknown', 'response-recorded') for item in records)
and sum(item['kind'] == 'write' for item in traffic) == 1
and cases[-1]['native_counters']['write_transport_entries'] == 1, 'demo.selftest-two-process-race')
+ # A contender can inspect the winner after durable entry but before its
+ # response is recorded, so Unknown is a truthful interim projection. Once
+ # both calls finish, exact replay must converge without any external I/O.
+ race_before, race_entries = operator.witnesses(), len(entries())
+ replayed = operator.application('submit', '--experiment', 'race')
+ require(replayed.returncode == 0
+ and decode(replayed.stdout)['records'][0]['result']['outcome'] == 'response-recorded'
+ and race_before == operator.witnesses() and race_entries == len(entries()),
+ 'demo.selftest-race-final-history-without-new-entry')
operator.restart(0)
_, records, traffic = execute('replay')
- require(records[0]['code'] == 'gateway.attempt.replay' and not traffic, 'demo.selftest-restart-replay')
+ require(records[0]['outcome'] == 'observed-by-provider' and not traffic
+ and not any(cases[-1]['native_counters'].values()), 'demo.selftest-restart-replay')
cases[-1]['case'] = 'restart-replay'
for name in ['count', 'sum']:
_, records, traffic = execute(name + '-positive')
- require(records[0]['outcome'] == 'observed-by-provider'
+ require(records[0]['outcome'] == 'response-recorded'
and sum(item['kind'] == 'write' for item in traffic) == 1, 'demo.selftest-budget-positive')
_, records, traffic = execute(name + '-overflow')
require(records[0]['outcome'] == 'not-entered' and not traffic
@@ -111,14 +182,15 @@ def execute(label):
and cases[-1]['native_counters']['write_transport_entries'] == 1
and any(item['response_dropped_after_effect'] for item in traffic), 'demo.selftest-loss-after-effect')
_, records, traffic = execute('lost-response')
- require(records[0]['outcome'] in ['not-entered', 'unknown']
- and not any(item['kind'] == 'write' for item in traffic), 'demo.selftest-unknown-never-retries')
+ require(records[0]['outcome'] == 'unknown' and not traffic
+ and not any(cases[-1]['native_counters'].values()), 'demo.selftest-unknown-never-refetches-or-retries')
return cases
def test_application_isolation_and_native_audit(operator):
from operate import APP_UID, APP_GID, ENVIRONMENT, HOME, child
- private = [HOME / 'private' / 'root.seed',
+ private = [HOME / 'private' / 'root.seed', HOME / 'private' / 'reader.seed',
+ HOME / 'private' / 'read-grant.json',
*(HOME / 'private' / (name + '.seed') for name in
['manager-a', 'manager-b', 'manager-c']),
HOME / 'gateway0' / 'credentials.cbor',
@@ -143,24 +215,116 @@ def test_application_isolation_and_native_audit(operator):
require(checks['observer-custody']['state'] == 'not-ready',
'demo.selftest-software-observer-not-production-ready')
pins = decode(read(HOME / 'private/pins.json'))
- audit = child([operator.binary, 'audit', '--bundle', HOME / 'app/audit.json',
+ private_bundle = operator.export_audit(HOME / 'private/verified-audit.json')
+ audit = child([operator.binary, 'audit', '--bundle', private_bundle,
'--trusted-context-sha256', pins['trusted_context_sha256'],
'--observer', pins['observer']['observer_anchor']['principal']])
+ private_report = HOME / 'private/verified-audit-report.json'
+ write(private_report, audit.stdout, new=True)
+ for protected in [private_bundle, private_report]:
+ probe = subprocess.run(['/opt/consumer/bin/python', '-c',
+ 'import sys; open(sys.argv[1], "rb").read()', str(protected)],
+ user=APP_UID, group=APP_GID, extra_groups=[], env=ENVIRONMENT, capture_output=True)
+ require(probe.returncode != 0 and not probe.stdout, 'demo.selftest-private-audit-readable-by-app')
+ require(protected.stat().st_mode & 0o777 == 0o600, 'demo.selftest-private-audit-mode')
result = decode(audit.stdout)
require(result['inconsistent'] == 0 and result['unverified'] == 0
and len(result['approval_responses']) >= 2,
'demo.selftest-audit-open')
- return {'doctor': doctor_value, 'audit': result}
+ # Full native report is protected. Publish only aggregate verification
+ # counts; preserve all detailed assertions above in the operator process.
+ summary = {key: result[key] for key in ['verified', 'refused', 'unverified', 'inconsistent']}
+ summary['private_report_retained'] = True
+ return {'doctor': doctor_value, 'audit': summary}
+
+
+def test_exact_operator_consent_binding(operator):
+ """Exercise private consent transport; scripted input is not a human trial."""
+ from operate import HOME, operator_consent
+ from demo_common import Refusal
+ before = operator.witnesses()
+ for accepted in [False, True]:
+ with tempfile.TemporaryDirectory(prefix='consent-test-', dir=HOME / 'private') as temporary:
+ directory = Path(temporary)
+ write(directory / 'literal-review.json',
+ read(HOME / 'private/effect-exact/literal-review.json'), new=True)
+ write(directory / 'literal-review.txt',
+ read(HOME / 'private/effect-exact/literal-review.txt'), new=True)
+ with ThreadPoolExecutor(max_workers=1) as pool:
+ pending = pool.submit(operator_consent, directory, 'exact-consent-test')
+ fifo, deadline = directory / 'consent', time.monotonic() + 5
+ while not fifo.exists() and time.monotonic() < deadline:
+ time.sleep(0.01)
+ require(fifo.exists() and fifo.stat().st_mode & 0o777 == 0o600,
+ 'demo.selftest-consent-not-private')
+ # The trusted operator channel supplies the exact operation
+ # token; changing it cannot approve the saved native request.
+ with fifo.open('wb', buffering=0) as stream:
+ stream.write(b'approve exact-consent-test\n' if accepted else b'approve another-operation\n')
+ if accepted:
+ require(pending.result(timeout=5) is None, 'demo.selftest-consent-not-accepted')
+ else:
+ try:
+ pending.result(timeout=5)
+ raise AssertionError('substituted operator consent accepted')
+ except Refusal as error:
+ require(str(error) == 'demo.operator-review-declined',
+ 'demo.selftest-consent-wrong-refusal')
+ require(not fifo.exists() and not (directory / 'approval').exists(),
+ 'demo.selftest-consent-created-authority')
+ require(before == operator.witnesses(), 'demo.selftest-consent-entered-provider')
+ return {'cases': 2, 'changed_operation_refused': True, 'owner_only_fifo': True,
+ 'provider_entries': 0, 'human_trial': False}
+
+
+def test_private_reviewer_refuses_unlisted_inputs(operator):
+ """Unlisted or malformed app requests receive no read proof or exact approval."""
+ from operate import APP_UID, APP_GID, ENVIRONMENT, HOME
+ script = r"""
+import json, socket, struct, sys
+results = []
+for value in [{'source': 'unlisted'}, {'source': []}, {'source': 'exact', 'url': 'https://other.example'}]:
+ with socket.socket(socket.AF_UNIX, socket.SOCK_STREAM) as channel:
+ channel.settimeout(5)
+ channel.connect(sys.argv[1])
+ raw = json.dumps(value).encode()
+ channel.sendall(struct.pack('>I', len(raw)) + raw)
+ def receive(length):
+ output = b''
+ while len(output) < length:
+ part = channel.recv(length - len(output))
+ if not part:
+ raise RuntimeError('incomplete reviewer response')
+ output += part
+ return output
+ length = struct.unpack('>I', receive(4))[0]
+ if not 0 < length <= 1024:
+ raise RuntimeError('reviewer response bound')
+ results.append(json.loads(receive(length)))
+print(json.dumps(results))
+"""
+ before = operator.witnesses()
+ result = subprocess.run(['/opt/consumer/bin/python', '-c', script,
+ str(HOME / 'sockets/review.sock')], user=APP_UID, group=APP_GID, extra_groups=[],
+ env=ENVIRONMENT, capture_output=True, timeout=20)
+ require(result.returncode == 0, 'demo.selftest-review-channel-unavailable')
+ refusals = decode(result.stdout)
+ require(refusals == [{'code': 'demo.operator-recovery-review-refused'}] * 3
+ and before == operator.witnesses(), 'demo.selftest-review-channel-minted-authority')
+ return {'cases': 3, 'app_uid': APP_UID, 'native_counters_unchanged': True,
+ 'returned_proof_or_approval': False}
def run(operator):
from operate import HOME
+ consent = test_exact_operator_consent_binding(operator)
+ reviewer = test_private_reviewer_refuses_unlisted_inputs(operator)
cases = test_native_refusals_replay_race_and_loss(operator)
verified = test_application_isolation_and_native_audit(operator)
report = {'schema': 'auths.stripe-gateway-demo-evidence/1',
'mode': 'simulation', 'production_qualification': False,
'independent_human_trial': False, 'source_commit': decode(read(Path(__file__).parent / 'manifest.json'))['source_commit'],
- 'cases': cases, **verified,
+ 'operator_consent': consent, 'operator_review_channel': reviewer, 'cases': cases, **verified,
'excluded': ['live Stripe writes', 'production AWS custody and PostgreSQL',
'protected full-corpus qualification', 'provider key rotation',
'independent security review', 'GitHub release installation']}
@@ -170,8 +334,34 @@ def run(operator):
capture_output=True, timeout=30)
require(support.returncode == 0, 'demo.selftest-support-refused')
published = data + audit + support.stdout
+ protected_report = decode(read(HOME / 'private/verified-audit-report.json'))
+ protected_digests = []
+ for entry in protected_report['entries']:
+ result = entry.get('provider_result') or {}
+ for key in ['response_digest', 'evidence_digest', 'relative_basis_digest']:
+ if result.get(key):
+ protected_digests.append(bytes.fromhex(result[key]))
+ public_buffers = [published]
+ def collect_public(value):
+ if isinstance(value, dict):
+ for item in value.values():
+ collect_public(item)
+ elif isinstance(value, list):
+ for item in value:
+ collect_public(item)
+ elif isinstance(value, str):
+ try:
+ public_buffers.append(base64.urlsafe_b64decode(value + '=' * (-len(value) % 4)))
+ except (ValueError, UnicodeError):
+ pass
+ for payload in [decode(data), decode(audit), decode(support.stdout)]:
+ collect_public(payload)
+ for raw in protected_digests:
+ for encoded in [raw, raw.hex().encode(), base64.b64encode(raw), base64.urlsafe_b64encode(raw)]:
+ require(all(encoded not in payload for payload in public_buffers),
+ 'demo.selftest-public-private-body-digest-exposure')
private_seeds = [read(HOME / 'private' / (name + '.seed'))
- for name in ['root', 'manager-a', 'manager-b', 'manager-c']]
+ for name in ['root', 'reader', 'manager-a', 'manager-b', 'manager-c']]
for secret in list(operator.keys.values()) + private_seeds + [read(HOME / 'app/actor.seed')]:
raw = secret.encode() if isinstance(secret, str) else secret
for encoded in [raw, raw.hex().encode(), base64.b64encode(raw), base64.urlsafe_b64encode(raw)]:
diff --git a/demos/stripe-gateway/submit.mts b/demos/stripe-gateway/submit.mts
index 4ebe7a679..fdb7e2147 100644
--- a/demos/stripe-gateway/submit.mts
+++ b/demos/stripe-gateway/submit.mts
@@ -10,7 +10,7 @@ for await (const chunk of process.stdin) {
if (input.length > 8 * 1024 * 1024) throw new Error('demo.packet-bound');
}
const packet = JSON.parse(input) as Record;
-if (Object.keys(packet).sort().join(',') !== 'action_b64,proof_b64') throw new Error('demo.invalid-packet');
+if (Object.keys(packet).sort().join(',') !== 'action_b64,approval_b64,proof_b64') throw new Error('demo.invalid-packet');
function bytes(value: unknown): Uint8Array {
if (typeof value !== 'string' || !/^[A-Za-z0-9_-]+$/.test(value)) throw new Error('demo.invalid-packet');
const result = Buffer.from(value, 'base64url');
@@ -18,4 +18,4 @@ function bytes(value: unknown): Uint8Array {
return result;
}
const client = new GatewayClient(new GatewayEndpoint(socket));
-console.log(JSON.stringify(await client.submit({proof: bytes(packet.proof_b64), action: bytes(packet.action_b64)})));
+console.log(JSON.stringify(await client.submit({proof: bytes(packet.proof_b64), action: bytes(packet.action_b64), approval: bytes(packet.approval_b64)})));
diff --git a/demos/stripe-gateway/tests/browser.py b/demos/stripe-gateway/tests/browser.py
index 1057e887d..1a3de60ad 100644
--- a/demos/stripe-gateway/tests/browser.py
+++ b/demos/stripe-gateway/tests/browser.py
@@ -16,15 +16,15 @@ def test_packaged_workbench_and_unsigned_evidence():
expect(page.locator('#approval-review')).to_contain_text('create_refund_v1')
for case, outcome in [('approval-insufficient', 'DENIED'),
('approval-duplicate', 'NOT-SUBMITTED'),
- ('above-boundary', 'NOT-ENTERED'), ('exact', 'OBSERVED-BY-PROVIDER'),
- ('replay', 'NOT-ENTERED'), ('lost-response', 'UNKNOWN')]:
+ ('above-boundary', 'NOT-ENTERED'), ('exact', 'RESPONSE-RECORDED'), ('recover-exact', 'RECOVERY-REPORT'),
+ ('replay', 'OBSERVED-BY-PROVIDER'), ('lost-response', 'UNKNOWN')]:
page.locator('#experiment').select_option(case)
page.locator('#submit').click()
expect(page.locator('#outcome')).to_have_text(outcome, timeout=45000)
expect(page.locator('#submit')).to_be_enabled()
expect(page.locator('#explanation')).to_contain_text('no automatic retry')
page.locator('#receipt-link').click()
- expect(page.locator('#result-json')).to_contain_text('auths.gateway-audit-bundle/2')
+ expect(page.locator('#result-json')).to_contain_text('auths.gateway-audit-bundle/3')
expect(page.locator('#result-json')).to_contain_text('outcome_b64')
page.screenshot(path='/evidence/workbench-evidence.png', full_page=True)
page.goto('http://localhost:8080')
diff --git a/demos/stripe-gateway/web/app.js b/demos/stripe-gateway/web/app.js
index 219b13925..c06c2e513 100644
--- a/demos/stripe-gateway/web/app.js
+++ b/demos/stripe-gateway/web/app.js
@@ -3,6 +3,9 @@ const element = id => document.getElementById(id);
const descriptions = {
'observed-by-provider': 'A fresh provider read matched the verified value and this action’s echo. The link assumes no other writer copied the echo.',
observed: 'The gateway read the declared field. Value equality alone does not link that value to this action or establish its effect.',
+ 'recovery-report': 'A separately authorized read completed. Inspect the original write stage, read facts and evidence disposition below; a state match alone is not causation.',
+ refused: 'The gateway refused this recovery action. No new write was authorized.',
+ reserved: 'The operation is reserved. Another process may still be working; this is not permission to repeat delivery.',
'response-recorded': 'The gateway recorded an HTTP response. This alone does not establish an effect.',
unknown: 'The outcome remains unknown. The provider may have received the write; its response or locator is unavailable. Authority remains consumed; no automatic retry is allowed.',
'not-entered': 'The gateway refused before write transport. A provider-read guard may already have acquired a credential and read evidence.',
@@ -20,6 +23,8 @@ function display(record) {
: item.result.outcome.toUpperCase()).join(' / ');
element('explanation').textContent = outcomes.map(value => descriptions[value] || 'Inspect the exact gateway result below.').join(' ');
element('result-json').textContent = JSON.stringify(record, null, 2);
+ const reviewed = record.records.find(item => item.result.literal_review);
+ if (reviewed) element('approval-review').textContent = reviewed.result.literal_review;
}
async function initialize() {
const response = await fetch('/api/status');
@@ -47,8 +52,9 @@ async function initialize() {
const value = element('experiment').value;
const action = status.proposals[value] || status.proposals.exact;
element('proposal').textContent = `${action.payment_intent} · ${action.amount / 100} ${action.currency.toUpperCase()} · operation ${action.operation_id}`;
- const review = status.approval.reviews[value] || status.approval.reviews.exact;
- element('approval-review').textContent = `${review.title} · ${review.fields.map(([label, field]) => `${label}: ${field}`).join(' · ')} · expires ${new Date(review.valid_until * 1000).toLocaleTimeString()}`;
+ const source = value.startsWith('recover-') ? value.slice(8) : value;
+ const review = status.effect_review.reviews[source] || (['replay', 'typescript', 'observe'].includes(value) ? status.effect_review.reviews.exact : null);
+ element('approval-review').textContent = value.startsWith('recover-') ? 'A separate native read review and operator consent are required. The exact read review is retained with its result.' : review || 'Native review refused this hostile input. This experiment substitutes another effect’s approval and must be rejected.';
const link = element('provider-link');
link.hidden = status.mode !== 'live' || !/^pi_[A-Za-z0-9]+$/.test(action.payment_intent);
if (!link.hidden) link.href = `https://dashboard.stripe.com/test/payments/${action.payment_intent}`;
diff --git a/demos/stripe-gateway/web/index.html b/demos/stripe-gateway/web/index.html
index be03b5599..e40a9b50b 100644
--- a/demos/stripe-gateway/web/index.html
+++ b/demos/stripe-gateway/web/index.html
@@ -26,7 +26,7 @@
Let the app request a refund. Keep the Stripe key out of its hands.
-
Each experiment fixes its payment, operation and signed arguments.
+
Each experiment fixes its payment, operation and signed arguments.
The native review binds the exact action, authority and expiry.
Open on Stripe ↗
diff --git a/demos/stripe-gateway/web/styles.css b/demos/stripe-gateway/web/styles.css
index cafb35838..ddd8ea6fc 100644
--- a/demos/stripe-gateway/web/styles.css
+++ b/demos/stripe-gateway/web/styles.css
@@ -1,3 +1,5 @@
:root{font-family:Inter,ui-sans-serif,system-ui,sans-serif;color:#122035;background:#f7f8fa;font-synthesis:none}*{box-sizing:border-box}body{margin:0}header,footer{padding:24px max(24px,calc((100vw - 1120px)/2));display:flex;justify-content:space-between;gap:20px;border-bottom:1px solid #dde2ea}header a{font-weight:800;letter-spacing:.08em;text-decoration:none;color:#122035}header span{font-size:11px;letter-spacing:.09em;font-weight:700;color:#25615a}main{max-width:1120px;margin:auto;padding:64px 24px}.eyebrow{font-size:12px;text-transform:uppercase;letter-spacing:.15em;color:#416489;font-weight:700}h1{font-size:clamp(30px,4vw,49px);line-height:1.16;letter-spacing:-.045em;margin:14px 0 22px}h2{font-size:21px;margin-top:0}.intro{font-size:18px;line-height:1.6;max-width:750px;color:#4a5c72}#connection{margin:30px 0;padding:14px 18px;border-left:3px solid #3574d5;background:#eef3fb;font-size:14px}.workbench{display:grid;grid-template-columns:1fr 1fr;gap:24px;margin:32px 0}article,.result{border:1px solid #dbe2ed;background:white;border-radius:12px;padding:28px}article p,.limits p{font-size:14px;line-height:1.7;color:#526074}dl{display:grid;grid-template-columns:1fr 1.4fr;gap:14px;font-size:13px}dt{color:#637389}dd{margin:0;font-weight:650;overflow-wrap:anywhere}.digest{font-family:ui-monospace,monospace;font-size:11px}label{display:block;font-size:13px;font-weight:700;margin:24px 0 10px}select{width:100%;padding:14px;background:#f7f9fd;border:1px solid #b7c6da;border-radius:7px;font:inherit;font-size:14px}button{width:100%;padding:16px;border:0;border-radius:7px;background:#2563eb;color:white;font:inherit;font-size:14px;font-weight:700;cursor:pointer}button:disabled{opacity:.5;cursor:wait}.small{font-size:12px!important}.stages{display:flex;flex-wrap:wrap;gap:8px;list-style:none;padding:0;color:#55708c;font-size:12px}.stages li{border:1px solid #dde5ef;padding:8px;border-radius:5px}.result #outcome{overflow-wrap:anywhere;font-size:26px}.result p{line-height:1.7;color:#526074}pre{white-space:pre-wrap;overflow-wrap:anywhere;background:#111f34;color:#d7e5f7;border-radius:8px;padding:22px;font:12px/1.65 ui-monospace,monospace;max-height:450px;overflow:auto}.result a{display:inline-block;padding-top:12px;font-size:14px;font-weight:700;color:#2563eb}.limits{padding:34px 2px 0;max-width:860px}footer{border-top:1px solid #dde2ea;font-size:12px;color:#637389;justify-content:center}@media(max-width:760px){main{padding-top:34px}.workbench{grid-template-columns:1fr}header{flex-direction:column}article,.result{padding:22px}dl{grid-template-columns:1fr 1fr}.result #outcome{font-size:21px}}
#provider-link{display:block;margin:14px 0;color:#2563eb;font-weight:700;font-size:14px}#provider-link[hidden]{display:none}
+
+#approval-review { white-space: pre-wrap; overflow-wrap: anywhere; }
diff --git a/deployment/gateway/tools/operator-simulation.py b/deployment/gateway/tools/operator-simulation.py
index 590a156bf..418aec295 100644
--- a/deployment/gateway/tools/operator-simulation.py
+++ b/deployment/gateway/tools/operator-simulation.py
@@ -84,7 +84,7 @@ def json(self, label, args, **kwargs):
return json.loads(self.run(label, args, **kwargs))
def install(self, state, store=None, join=False, secret=None, extra=(), label="install"):
- args = ["install", "--state-dir", state, "--recipe", self.inputs / "recipe.json",
+ args = ["install", "--effect-policy", self.inputs / "effect-policy.json", "--state-dir", state, "--recipe", self.inputs / "recipe.json",
"--profile-lock", self.inputs / "profile.lock.json",
"--trusted-context", self.inputs / "trusted.context.cbor",
"--approve-digest", self.digest, "--provider", "airtable",
@@ -174,7 +174,7 @@ def exercise(self):
private_tree(first, self.uid, self.gid)
self.json("offline-support", ["support-bundle", "--state-dir", first])
# These are shipped builds: neither test transport nor plaintext production is allowed.
- args = ["install", "--state-dir", self.work / "production", "--recipe",
+ args = ["install", "--effect-policy", self.inputs / "effect-policy.json", "--state-dir", self.work / "production", "--recipe",
self.inputs / "recipe.json", "--profile-lock", self.inputs / "profile.lock.json",
"--trusted-context", self.inputs / "trusted.context.cbor", "--approve-digest",
self.digest, "--provider", "airtable", "--alias", "simulation",
@@ -231,8 +231,11 @@ def exercise(self):
and one["credential_held"] and two["credential_held"], "simulation.rotation-diverged")
self.json("redacted-support", ["support-bundle", "--state-dir", first],
env={"AUTHS_SIMULATION_CANARY": self.secret.decode()})
- self.admin(first, "reobserve", extra=["--operation-id", "simulation-no-such-operation"],
- success=False, code="gateway.reobserve.not-observable")
+ # AP074 removed raw re-observation. An operator must author and approve
+ # an independent recovery effect; the obsolete command stays unavailable.
+ self.run("refuse-obsolete-reobserve", ["reobserve", "--state-dir", first,
+ "--operation-id", "simulation-no-such-operation"], success=False,
+ code="unrecognized subcommand 'reobserve'")
tuple_bytes = self.run("deployment-tuple", ["qualification-status", "--state-dir", first, "--tuple"])
tuple_path = self.work / "tuple.json"
tuple_path.write_bytes(tuple_bytes)
diff --git a/docs/adr/0013-recipe-capabilities-and-sum-budget.md b/docs/adr/0013-recipe-capabilities-and-sum-budget.md
index f342078ad..c8479472a 100644
--- a/docs/adr/0013-recipe-capabilities-and-sum-budget.md
+++ b/docs/adr/0013-recipe-capabilities-and-sum-budget.md
@@ -249,7 +249,7 @@ evidence pipeline for the recipe it names.
- Epics 2–7 of AP-SPEC-063 implement the capabilities. Epic 1 commits their
vectors under `bindings/fixtures/gateway/` (`hostile-recipes-v2.json`,
- `attempt-scenarios-v3.json`, `bounds-aggregate.json`, `outcome-v2.json`,
+ `attempt-scenarios-v3.json`, `bounds-aggregate.json`, `outcome-v3.json`,
`key-identity.json`, and `codes.json`), each with a test showing that
current code does not satisfy it.
- ADR 0012 stays in force. Epic 8 rewrote its decision and consequences
diff --git a/docs/adr/0014-stripe-refund-recipe-qualification.md b/docs/adr/0014-stripe-refund-recipe-qualification.md
index 9a6713e53..8030ec5b7 100644
--- a/docs/adr/0014-stripe-refund-recipe-qualification.md
+++ b/docs/adr/0014-stripe-refund-recipe-qualification.md
@@ -61,7 +61,7 @@ the gateway receives only the restricted test key.
## Corpus, recovery and publication gates
The maintained platform profile/recipe are accompanied by gateway
-`attempt-scenarios-v3.json`, `bounds-aggregate.json`, `outcome-v2.json` and
+`attempt-scenarios-v3.json`, `bounds-aggregate.json`, `outcome-v3.json` and
`hostile-recipes-v2.json`. They must be expanded into the family-owned
`auths.qualification-corpus/3` with executable coverage of every mandatory wall
scenario; they are not the missing protected family corpus.
diff --git a/docs/adr/0017-bound-effect-and-recovery-ownership.md b/docs/adr/0017-bound-effect-and-recovery-ownership.md
new file mode 100644
index 000000000..98d680486
--- /dev/null
+++ b/docs/adr/0017-bound-effect-and-recovery-ownership.md
@@ -0,0 +1,1408 @@
+# ADR 0017: One bound effect and gateway recovery owner
+
+- Date: 28 October 2026
+- Status: Implementation decision for AP-SPEC-074; evidence gates remain open
+
+The gateway already owns production provider writes. AP070 separately decides
+read admission, release and conservative accounting in Python/SQLite. The
+lifecycle package has useful preparation and durable-acknowledgement contracts,
+but its profile states include partial effects and proven non-effect with
+different recovery semantics. None can be treated as interchangeable merely
+because each has a journal.
+
+### Scope clarification: independent application-owned references
+
+The owner's subsequent decision retains AP070 as an independent runnable
+application-owned reference. Its Python/SQLite broker is not a second authority
+for a managed gateway operation and need not migrate to establish AP074.
+Preserve its existing narrower claims and differential value. No managed grant,
+credential, permit or native recovery decision may be delegated to it; its
+receipts cannot be substituted for gateway evidence. Inventory the distinct
+channels, state and credentials and enforce the managed deployment boundary.
+Independent external credentials/writers remain an explicitly excluded threat.
+
+Future managed Google productization requires a reviewed native integration and
+one native journal for those managed effects. Google metadata/refresh and AP070
+owner removal are deferred, while every generic native, binding/UI, file and
+PostgreSQL concurrency/crash/fault, formal and containment gate remains required.
+Historical checkpoints below describe earlier implementation scope; this decision
+and the current specification/acceptance ledger govern remaining work.
+
+Keep pure managed transitions in `auths-gateway-kernel::transition`, orchestration
+in `auths-gateway::submit`, and the logical journal in `auths-gateway::store`.
+Extend those owners; do not introduce a parallel executor or task ledger.
+`auths-stores` implements durable PostgreSQL mechanisms. The existing file
+adapter stays beside the gateway until moving it has a concrete dependency
+benefit. Store adapters never decide provider truth or refund policy.
+
+The gateway owns the canonical bound effect, built from the native sealed action
+and closed recipe output. It includes selected branch/evidence identity, trust,
+scope, installation, typed charges and protocol. Exact approval signs that
+identity after compilation; the signature is not an input to its own identity.
+The stored effect is data. Only a successful new durable entry transition yields
+a non-cloneable, non-deserializable, consumed transport capability.
+
+### Retained submission identity and current authorization
+
+Retrieving an admitted fact is independent of authorizing another entry. The
+native journal retains a bounded, canonical portable projection of the original
+native verification result, including its exact proof and action digests. Its
+normalized authority commitment must equal the immutable binding's existing
+native-authority commitment. The projection's portable self-digest is validated;
+the original commitment preimage is reconstructed with the native codec, keeping
+the existing binding identity unchanged. The private projection is at most
+65,536 bytes and is immutable under every protocol transition.
+
+The application action supplies bounded lookup hints only. Native history
+retrieval checks the partition/epoch, namespace, operation, recipe, both original
+input digests and a signature for the retained effect under its retained approver
+pin. Approval validity intervals and installed keys are current-entry conditions;
+an expired original approval can identify history without granting current
+authority. Additional API resource, recovery-source and disclosure-target inputs
+must still match the historical signed action. No credential lease, native
+reauthorization, peer authentication or provider/delivery request is needed to
+return already entered or terminal facts.
+
+A reserved operation may continue only through the existing full fresh native
+compilation, current exact approval and atomic writer fence. Failure to acquire
+current authority leaves its retained reservation visible; it does not rewrite
+history as a new non-entry fact. This history path returns only typed snapshots,
+never `BoundEffect`, `ApprovedEffect`, an authenticated channel or a permit.
+The journal cuts over directly to `auths.gateway-attempt/16`; obsolete disposable
+records are rejected. There is no legacy reader or second authority cache.
+
+Retain lifecycle identity/commitment helpers only where their exact contract
+matches. `PreparationBindingV1` is not reused as the effect: its optional
+connection/evidence and single principal do not encode disclosure/source/channel,
+selected native branches, exact compiler output or typed reservation units.
+`ProviderCallAuthorizationV1::from_durable(&DurableTransitionV1)` illustrates
+acknowledgement validation but a borrowed acknowledgement can be reused. AP074
+must consume the newly applied acknowledgement exactly once. Do not change
+reference consumers' lifecycle meaning to make these APIs fit.
+
+The extraction is limited to identity preservation, finite typed reservation
+arithmetic, single grant and immutable evidence. This is a domain-independent
+product mechanism; the smaller-than-three-production-consumer exception is
+justified by these exact properties, not inferred provider similarity. Provider
+result interpretation remains in closed reviewed recipe declarations and
+independent references. The [comparison case](../research/domains/abstraction-cases/0009-bound-effect-and-recovery.md)
+defines divergences and the evidence still required before promotion is complete.
+
+File and PostgreSQL admission must order configuration/revocation with entry.
+Read-and-CAS of just the attempt does not fence a concurrently changed connection.
+The transaction derives current state, samples the deployment clock after lock
+acquisition, validates full native authority and pinned dependencies, and commits
+the grant and charges before acknowledgement. A committed revocation blocks later
+grants; an earlier grant may still be exercised. Process-local in-flight counters
+alone do not establish the multi-process order.
+
+Before the durable mutation cutover, this ADR must record the implemented file
+lock and PostgreSQL transaction scopes, lock order, versioned dependency records,
+configuration/revocation writer participation, after-lock clock sampling,
+verification deadlines and durable acknowledgement. The P2 design below fixes
+that mechanism; its source implementation alone does not satisfy final entry
+until the shared configuration record and grant owner are connected. The present
+attempt CAS and connection reload do not satisfy it.
+Cross-store state requires demonstrated fencing; unsupported atomicity fails
+closed. Capacity reserved earlier is checked for intact conservation at grant,
+not allocated a second time.
+
+The installed gateway boundary must exclude untrusted agents from provider
+credentials, journal mutation, operator channels and alternate managed transport.
+Keep an interface inventory and exercise it from a separate restricted process.
+Rust sealed values do not establish process isolation, and independent provider
+writers or a malicious trusted host/operator are outside this containment claim.
+
+Native review must implement a versioned recipe-bound completeness contract.
+Missing mandatory fields or an unsupported renderer prevents approval; signing
+an effect digest alone does not establish complete presentation. Domain profiles
+own the meanings and units of mandatory fields. Faithful presentation makes no
+claim about human comprehension.
+
+The managed Airtable demo's optional owner HTTP interface is consent transport,
+not a second semantic owner. The operator retains the complete native review and
+request, authenticates the owner through a private bootstrap credential, and
+signs only after an exact pending-review decision. The application receives no
+preissued exact approvals in this mode. Native code still owns approval bytes,
+resource authority, the immutable effect and every durable transition. Owner
+authentication stays in explicit origin-bound headers rather than cookies that
+cross localhost ports. Session/consent time limits never refresh native signed
+times or create an entry permit. This remains a simulation demonstration with
+hosted tests authored; complete installed acquisition/disclosure UI, broader
+containment and exact-head evidence remain required.
+
+The installed local owner journey now composes an additional data-only reviewed
+recipe with the existing content capability. It uses one native control/journal
+for acquisitions, separate recipient release and writes, and the same native
+agent actor. Python and TypeScript submit native acquisition inputs; the native
+recipient CLI owns the authenticated channel, exact byte check and ACK. The
+application stores authorized plaintext only in recipient-private files and
+shows the gateway's retained facts. No Python/TypeScript effect hash, permit,
+accounting decision, ACK codec or second admission owner is introduced. The
+root-private development actor simulator uses a fixed copy created before app
+file ownership changes; it never opens an app-writable signer configuration as
+root. Local recipe selection is an explicit capability choice, with no legacy
+reader or alternate runtime implementation. Exact-head hosted execution and
+broader containment/adverse-path obligations remain required.
+
+Recovery exposes bounded observation work, retained charges and escalation.
+Administrative annotations are not provider evidence; no administrative path can
+convert post-grant uncertainty into non-entry or silently release capacity.
+Future delegated scope and ancestor charges feed this same admission authority,
+under atomicity or demonstrated fencing, without a second task/effect journal.
+This is an integration boundary only; AP073 remains outside the implementation.
+
+Acquisition, disclosure and mutation are closed typed alternatives. Ancillary
+credential/account/preflight/observation reads are linked bounded acquisitions.
+Each disclosure has its own recipient authority, source binding and charge.
+An ACK records recipient evidence and never creates another delivery grant.
+State-match observation may coexist with unknown causation.
+
+Cut over directly and reject obsolete disposable journal formats. No legacy
+decoder, conversion, dual-write path or compatibility switch. Preserve rollback
+floors and PostgreSQL transaction rollback; these are current safety semantics.
+Retain reference implementations and original formal evidence only with accurate
+scope. Acceptance requires A01–A22, updated source-linked proofs and hosted CI;
+this ADR itself is not evidence that those gates passed.
+
+## Installed mutation journal identity cutover
+
+Attempt record `auths.gateway-attempt/4` retains the complete native canonical
+binding and original canonical exact approval as immutable fixed fields. The
+existing transition predicate compares those fixed bytes; a later observation
+or outcome cannot substitute them. Historical approval validation uses the
+recorded evaluation time and the approver pinned in that binding, not a new
+approval supplied by a replay. This preserves evidence; it does not grant entry.
+
+The only operation lookup key is the repository domain commitment
+`auths.effect-replay-key/1` over canonical `(namespace, partition, epoch,
+operation)`. It excludes the effect digest. An insertion collision loads and
+validates the retained record and compares the complete binding: equality is
+replay, inequality is `gateway.effect.operation-conflict`. Neither changes the
+record or allocates another slot. Malformed state is unavailable, never replay
+success or a free operation. There is no legacy lookup or decoder.
+
+The runtime constructs the attempt owner from the same installed effect policy
+as the engine and refuses a mismatch. Support listings filter the installed
+partition/epoch within their bounded scan. Native count/sum counter keys remain
+unchanged: changing an operation epoch must not reset a shared grant's capacity.
+Old attempt tombstones remain durable. A higher epoch is an independently
+installed namespace, not evidence of rollback protection or permission to retry
+an old unknown operation. Store conformance source covers both adapters, but
+its new binding/scope cases have not yet been executed on hosted CI.
+
+This cutover does not replace `Attempting` with a durable entry grant, provide a
+configuration fence, or finish independent recovery authority. Those remain the
+explicit P2 requirements above.
+
+## P2 adapter fence design
+
+The gateway keeps lifecycle interpretation in its existing transition owner.
+The adapter adds one mechanism: read a bounded set of records under the common
+writer fence, invoke one bounded in-process decision, and durably replace the
+first record before returning. The decision receives opaque snapshots, performs
+no provider, custody or nested store I/O, and returns only the replacement bytes.
+The adapter calls it at most once and never retries a failed/ambiguous commit.
+The gateway alone interprets those bytes and may consume a newly acknowledged
+`Reserved -> EntryGranted` transition to construct one non-cloneable permit.
+Reading `EntryGranted` from disk cannot construct that permit.
+
+File scope is the existing host-wide `.replace.lock`. Every insert, batch,
+replacement, sweep and fenced decision uses it; fenced snapshots use lock-held
+read helpers instead of acquiring the lock recursively. Acquisition uses
+nonblocking flock attempts with a two-second deadline. The decision's work is
+bounded by the native input limits and a two-second elapsed deadline; the store
+checks the deadline before publication. Atomic replacement uses synced temporary
+bytes, rename and directory sync. A failure after publication returns no permit;
+the surviving grant prevents reminting after restart. OS scheduling and storage
+sync are trusted platform operations, not preemptible hard real-time guarantees.
+
+PostgreSQL scope is an advisory transaction lock keyed by the qualified gateway
+record table's OID and a fixed gateway domain. All gateway insert, replacement
+and expiry-sweep writers acquire it before row work; this includes connection
+revocation and credential rotation. Row insertion retains sorted key order.
+The fence uses READ COMMITTED: snapshots are loaded only after the advisory lock
+is acquired, so a committed predecessor is visible. Local lock, statement and
+idle-transaction timeouts bound database waits and preserve any tighter installed
+pool timeout; values below SQL's millisecond precision are rejected rather than
+silently disabling a deadline. The same two-second decision
+limit applies before UPDATE and COMMIT. COMMIT must acknowledge durability before
+the caller receives success. Pool failures, serialization errors, timeouts and
+ambiguous commits return no permit. No transaction spans credential access or a
+provider call. Shared-store throughput is deliberately serialized initially;
+partitioned locks require later evidence for shared-budget and configuration
+interactions, not an opportunistic optimization.
+
+The final decision's bounded read set must contain the attempt, its connection,
+the shared installed-configuration/revocation record and every reserved counter
+slot. The native owner samples the deployment clock after obtaining those
+snapshots, revalidates the original proof, exact approval and current policy,
+checks pinned configuration/revocation and intact reservation membership, then
+records the grant and audit event in the same attempt replacement. Reserved
+slots become consumed by the grant stage, without allocating or charging them
+twice. Unsupported split-store installations fail closed.
+
+A new shared configuration record must pin recipe/compiler, trust context,
+operator effect policy and qualification/revocation versions. Installation,
+qualification refresh and operator configuration writers must participate in
+the same fence; a host's cached configuration must match that record. Local
+manifest comparison and process-local revocation caches are insufficient.
+The primitive described here is only preparation: until the shared record,
+writer participation and grant reducer are connected, the current engine does
+not claim this final-entry guarantee.
+
+## Shared installed configuration and permanent qualification memory
+
+The sixth closed gateway record kind, `effect-configuration`, is non-expiring.
+Its key commits to namespace and partition independently of the candidate
+configuration and epoch. Configuration schema `/2` stores the active epoch;
+configuration-key domain `/2` makes old and new hosts contend on one control.
+Operator activation cannot decrease that epoch, including with a fresh review
+token. Installation manifest `/7` rejects obsolete disposable installations;
+there is no old-schema reader or migration path. The installed pin commits to recipe, native verifier trust,
+operator effect policy, connection identity, workload, profile, qualification
+contract and gateway semantic closure. Bound-effect `/2` includes that pin;
+compiler identity `/2` rejects approvals over the earlier binding format.
+PostgreSQL schema `/7` directly replaces `/6`; no old-schema reader is retained.
+
+Only trusted install/join initializes a missing record. Joining an existing
+record must match its pin. Changing the pin requires explicit operator review
+of both the candidate and a commitment to the entire previous record, followed
+by activation under the writer fence. A concurrent publication invalidates that
+review token. Application frames expose neither initialization nor activation.
+
+Qualification import/reload publishes the authenticated bundle identity and
+permanent revocation memory through that same fence before reporting success.
+Accepted sequence/issue-time floors take maxima; signer and qualification
+revocations take unions. The record bounds these sets to the native release
+limits (64 signers and 1024 qualifications). Saturation persists a terminal
+blocked marker; neither an empty local cache nor configuration activation can
+reset it. Each record is canonical and bounded to 128 KiB. Revision saturation
+also blocks authorization. This preserves bounded work without forgetting a
+revocation and then authorizing from a truncated set.
+
+A host adopts shared memory and must hold the published bundle identity before
+ordinary qualification or separate commissioning-permit verification. Shared
+memory itself grants neither. Publishing an unloaded bundle can disable other
+hosts until their inputs are synchronized; this is a deliberate fail-closed
+availability tradeoff. Earlier local verifier-state files remain rollback
+witnesses, not an alternate shared admission authority.
+
+Compilation and entry preparation now perform early checks. The final grant
+must still repeat these checks over the configuration snapshot under its own
+attempt fence. No final-entry guarantee is claimed by these early checks.
+
+## Mutation grant source cutover
+
+Attempt schema `/5` separates `Reserved` from `EntryGranted`. Admission creates
+only a reservation. The existing kernel transition owner requires immutable
+entry-audit bytes for every post-grant stage and forbids post-grant non-entry.
+The existing submission step machine now has an explicit `GrantEntry` action;
+only its successful acknowledgement can lead to `Send`.
+
+`ClaimedGatewayAttempt::grant_entry` holds the common adapter fence while it
+checks the exact attempt bytes, installed connection, shared configuration and
+every reserved count/sum slot. Commissioning additionally reads its permanent
+budget record and checks it against the acknowledged pre-custody floor. The
+native validation samples the deployment clock after lock acquisition, freezes
+one published qualification bundle plus monotonic memory, re-runs the original
+proof and exact approval, and checks the connection/workload/profile and time
+bounds. A concurrent process-local reload cannot replace that frozen evidence.
+No provider, custody or nested store I/O runs in the decision. Deployment clock
+sampling retains the clock implementation's explicit platform assumptions.
+
+Reservation membership checks cover exact key, kind, operation, window, slot,
+amount and capacity, including the one-to-one correspondence with native count
+and sum obligations. The grant allocates no second slot. Its private immutable
+audit includes the effect identity, grant time and domain-separated commitments
+to dependency snapshots. The adapter must acknowledge durable publication before
+one non-cloneable/non-deserializable permit is constructed. Every managed write
+port, through the final HTTPS transport, consumes that permit and obtains its
+request from the retained native command. Callers cannot supply substitute bytes.
+
+After a grant, even a transport failure before a socket opens is unknown; it
+cannot become `NotEntered`. A lost commit acknowledgement returns no permit.
+Reading a granted record after restart projects unknown and never remints one.
+The native corpus exercises one file-backed grant, stale-grant refusal and
+crash-before-send projection; separate file/PostgreSQL race tests are authored.
+This is not yet full P2 evidence. Reopening a reservation with fresh approval,
+explicit journal revisions, installed after-lock expiry/revocation races,
+ambiguous-commit fault injection and translated/refinement regeneration remain
+open. Acquisition/disclosure and independent recovery are later cutovers.
+
+
+## Reservation restart and immutable approval history
+
+Attempt `/6` requires monotonic exact-JSON-integer revisions. Every update uses
+both the complete predecessor bytes and one revision increment; saturation
+refuses progress. An identical reserved operation may reopen with its existing
+allocation. It does not acquire entry authority and cannot reopen a granted or
+terminal operation. Full-binding differences still conflict.
+
+A renewed exact approval does not rewrite the admission approval or first
+pre-entry checkpoint. Grant audit `/2` separately freezes the actual approval,
+fresh authorization start time, revision and fresh reads used at final entry.
+The original proof is revalidated under the fence, using the new submission's
+bounded deadline. The public snapshot uses the grant's pre-entry evidence after
+entry. Stale handles lose the same atomic replacement; stored grants still
+cannot recreate transport permits. Count/sum slots are checked and retained,
+never charged a second time by resumption.
+
+The synthetic native file corpus now covers restart beyond the first approval
+and submission deadline, fresh approval, unchanged exhausted-capacity slots,
+immutable initial evidence and refusal of a second grant. PostgreSQL parity
+source is authored; installed authority, process faults and full recovery
+remain separate evidence gates. A revision in the same store cannot detect
+whole-store rollback.
+
+
+## Recipe-owned review semantics
+
+Source `/3` adds one mandatory versioned review declaration. Native validation
+requires complete typed field coverage, bounded printable labels, explicit
+integer units and operator-authored consequences. These are presentation facts,
+not gateway-interpreted financial or provider semantics. The recipe digest and
+effect compiler `/3` bind them to the exact effect. Compilation refuses a draft,
+unknown version, duplicate/missing field, incompatible type or unsafe label.
+
+Private review `/2` carries the declaration's values from the sealed command,
+complete request, structured binding and actual connection identity/generation.
+The native literal renderer escapes untrusted text as JSON data and refuses
+unsupported versions or output limits. The CLI emits only that safe presentation
+with its exact signing statement and writes the request file after output succeeds.
+This is a local presentation guarantee, not proof of an authenticated external
+reviewer or of human understanding; installed channel/UI evidence remains open.
+
+OpenAPI derivation owns transport shape only. Its generated review declaration
+has empty consequences/units and cannot compile until the operator completes
+it. Test-only declarations support independent transport comparisons without
+silently approving business meaning from untrusted descriptions. The full
+Todoist vendor input currently fails the pinned digest; equivalent frozen native
+excerpt outputs update the render artifacts while full vendor execution remains
+explicitly pending. No source pin or verification rule was weakened.
+
+
+## Public replay and unacknowledged grants
+
+The mutation submission driver returns stored facts for equal bindings that
+already have a grant or terminal result. Replay stops before a credential lease
+and cannot invoke the operator observation path. An equal pre-grant reservation
+may continue only through the separate fresh authorization and grant checks.
+
+The native result represents that reservation explicitly with its bounded
+revision. If admission or grant acknowledgement is unavailable, reload the exact
+bound journal: reserved remains reserved; a retained grant without conclusive
+response remains unknown; missing/corrupt/unavailable state is indeterminate.
+These reads cannot return a permit or infer non-entry. Historical observation
+returned on replay must not be counted as a new read-back confirmation.
+
+The old explicit operator observation mechanics remain temporarily reachable
+only on the operator surface. They are not the final AP074 recovery contract:
+that direct path, automatic post-write observations and all ancillary reads must
+be replaced by independently authorized bounded acquisitions before completion.
+
+
+## Independently authorized recovery compilation
+
+Bound-effect `/3` and compiler `/4` retain one shared identity and exact approval
+owner, with sealed mutation/acquisition compiler alternatives. The source
+journal supplies protected request/locator provenance; the new native read proof
+supplies current authority. The read contract pins the exact GET, source binding,
+observation projection, echo and finite bounds. Mutation recipes cannot use the
+reserved read tool name, and mutation admission refuses a sealed acquisition.
+Neither decoded source data nor compilation can mint a transport permit.
+
+Source-lifetime reservations bind acquisition count and maximum response bytes
+separately. Their keys include the immutable source effect and unit, excluding
+the new read operation so new IDs cannot reset the declared capacity. The native
+contract allows at most 16 attempts, at most 65,536 response bytes per attempt,
+15 seconds per transport, and 30 seconds of evidence age. Entry accounting and
+evidence freshness enforcement remain pending; a declaration is not evidence
+that those boundaries have been executed.
+
+The first compiler requires the same source connection commitment and credential
+generation. Rotation or a narrower replacement credential currently refuses
+compilation; safe replacement recovery still needs its own reviewed contract.
+Managed bounded-policy extensions also fail closed until acquisition accounting
+can enforce them. These are incomplete implementation boundaries, not permission
+to drop the full acquisition/recovery requirements of AP074.
+
+The installed operator can author the private native read call and compile it
+for review using the same deployment clock, current configuration, and connection
+checks as mutation compilation. The literal renderer and signing-request path
+remain shared. No new provider endpoint or read execution is introduced by this
+checkpoint; direct operator re-observation and ancillary reads still require
+the acquisition admission/entry cutover before completion.
+
+
+## Recovery acquisition journal and entry
+
+The next cutover uses attempt journal `/7`: a common immutable header and entry
+history with a closed typed mutation/acquisition payload. It uses the same
+transition predicate, record codec and durable transaction ports. Acquisition
+responses cannot become mutation observations through a protocol or stage tag.
+Support bundle `/2` distinguishes the protocol in both counts and retained stages.
+
+One atomic admission stores the read claim and a permanent allocation holding
+one invocation and the maximum response bytes. The allocation binds distinct
+count and byte keys to the source effect; new read IDs share its fixed 16 slots.
+The existing file batch and PostgreSQL transaction implement this insert.
+PostgreSQL source schema `/8` adds the non-expiring allocation kind directly;
+there is no migration/legacy reader. Sweeping does not refund allocations.
+
+The final read grant uses the existing writer fence over the read attempt,
+source attempt, allocation and current installed dependencies. It rechecks the
+exact source-derived target and fresh read authority. A durable acknowledgement
+returns a one-use permit that the bounded transport consumes; a persisted grant
+alone returns unknown history. Response bytes stay in the private journal and
+do not themselves refine the mutation or authorize content disclosure.
+
+The runtime path refuses recipes requiring additional credential reads until
+those have their own accounted acquisition paths. Source-evidence freshness,
+credential replacement, managed bounds, public execution routes and removal of
+existing raw observation/ancillary paths remain required work. The native file
+corpus uses synthetic entry validation and response bytes; it is narrow durable
+mechanism evidence, not an installed authority or transport claim.
+
+
+## Retained recovery evidence and operator cutover
+
+Journal `/8` adds immutable acquisition provenance to source observation facts.
+The existing adapter fence checks source/read/allocation and samples the deployed
+clock before the same transition predicate can advance the mutation. Read
+receipt time is preserved. Newly recorded evidence must have age 0–30 seconds;
+exact already-recorded history does not expire or refresh. Unknown sends require
+supported provider linkage; state equality alone does not establish execution.
+Duplicate object members are refused by the bounded native observation decoder.
+
+The production operator `reobserve` path is removed. Admin request/response `/2`
+uses `recover` with a separately signed native read and exact approval, or
+`recovery-record` with already retained read evidence and no external I/O.
+Application and operator input frames share a bounded native decoder. Redacted
+recovery report `/1` separates read facts, source facts and evidence disposition
+without private content, locators or commitments. The old observation mechanics
+remain only as an explicit store-conformance test oracle. The shipping submission
+step machine has no read-back/resume mode: recording a response stops it, and
+observation requires the independent recovery acquisition. Ancillary/application
+reads remain the next production cutovers. The native corpus is file-journal mechanism evidence
+with synthetic entry/response/clock inputs, not installed transport qualification.
+
+
+## Prerequisite acquisition compilation
+
+Compiler `/5` retains each mutation prerequisite as a closed compiler request
+and a typed purpose. The ordered plan is part of the mutation transport
+commitment and appears in native private approval review. Account reads pin the
+installed account commitment and comparison domain; denied reads pin their
+ordinal and refused statuses; relative-ceiling reads resolve the verified
+argument and bind values; pre-entry reads pin their pointers and complete
+selected native authority. There is no deserializer or application constructor
+for this compiler output.
+
+The maximum plan is seven reads, each allowed once with its exact response bound,
+a 15-second transport deadline, a 30-second evidence-age ceiling and no release
+of content. Existing proof freshness conditions can be stricter. Onboarding
+probes are outside this mutation plan. The same exact-approval and native
+recompilation machinery binds these declarations; no separate token format or
+approval owner is introduced.
+
+This checkpoint does not implement prerequisite admission or grants. The
+existing raw prerequisite calls still require direct cutover to journaled,
+charged acquisition entries and one-use permits. The parent approval is evidence
+for only its enumerated recipe-derived reads; it is not general read authority
+or authority to repeat an uncertain read. The native fixture establishes signed
+compilation/review only, not entry eligibility, accounting, transport or the
+installed human channel.
+
+
+## Durable prerequisite acquisitions
+
+Journal `/9` stores a closed acquisition-plan alternative: independently
+approved recovery, or one enumerated prerequisite of the exact approved
+mutation. Both use the existing acquisition states, transition predicate,
+permanent allocation slots, fenced grant and response persistence. There is no
+new approval envelope, task ledger or retry decision owner. Old journal records
+are refused directly; the PostgreSQL storage schema remains `/8` because its
+existing record kinds and transaction contract suffice.
+
+A prerequisite record retains the parent binding and exact approval plus its
+ordinal and a bounded canonical transport membership witness. Decoding checks
+the witness against the parent transport commitment and checks the declaration's
+one-call, response-byte, transport, freshness and no-disclosure limits. Audit
+witnesses never construct transport requests. The permit carries the original
+sealed compiler read. The component replay key derives from the parent's logical
+operation key and ordinal, not its effect digest; a changed binding cannot
+create another component under the same operation. Its permanent allocation
+charges one invocation and the full response ceiling, including unknown entry.
+
+The acquisition grant fence includes its record, the reserved parent, its
+allocation and current native entry dependencies. A parent that already entered
+or became not-entered cannot issue a prerequisite permit. Only acknowledged
+publication returns a permit. Concurrent stale handles, replay and reopened
+entry grants cannot issue another. Recorded responses retain their receipt
+time; pre-entry evidence uses the earlier grant time conservatively. Loading
+history does not refresh either timestamp.
+
+Submission's account, denied-capability, pre-entry and relative-ceiling reads now
+consume these permits. It evaluates a response only after durable recording. The
+final mutation fence includes every required child record and allocation and
+refuses missing, uncertain or stale responses. Seven prerequisites add at most
+14 dependencies; together with 34 mutation slots, the parent and three installed
+dependencies this uses at most 52 of the existing 64-record fence limit.
+
+This source cutover does not close installed evidence: tighter native observer
+freshness at the final lock boundary, complete selected-evidence provenance,
+ordered initial admission, all application/onboarding reads and disclosure still
+need completion. The file fixture uses native compilation but explicit-time
+oracle grants and synthetic responses; it is not provider transport, installed
+native entry or PostgreSQL qualification.
+
+
+## Final prerequisite evidence validation
+
+The installed mutation fence now reuses the existing account, denied-capability,
+relative-ceiling and selected-observation predicates on the exact retained child
+responses. The acquisition validator returns typed, bounded evidence only after
+checking its parent binding, plan, allocation, recorded grant and original times.
+The installed entry validator runs these predicates after full native authority
+validation and before publishing the mutation grant, at the same sampled time.
+
+For pre-entry observations, the existing canonical signature verifier checks the
+pinned observer. Every signed subject, fact set and observation time must match
+the exact acquired response projection and its conservative grant time. The
+selected requirements are resolved again from the approved native command and
+current pinned trust, then evaluated with their own maximum ages and anchor
+validity. The 30-second acquisition ceiling cannot override a tighter native
+requirement. No observer signing, custody or provider I/O occurs under the fence.
+The existing four-observation limit and bounded canonical decoder bound this
+additional signature work. The selected gateway observer identity is now included in the installed
+configuration commitment as well as checked against native observer trust.
+
+Relative-ceiling checkpoints must match the retained response's basis, digest
+and original grant time, and still satisfy the original ratio and bound values.
+Extra observation or basis checkpoint data is refused when the compiled plan
+has no corresponding prerequisite. Account, ceiling and pre-entry JSON now use
+the same bounded duplicate-rejecting response decoder as recovery observations.
+
+The native predicate fixtures cover canonical signature substitution, subject,
+observer, fact and timestamp mismatch, tighter age boundaries, missing/extra
+observations, account mismatch, capability excess, bound currency/ratio changes,
+checkpoint substitution and duplicate JSON. They use signed native effects and
+synthetic provider bodies, not an installed transaction or transport. Independent
+process lock-wait histories and both adapter fault/race suites remain required.
+
+## Ordered initial admission design
+
+Initial effect admission uses an extension of the existing insert-once batch
+mechanism. Under the same file lock or PostgreSQL advisory transaction lock,
+the adapter checks the proposed keys for conflicts, reads a bounded dependency
+set, invokes one native decision and publishes the claim and all capacity rows
+atomically. The decision may replace only the first record's opaque bytes;
+keys, kinds, expiries and remaining capacity rows stay fixed. Existing keys
+return their original conflict index without invoking the admission decision.
+The combined insertion/dependency set retains the existing 64-record limit,
+and the same two-second fenced decision deadline applies. No nested store,
+credential or provider I/O is permitted in the decision.
+
+The native decision reuses final-entry authority/configuration validation,
+sampling time after the writer fence is held. A new record's `evaluated_at` is
+that admission time, not the earlier candidate compilation time. All current
+approval, native proof, configuration, connection and qualification constraints
+must hold before a charge is published. Subsequent entry remains a separate
+fresh check and one-use permit. Equal replay returns immutable history and does
+not rerun admission or charge again.
+
+Counter frontiers remain bounded candidate selection outside the fence. The
+insert transaction confirms target vacancy; immutable sum predecessors are
+included in the dependency snapshot and their validated cumulative amounts must
+match the proposed charge. Full native recompilation rejects a changed window
+before publication, so expired-slot collection cannot authorize an old-window
+admission. Exhausted claims are also recorded only after the same ordered
+checks. Acquisition admission additionally fences the eligible source record.
+
+
+Mutation admission has at most one claim, 34 capacity rows, 17 sum predecessors
+and three native dependencies: 55 records within the existing 64-record bound.
+The later entry boundary retains its independent 52-record maximum. Retained
+history is checked before constructing a new candidate so an expired current
+approval cannot hide an existing reservation; that handle still requires fresh
+entry validation. Runtime grant timestamps are sampled after admission rather
+than reused from the earlier candidate evaluation.
+
+`effect-admission.json` constructs real file-store ordering and atomicity
+histories with an explicit-time approval oracle. It does not establish installed
+native revocation, PostgreSQL execution or process fault qualification. The same
+adapter conformance source is available for the hosted PostgreSQL fixture.
+
+
+## Application recovery cutover
+
+The application observation socket must not turn caller-supplied recipe path
+arguments into unapproved credential-bearing reads. Remove its `read-back`
+request and CLI/SDK convenience methods directly. `observe` retains only local
+outcome and pre-entry history projections. A new closed application recovery
+frame supplies the source operation and the existing bounded native
+proof/action/exact-approval envelope; it calls the same engine recovery method
+as the operator interface and returns only the native redacted recovery report
+or native refusal. It cannot supply a URL, request body, credential or time.
+No provider body or newly signed provider value is disclosed by this route.
+
+The old observation evaluator remains a test-only reference, reachable only
+through direct tests and the private synthetic harness control plane. It is
+removed from the shared application protocol even in harness builds. General
+acquisition/disclosure and the installed approval journey remain separate
+unfinished requirements; this cutover must not present a report as disclosure
+or a provider response as a successful mutation.
+
+## Managed recovery read budgets
+
+Recovery compilation consumes the same selected native bounded branch and the
+same tightening, count and sum arithmetic as mutation compilation. Its numeric
+policy argument is exactly `maximum_response_bytes`; a sum reserves that full
+response ceiling, including lost responses. The mutation recipe's spend-sum
+requirement does not apply to the separate recovery action. Read-policy count
+and byte counters use distinct domain-separated identities from mutation
+count/amount counters, retaining namespace, subject, evaluator, partition and
+fixed-window components. Epochs and operation IDs do not reset these budgets.
+
+The existing slot candidate, slot encoding, sum-predecessor check and reservation
+membership logic is shared inside the gateway store. Admission inserts one
+acquisition journal record, its permanent source allocation and every managed
+slot atomically. A collision advances only its candidate; no partial insert or
+refund is possible. Final entry validates every slot with fresh native authority
+inside the same fence. Prerequisite reads retain their parent's declared
+allowance and do not charge the parent's managed policy again.
+
+At most sixteen bounded links produce 32 managed reservations plus two source
+reservations, preserving the 34-reservation limit. Admission requires at most
+34 inserted records, 16 predecessors, one source and three native dependencies
+(54 total); entry requires at most 38 records. Both fit the existing 64-record
+fence. Journal `/10` rejects the superseded acquisition layout. Compiler `/6`
+binds the read's managed reservation set. Independent acquisition/disclosure,
+credential replacement and installed race/fault evidence remain separate work.
+
+## Recovery credential prerequisites
+
+An independently authorized recovery read uses the same installed account and
+capability checks as mutation. Its native request commitment and human review
+include an ordered account/denied-capability acquisition plan derived from the
+pinned recipe and connection. The installation-only probe remains a separate
+onboarding obligation. The recovery action's source and recipe commitments fix
+which declarations are eligible; approval binds their complete resolved bytes.
+
+Use the existing prerequisite records, one-call/full-byte allocations, component
+keys, shared reducer and consumed transport permits. A child may name a reserved
+mutation or a reserved independent recovery acquisition; it cannot name another
+child. The parent must remain reserved at child entry. Recovery's final entry
+fences all required child records/allocations, freshness and actual account or
+capability results, together with its managed policy slots and original source.
+A failed or lost child never authorizes replayed I/O. Ancillary reads consume
+only their explicit child allowance; the parent managed read policy is not
+charged for every component.
+
+Compiler `/7` binds the complete read transport witness. Journal `/11` rejects
+old recovery plans and authenticates child membership against either closed
+parent protocol. Recovery has at most five prerequisites, for at most 48 final
+entry records including 32 managed slots and three native dependencies. The
+existing seven-prerequisite and 64-record bounds remain unchanged. No operator
+attestation is silently promoted into native effect authority: first-install
+and candidate-rotation authorization remain separate unfinished cutovers.
+
+## Candidate credential check compilation
+
+Candidate checks are a distinct acquisition target. They do not borrow the
+current writer's proof or turn an operator attestation into native authority.
+The target pins the candidate connection descriptor, expected previous record
+(or absence on first install), and exact recipe-declared checks. The native
+reserved tool `auths_gateway_candidate_check_v1` requires its own proof and exact
+human approval. Compilation performs no custody or provider I/O.
+
+The ordered provider sequence remains probe, account, denied-capability reads.
+The final declared read is the main acquisition; preceding reads are the same
+bounded prerequisite declarations used elsewhere. This keeps every child before
+parent entry without inventing an aggregate provider response. Each purpose
+retains its concrete predicate, and a complete main response will still need
+its purpose evaluated before credential publication. Recipes with no network
+checks remain an offline operator path; compilation cannot invent a read.
+
+Lifetime capacity is keyed by candidate credential identity (connection ID,
+credential generation and reference commitment), not operation ID or timestamps.
+The maximum is sixteen explicitly approved attempts, each with one main-response
+ceiling and declared one-call child allowances. Current native count/byte policies
+also apply. This checkpoint adds compilation and review only. Candidate staging,
+fresh absent/current-record fencing, shared journal admission, actual checks and
+publication are required before production onboarding can use this target.
+
+
+## Candidate acquisition admission and execution
+
+Candidate checks now use a typed plan in the existing acquisition journal.
+Their source dependency is the exact expected connection row, including a
+fenced absence for first install. The same native entry checker handles a
+published connection or a sealed candidate; candidate reads may preserve a
+disabled predecessor, but require the pinned workload/profile and refuse
+revocation or any predecessor change. Child entry checks both its reserved
+parent and the candidate's current connection dependency. Final primary entry
+also checks all retained child predicates and freshness.
+
+The source and native connection dependency can name the same physical row.
+The fence reads it once and maps that snapshot to both checks. This is index
+bookkeeping, not a second validation authority. Entry audit `/3` records a
+checked absence as null, distinct from a present record's commitment. Journal
+`/12` rejects the superseded encoding. No missing dependency is implicitly
+accepted; the typed candidate predicate alone admits first-install absence.
+
+Lifetime allocations are append-only and retain a fixed byte ceiling for one
+credential identity. A competing proposal with that identity cannot change the
+ceiling after an allocation exists. The shared read runtime executes ordered
+prerequisites, evaluates retained native evidence, consumes one primary permit
+and records response/unknown. Mutation prerequisites, recovery and candidates
+use the same helper. `acquire_candidate_credential` leases only the exact already
+staged candidate and returns redacted acquisition facts. It does not publish the
+credential. CLI staging, final-purpose validation, safely fenced publication and
+removal of the old onboarding/rotation read routes remain required.
+
+
+## Candidate credential publication
+
+Publication uses the existing insert/replace writer fence over the connection,
+root and child acquisition records/allocations and native dependencies. It is an
+explicit trusted operator action, with no application route and no new lifecycle
+state. The primary response now receives its actual recipe purpose check, just
+like each retained child. All timestamps remain original; publication cannot
+renew evidence or spend capacity again. Missing/unknown, bad and stale evidence,
+changed configuration and changed predecessor refuse the write.
+
+The runtime verifies exact staged custody, invokes this fence and reloads the
+published connection through the generation-floor owner before scheduling
+superseded credential retirement. Independent custody availability is not part
+of the store's atomicity guarantee. Every later lease still checks the exact
+credential binding. `AlreadyCurrent` is an exact-state observation for an
+explicit retry, including lost acknowledgement; it makes no causal claim and
+issues no capability. Staging/CLI cutover and installed custody/transport fault
+coverage remain separate required work.
+
+
+## Private candidate proposal persistence
+
+`auths.gateway-credential-candidate/1` stores exact canonical candidate and
+predecessor connection bytes plus the pinned recipe digest. The bounded native
+decoder reconstructs only proposal data and derives the closed check plan from
+the installed recipe. It validates the first-install or rotation relationship
+and rejects noncanonical/duplicate/unknown/obsolete data. No persisted proof of
+current custody, predecessor, authority or completed checks is implied. This
+format prepares the operator CLI to retain a stable target across approval and
+restart without adding a second execution path or serializable capability.
+
+
+## Candidate staging and operator CLI
+
+Rotation staging now records cleanup obligations before a non-overwriting
+custody replace, preserving the current connection and secret. The exact native
+proposal feeds call authoring, `effect-review --candidate`, acquisition under
+runtime read identity and publication under operator identity. Read-bearing
+recipes refuse the old unapproved rotation/commit routes; no-read recipes retain
+an explicit offline operator flow. Joining an already-published secret uses
+exact reference confirmation and does not acquire new provider evidence.
+
+Independent CLI processes exposed a local file-custody lifetime problem:
+previous handles retained snapshots that could overwrite a later successor.
+The existing persistent custody owner now uses a stable sidecar lock, reloads
+bounded canonical state under it, and fails immediately on lock contention.
+Atomic database rename remains the persistence boundary. Readers refresh before
+selecting the exact credential; a completed deletion cannot be hidden by a stale
+process cache. This is a store mechanism, with no change to credential identity,
+read authorization or publication policy. Staged/CLI fixtures cover real file
+custody, independent competing processes, held-lock refusal and stale handles;
+they do not establish AWS, PostgreSQL or full process-crash qualification.
+
+
+### First-install staging and floor binding
+
+A read-bearing first install now stores a private canonical connection draft
+before its manifest or custody mutation. The draft fixes the connection identity,
+account, descriptor and timestamps, and carries no read or publication authority.
+The existing candidate runtime derives the credential reference with the native
+custody commitment, initializes/registers cleanup notes, stores only an absent
+exact generation and binds the host floor to that unpublished generation-one
+record. The shared connection remains absent until fenced candidate publication.
+An exact custody `holds` result supports explicit retry from the same retained
+draft after an ambiguous write acknowledgement; another secret cannot replace it.
+
+`install` saves `candidate-installation.json`; `candidate-stage-installation`
+resumes from `candidate-installation.cbor` into a new private output file. Only
+this explicit operator action may initialize the installation's configuration,
+cleanup journal or floor. Candidate compilation, acquisition and publication
+check an existing floor, and ordinary entry still requires a live connection.
+There is no cross-store atomicity claim: a floor or output failure may leave
+custody with a retained cleanup obligation and no published connection.
+
+Production raw onboarding exports are removed. The previous checker remains a
+test-only reference; the pure probe predicate remains shared by managed entry.
+The offline install helper refuses recipes with reads before custody, while
+join records cleanup notes before confirming an exact already-published secret.
+The loopback option moves from unapproved install to approved candidate acquisition
+and is limited to development installations in feature-enabled builds.
+
+Native staging artifacts exercise absent connection, exact restart, conflicting
+secret, bad prefix, corrupt cleanup notes, substituted floor and runtime refusal
+to repair a missing floor. The production CLI artifact exercises actual install
+and a separate resume process. Full approved provider journeys, process faults
+and PostgreSQL evidence remain separate unfinished requirements.
+
+
+## Recipient acknowledgement ownership
+
+`effect::disclosure` constructs and verifies the bounded native
+`auths.gateway-disclosure-ack/1` statement. Recipient identity is the existing
+raw-key principal derived from the statement's nonweak Ed25519 public key and
+must equal the immutable disclosure binding. The exact protected-content length
+comes from authoritative retained data, not the ACK. Successful verification
+returns a sealed historical assertion, never a transport capability. It accepts
+no clock/current-authority input, permitting late evidence after expiry or
+revocation. This cannot authorize a new delivery.
+
+`auths-gateway-kernel::transition::valid_transition` remains the sole transition
+owner. Disclosure shares the fixed identity/allocation/revision/grant contract
+and adds a separate immutable acknowledgement field and terminal stage. It
+cannot acquire provider responses, pre-entry observations, locators or mutation
+facts. Provider journal projections supply an empty acknowledgement field.
+`exact_acknowledgement_replay` is a read-only equality predicate on consistent
+terminal disclosure history, not a second state transition rule. Persistence
+must still fence the original record; duplicate ACKs must return retained bytes
+and revision without rewriting or recharging.
+
+This checkpoint implements the evidence contract and pure relation. The
+protected-content owner, disclosure admission/entry, recipient transport and
+runtime journal projection remain required before this is a managed release
+path. Kani oracle source includes disclosure; source-linked Lean regeneration
+and refinement are still open P4 obligations, with no new proof claim.
+
+
+## Explicit content recipe compilation
+
+General content reads are an explicit digest-bound recipe capability rather
+than reinterpreted credential checks or mutation observations. The versioned
+`content` declaration is optional within the existing source envelope; absent
+recipes retain their original canonical bytes/digest and cannot compile content
+reads. The declaration has its own closed schema and complete operator review.
+`recipe::content` validates resource fields, fixed query pairs and bounded
+UTF-8/JSON-string projection. It reuses the existing path/header lowering,
+percent encoder and duplicate-rejecting provider parser. No provider-name branch
+or new runtime callback is introduced.
+
+`effect::content` participates in the same native acquisition compilation,
+review, exact approval and fresh recompile owner. Its native read call binds the
+explicit resource/account arguments and full private contract. A sixteen-entry
+source allowance excludes logical operation IDs and administrative metadata;
+it includes recipe, connection ID, credential generation/reference and exact
+resource/account arguments. The full connection record still binds approval,
+so administrative changes demand recompilation without silently refreshing that
+allowance. Fixed-window native managed bounds are additionally retained.
+
+Content projection returns private bytes and typed response errors, not a
+receipt or release grant. Runtime must authenticate/retain HTTP metadata and
+content under the journal contract and issue a distinct disclosure capability.
+The initial compiler slice supplied no admission/entry; the following section
+records that integration. Read-only installation, AP070 metadata predicates,
+recipient transport and protected projected-content custody remain required
+before migrating the maintained demo.
+
+
+## Content acquisition journal and transport
+
+The content plan is a concrete variant of the existing acquisition journal.
+Its current source is the exact connection row, checked under the same fence as
+native authority, configuration, qualification, all source/policy allocations
+and credential prerequisites. The plan validates its reconstruction data against
+the full signed request/projection commitments; decoding it grants no authority.
+The engine uses the common acquisition driver. A content permit can enter only
+the metadata-preserving content GET path and is consumed once.
+
+Content responses have a distinct private journal variant mapped to the existing
+kernel acquisition response stage. They retain bounded raw bytes and typed
+media/encoding facts, without upgrading them to usable content or a disclosure.
+Missing, duplicate, oversized and non-text header values cannot become a valid
+single header. Transport requests identity encoding and explicitly disables
+automatic decompression. Query support is confined to the compiler-owned content
+request; other transports retain their previous URL restriction. No new retry,
+transition owner, decoder compatibility path or generic request callback is added.
+
+The `content-journal.json` corpus exercises installed native entry against the
+file adapter, actual loopback HTTP, durable source exhaustion and reopened
+journal replay. Protected projected-content construction, separate disclosure,
+full engine/CLI process journeys, PostgreSQL faults and AP070 differential
+cutover remain required; this evidence grants no provider qualification.
+
+
+## Protected content in the response transaction
+
+Attempt `/15` retains the existing content response fact with one closed
+protected/refused result. Projection and key generation happen before the same
+response acknowledgement; a crash cannot leave a published content identifier
+without its retained response, or re-create a provider permit. Loading recomputes
+projection/refusal from the bound contract and retained transport facts. No
+separate content lifecycle, external write, schema migration or retry rule is
+introduced. Ready projected bytes are reconstructed only inside the native owner.
+
+The receipt disclosure protector is a storage port, and MCP execution references
+are bound to another protocol; neither supplies the required content commitment.
+The gateway therefore uses the already-pinned workspace `hmac` primitive with
+SHA-256 and the existing native canonical/domain codec. A fresh private 32-byte
+key scopes each commitment to one source effect, projection, exact length and
+private content digest. The key is stored under the same private-journal policy,
+not returned in public snapshots. The snapshot exposes only readiness/length or
+a stable refusal. This is hiding against public offline guesses while the key is
+private; it is not journal encryption or protection against malicious journal
+administration. No new cryptographic algorithm or general protection framework
+is extracted.
+
+`content-journal.json` now includes fifteen HTTP cases, empty/exact/over-limit
+content, bad UTF-8/version/encoding/media, original-byte replay, altered protected
+fields and substitution from another valid acquisition with identical content.
+The same pure projection leaf serves compiler-only and durable response paths.
+Disclosure still needs separate native authority, approval, accounting, recipient
+channel authentication and the shared final-entry fence.
+
+## Sealed source and separately authorized disclosure compilation
+
+`GatewayAttempts::retained_content` reconstructs only private provenance after
+normal journal decoding, response projection and constant-time MAC validation.
+No public plaintext accessor, provider read or delivery authority is exposed.
+`DisclosureTarget` fixes recipient raw-key identity, channel and instance over
+that source. Fresh native authorization uses a reserved release tool; exact
+approval and entry-time re-verification use the existing bound-effect owner.
+The old source proof is history, never a release permission.
+
+Count and byte units are explicitly disjoint from acquisition and mutation.
+Sixteen source-lifetime deliveries share capacity regardless of changed operation
+ID, recipient or channel. Empty content remains an actual zero-byte delivery
+with one count. Managed policy bounds use the same arithmetic leaf with closed
+protocol domains; no provider semantic dispatch or new lifecycle is introduced.
+This source/compilation slice does not create a transport or journal grant.
+The shared store must next fence retained source facts, current release authority,
+recipient authentication and these reservations at the unique delivery grant.
+
+The content journal fixture constructs both empty and nonempty sources using
+real local HTTP, then exercises native release compilation, review, fresh exact
+approval, full recompile, managed count/byte declarations, ACK binding and changed
+recipient/channel/instance/partition/operation refusals. It establishes no actual
+delivery, persistent ACK, isolated-agent or PostgreSQL claim.
+
+### Shared source allocation leaf
+
+`effect/capacity.rs` now owns the fixed source count/byte declaration arithmetic,
+with the protocol supplying its typed units. `store/capacity.rs` owns the existing
+canonical append-only allocation record, key and occupied-slot validation. The
+acquisition protocol continues to derive source eligibility, source identities,
+limits and transitions; disclosure supplies its distinct units and source keys.
+No schema, journal authority, admission decision or refund rule changes. The
+empty-disclosure ceiling convention does not alter nonempty acquisition bytes.
+This extracts the identical bounded record mechanism rather than copying the
+acquisition allocator into a second disclosure ledger.
+
+## Disclosure admission and historical ACK journal
+
+Attempt `/15` retains the closed disclosure state to the existing `JournalProtocol`
+projection. The existing kernel reducer remains the sole transition decision;
+file/PostgreSQL adapters retain their existing fenced insert/replace/CAS contract.
+The plan is data only and validates source scope, installation, projection, exact
+length, recipient key and typed reservations against the bound effect. Admission
+loads and validates the retained acquisition under the same writer fence as native
+configuration/authority. It atomically inserts one permanent source allocation
+and the existing count/sum policy slots. No disclosure ledger, callback, provider
+request or second capacity authority is introduced. Obsolete `/14` is rejected.
+
+The protocol's immutable ACK field is verified on every decode. Historical ACK
+recording needs an existing grant and the common reducer, not fresh execution
+authority. A concurrent identical ACK can be recognized after one bounded reload;
+revision and charges remain unchanged. Public snapshots and support listings keep
+provider/acquisition/recipient facts distinct and contain no private correlations.
+Engine proposal/compile/admit operations have no delivery capability or custody.
+The final recipient-authenticated grant and one-use transport use the boundary below.
+
+`disclosure-journal.json` constructs real HTTP source content then executes native
+installed file admission, duplicate claim concurrency, changed source/connection/
+approval refusal, policy/source exhaustion and reopened replay. Corrupt plan fields,
+cross-protocol records and obsolete schemas refuse. ACK cases use explicitly
+synthetic retained grants, exercise the real durable CAS/decoder/reducer, and assert
+pre-grant/foreign refusal, concurrent exact replay and late evidence after disabling
+the connection. They do not establish recipient authentication, actual send,
+independent-process isolation or PostgreSQL qualification.
+
+## Local authenticated disclosure channel boundary
+
+The first delivery transport uses an owned Unix stream, not an arbitrary callback
+or provider credential. A native channel commitment pins gateway UID, recipient
+UID and an operator-selected channel identifier. Both peers check kernel peer UID
+and their effective UID. The recipient additionally signs a freshly sampled
+challenge bound to the complete native ACK statement under a distinct
+`auths.gateway-disclosure-auth/1` domain. Authentication never signs an ACK or
+claims receipt. The owned stream and bounded authentication transcript must remain
+sealed through the final journal fence and the one-use send permit.
+
+The existing gateway admin listener already uses Tokio peer credentials. The
+exchange Unix adapter supplies peer observations but does not bind this recipient,
+effect, UID pair or fresh challenge; reusing its generic authenticated-server
+label would not establish the needed fact. Reuse the pinned Tokio/rustix/Ed25519
+primitives and native canonical/domain codec without adding a dependency or a new
+network authority framework. Native authority and exact approval still run at
+final entry; channel authentication cannot substitute for either.
+
+This is local transport under kernel/UID/socket-access isolation. A recipient must
+receive the expected effect and channel policy on its trusted private authoring
+channel and check the gateway UID before signing. A same-UID synthetic socket test
+establishes protocol behavior only, not hostile-process isolation. A malicious
+host, gateway UID or recipient key holder can bypass or forward their own data;
+that is outside this boundary. The retained MAC is opaque to the recipient: the
+trusted gateway binds its exact projected bytes to that identifier, and the
+recipient ACKs only after receiving the declared exact length on this channel.
+
+
+The concrete final grant fences the attempt, protected source, permanent source
+allocation, native connection/configuration dependencies and selected count/sum
+slots in the existing transaction. It validates full native authority/approval and
+the owned live channel, reconstructs exact content from the validated source, and
+persists entry plus authentication atomically. Only acknowledged persistence
+returns `DisclosureEntryPermit`; transport consumes it once. Neither a serialized
+transcript nor a retained grant can recreate it. The existing transition owner
+freezes `(EntryAudit, RecipientAuthentication)` as one immutable entry value.
+Attempt `/15` rejects missing/invalid authentication on a disclosure grant and
+rejects obsolete `/14`; no second reducer, store or compatible decoder exists.
+
+Framing is a four-byte unsigned big-endian length checked before allocation,
+4,096 bytes for canonical authentication/ACK and 65,536 bytes for content, with
+exact projected length. Authentication expires after five seconds; the whole
+session expires after fifteen seconds, including final fence wait and send/ACK.
+A transport timeout does not refresh native evidence or refund reservations.
+The gateway deployment clock remains the final authority/freshness owner.
+
+`disclosure-delivery.json` is constructed from actual socket pairs and native
+installed file grants over retained HTTP source content. It exercises actual
+receipt/ACK, loss of ACK with reopened no-redelivery and late refinement, dropped
+permits, competing authenticated channels, and connection/configuration/approval
+changes after authentication. Authentication substitutions, malformed frames and
+corrupt retained authentication refuse. This evidence is separate from the older
+synthetic-grant corpus; neither establishes independent-process containment,
+PostgreSQL parity, installed UI/bindings or completion of AP074 acceptance cases.
+
+
+## Content application and recipient-client boundary
+
+Four closed application schemas route acquisition, separate disclosure, local
+history and exact late ACK to the existing engine/journal. Content's resource
+arguments now come from its native-verified action. The existing acquisition
+compiler is split at the sealed verification result so target construction and
+binding use that same result without an additional verifier or weaker map. Native
+fixture parity compares this path with explicit target compilation. The initial
+verification precedes target validation; final entry recompilation remains the
+same full native check.
+
+The disclosure control path invokes native proposal/admission and emits either
+redacted history/refusal or Ready. Ready transfers the owned stream to the native
+engine and authenticating transport; it confers no entry permission. Current
+compilation/admission run again after the protocol switch, and the existing final
+writer fence still exclusively grants delivery. This bounded repetition avoids
+exporting an approved or live-channel capability to a generic application handler.
+The transport also honors the enclosing application's deadline. Lost control or
+ACK responses never schedule another read or delivery.
+
+The native recipient client checks UID/channel/key before sending its private
+request. CLI receipt writes new 0600 content and ACK files, syncing file and
+parent, before sending the ACK. It reports only that the recipient sent evidence;
+retained ACK persistence is queried separately. Native history restricts recipe,
+namespace, partition and epoch and exposes only typed redacted projections. Thin
+Python/TypeScript control methods pass proof/action/approval/ACK bytes without
+hashing, signing, recovery or accounting. No non-native recipient protocol is added.
+
+Application frame/fuzz fixtures exercise real sockets with a counting refusal
+oracle. The three control CLI cases capture exact envelopes against real refusal
+sockets. Native socket/file delivery cases now use the production recipient client
+with an explicit control-frame shim; the native installed file grant still uses
+real approved authority. These are not a complete installed app dispatcher,
+recipient CLI process, language runtime, authenticated UI or hostile-agent claim.
+
+
+## Native authoring and process evidence
+
+Content/disclosure call-authoring APIs reuse `ContentReadTarget` and
+`DisclosureTarget` with installed recipe/connection and retained provenance.
+The CLI does not construct contracts or hashes. Content JSON reuses the existing
+65,536-byte duplicate-rejecting parser as a mechanical leaf, then the native
+recipe validates its closed resource fields. These operator-only authoring
+surfaces return canonical private data, never an execution capability.
+
+The real daemon/recipient CLI artifact now covers acquisition, disclosure and
+ACK, restart replay without redelivery, an approved mutation and independent
+provider-linked observation. A separate test recipient uses the shipping native
+client to retain ACK evidence without sending it; both peers are killed, and
+reopen preserves unknown charges until exact late ACK. A separate operator
+process disables before a further release and no bytes are delivered. These
+histories use the shared file adapter and public fixed test signing keys.
+Authenticated UI, installed language packages, complete fault/concurrency tests,
+PostgreSQL and hostile process isolation remain distinct incomplete obligations.
+No production transition predicate or wire schema changed in this increment.
+
+
+## Bounded expiry maintenance and retained replay facts
+
+The existing store remains the only journal and reservation owner. Count/sum
+slots expire one complete window after the corresponding window ends. Attempts,
+configuration, commissioning budgets and lifetime effect allocations do not
+expire and are never eligible for this sweep. Full retained attempt records
+serve as replay tombstones; no compressed tombstone decoder or alternative
+admission path is added. Total journal storage may grow with retained operations.
+
+The file adapter now inspects at most 4,096 directory entries in a sweep and
+checks the existing two-second cooperative work deadline between entries. It
+retains one transient directory iterator, so later calls can advance past
+permanent and unexpired records without collecting or sorting the directory.
+All inspection and deletion participate in the common writer fence. Reopening
+loses only scan progress, never journal authority. Each record read remains
+byte-bounded; physical filesystem I/O/sync latency remains a host assumption.
+Deletion of expired slots is directory-synced even if a later read fails.
+The indexed PostgreSQL sweep keeps its existing writer fence and bounded LIMIT.
+
+The engine supplies the same deployment clock used for native effect evaluation;
+required qualification with untrusted time refuses cleanup before storage work.
+An independently sampled wall clock must not age out a still-current effect
+window. This is not a new clock rollback detector or a whole-store rollback claim.
+
+Read-only file inventory is bounded separately from expiry maintenance. It
+retains at most 4,096 candidate keys in a max-heap, inspects at most 16,384
+directory entries plus one overflow probe, and observes the same cooperative
+two-second work budget under the shared fence. Successful results retain the
+original globally sorted smallest-key prefix. Larger output requests, excessive
+directory work, unreadable entries or an expired work budget return unavailable;
+they never return an apparently complete partial inventory. Irrelevant filenames
+consume scan work too. Known-operation lookup, admission, replay and accounting
+do not consult this inventory. The existing support-bundle consumer represents
+an unavailable inventory as unavailable, distinct from a successfully empty list.
+PostgreSQL keeps its existing indexed bounded query. A large file archive may
+therefore require direct known-key inspection or the indexed adapter for a
+complete inventory; no journal record is removed to make a query fit its budget.
+
+Native file/PostgreSQL evidence acquires actual HTTP content using signed
+count/byte bounds and exact approval, then optionally performs authenticated
+Unix disclosure with ACK or ACK loss. Slots remain before their exact expiry
+and are removed at expiry while the original approval remains current. Replay
+returns the same original facts without new entry, recipient authentication,
+delivery or charges; another operation cannot consume the old approval. A
+reopened engine preserves those facts after full authority expiry. Separate
+adapter controls retain every nonexpiring kind, bound scan progress and refuse
+corrupt slots. Execution status belongs in the AP074 ledger; epoch/floor,
+reserved/mutation retention and remaining fault cases are not inferred from
+these read/release histories.
+
+## Standalone read-only recipe decision
+
+A source `/3` with no `write` declaration is a content-only installation. It must
+name the reserved native content tool and declare content; mutation-specific
+observation, echo, pre-entry, precondition, relative-ceiling and spend-sum fields
+are refused. Its profile lock names only the existing binding fields, resource
+and optional account field. Existing credential guards and registered headers
+retain their owners. No fake write URL, body or mutation permission is installed.
+
+The compiler lowers this into a closed `ContentOnly` execution alternative,
+which contains no `WritePlan`. Mutation request construction rejects that
+alternative before constructing a request. Content acquisition, candidate reads,
+release, approval and journaling remain existing native paths. Wire parsing may
+represent an absent write, but one compiler validates the relation and the sealed
+compiled execution kind is authoritative. This is an extension of source `/3`,
+not an old-schema reader or parallel runtime mode. Existing mutation canonical
+bytes and semantics must stay unchanged. Operator review represents no mutation
+with null method/mutation recovery, empty write path and zero write-body bound, while rendering
+the complete declared content GET separately. No `GET` is added to write methods.
+
+The required evidence is a bounded source/lock negative corpus, native
+mutation-refusal and real read-only install/acquisition/disclosure process
+journey. Google metadata, token refresh custody and AP070 owner migration are
+deferred integration work. Supporting a content-only recipe establishes neither
+those optional provider capabilities nor AP070 equivalence.
+
+## Native fixed-window contention and retained reservation evidence
+
+The existing native admission owner is exercised by independent OS contenders
+for mutation count/amount-sum and disclosure count/byte-sum capacity on file and
+TLS PostgreSQL. A prefix leaves exactly one unit; two separately approved native
+operations contend and exactly one wins. The counting HTTP provider performs no
+idempotent deduplication. Native recipient authentication and ACK retain typed
+disclosure facts, with no provider refetch or credential lease at release. Each
+disclosure contender hosts gateway/recipient tasks over same-UID Unix sockets;
+this is independent admission-process evidence, not hostile-process containment.
+
+The short signed window expires while original proof and approval times remain
+eligible. Native recompilation selects a new window-bound identity; old exact
+approval cannot authorize it. Cleanup removes only old fixed-window slots and
+keeps original attempt/allocation bytes. Terminal replay and reserved restart
+create no new grant, send, recipient bytes or recorded charge. Mutation reserved
+histories include SIGKILL after actual reservation acknowledgement. Disclosure
+reservation-only admission returns data and no permit. These tests add no
+production owner or new semantic abstraction. Source-specific adverse schedules,
+rollback floors, installed containment and final hosted evidence remain required.
+
+## Native unscoped contract and partition evidence
+
+The native verifier and managed compiler retain different responsibilities.
+Freshly signed acquisition and disclosure calls can be valid native authority
+while proposing unsupported task, participant, ancestor or charge inputs, or
+substituting the closed compiler contract. The managed compiler refuses these
+calls before admission. File and TLS PostgreSQL histories inspect unchanged
+attempt bytes and execution witnesses, then execute the original exact read and
+authenticated same-UID disclosure/ACK as positive controls.
+
+Two independently configured operator partitions share the physical adapter but
+have different effect identities and control keys. The same valid proof does not
+let an exact approval cross partitions, nor let a partition read or release the
+other's retained content. Original history remains replayable without new entry.
+An activated replacement recipe also requires an independently approved native
+connection rotation/CAS; activation alone cannot reuse the old descriptor. Even
+fresh proof and approval cannot overwrite the original operation. A separately
+approved new operation remains possible under the replacement. The proposal
+getter now checks the retained recipe digest before returning provenance; the
+compiler and entry fence retain their own existing installation checks.
+These tests exercise the present single-agent, unscoped protocol; they introduce
+no task protocol, alternate admission authority or AP073 dependency. Installed
+language framing, all binding dimensions and hostile OS containment remain
+separate acceptance obligations.
+
+## Mutation dependency order and wait-time validation
+
+Native Airtable mutation histories now schedule independent worker and operator
+processes against each maintained adapter's actual common writer fence. Connection
+revocation and configuration activation are ordered before versus after the final
+grant. Independent approval and native authority deadlines have eligible and
+expired controls while the worker demonstrably waits on a physically different
+connection record. Only the existing native callback computes and persists the
+grant; the scheduling wrapper supplies neither record bytes nor a decision.
+
+Refused final entry retains the original reservation with zero HTTP writes. A
+grant that linearizes first permits one exact captured write; terminal replay
+after reopen creates no new custody or transport calls. The test explicitly
+counts the custody lease that preceded the final fence; refusal does not undo it.
+This extends the acquisition fence evidence using the same test-only scheduling
+helper and adds no production owner. Selected observer/threshold timing,
+disclosure races, installed languages and OS containment remain separate work.
+
+## Disclosure dependency order and wait-time validation
+
+The same eight order/time schedules now exercise separately approved native
+disclosure from an actual HTTP acquisition. Independent worker/operator processes
+order the actual adapter fence; gateway and recipient tasks inside the worker
+use real same-UID Unix sockets. Recipient authentication completes before the
+native final-grant pause. Withdrawal or expiry that wins retains a reservation
+and produces no recipient content or ACK. A grant that wins yields exactly the
+retained projection and an exact native recipient ACK. Historical replay after
+reopen emits no authentication/content frames or new custody/provider calls.
+Attempts and lifetime delivery/byte allocations remain retained on refusal.
+An opaque same-process socket relay independently captures all gateway bytes.
+After the native authentication challenge, refusal must produce zero wire bytes,
+including partial frame prefixes; success must carry exactly one length-prefixed
+content frame with the original bytes. The relay forwards bytes without changing
+native decisions or authentication and provides no hostile-channel qualification.
+
+These file/TLS PostgreSQL histories establish native entry ordering, not hostile
+recipient process containment, separate recipient OS processes within each worker,
+selected observer/threshold timing or installed binding/UI coverage. Existing
+native authority, approval, recipient authentication and journal owners remain
+unchanged; the test scheduler constructs no grant or decision.
+
+## Selected observation and threshold parity at native entry
+
+Mutation and disclosure each extend their entry-order matrix with the existing
+native acquisition observation/threshold controls. At the inclusive five-second
+freshness boundary, a single signed observation and a genuine 1-of-2 proof retain
+the approved native selection. At six seconds the observation-only proof is stale;
+the threshold proof remains Authorized through its long-lived branch, while the
+native selected-authority commitment changes. Fresh native compilation establishes
+that only the bound authority changes, and the old approval's own signature/time
+remains valid for the original effect. It cannot authorize the new effect.
+
+Actual worker/operator lock waits re-run that native check at the final grant.
+Refusal retains original reservations and sends no mutation request or disclosure
+content/ACK. Positive controls perform the exact separately reviewed request or
+recipient delivery. Changing a mutation's signed proof also changes its declared
+provider echo, so the expected captured body comes from that proof's own native
+bound request. No approval, request body, input timestamp or grant is synthesized
+by the scheduling helper. The fixtures use explicitly public test observer keys
+and signed observation attachments; they do not qualify live-provider evidence.
+Installed language/UI parity, other authority/revocation scopes and hostile OS
+containment remain separate obligations.
+
+## Typed native admission rows roll back together
+
+Server-side insert exceptions now exercise every row position of the native
+mutation and disclosure admission batches. A nontransactional sequence in each
+owned fixture schema witnesses the actual rejected prefix; complete gateway-table
+counts and raw slot lookups verify rollback before any credential or transport
+entry. Mutation keeps its distinct attempt/count/amount-sum transaction. Disclosure
+adds its separate source-lifetime allocation and byte units, while its original
+acquisition and source charges remain byte-identical. No generic lifecycle or
+alternative admission authority is introduced.
+
+The sequence witness checks `is_called` as well as `last_value`. A fresh sequence
+can report its initial value before any insert; that state must never count as
+evidence that the first native admission row reached the trigger. This stronger
+oracle applies to the existing acquisition fixtures and both new typed matrices.
+
+Removing the owned test trigger permits explicit native reauthorization with
+unchanged proof/action/approval. Mutation sends one exact compiled request;
+disclosure authenticates a separately spawned native recipient over a real Unix
+endpoint and records its exact ACK. Terminal replay changes no inspected rows or
+execution witness and sends zero bytes. Failed release admission sends no native
+authentication challenge either. These tests use the maintained TLS transaction
+adapter and existing testkit setup/authoring, with no shipping fault branch.
+Commit acknowledgement ambiguity, actual durability/power faults, full file I/O
+matrices and hostile OS/provider qualification remain separate obligations.
+
+## Real lost COMMIT replies preserve typed uncertainty
+
+The existing acquisition COMMIT harness now observes typed mutation and
+disclosure record paths and exact operation keys. Its native store wrapper
+forwards unchanged native decisions and counts actual adapter errors. An opaque
+proxy never terminates TLS; a separate direct TLS observer waits on the selected
+backend's deferred transaction trigger, verifies its SSL session and independently
+observes commit versus backend termination/rollback. Snapshot keys include record
+kind as well as storage key so distinct record families cannot collapse together.
+No production error, clock, grant, record byte or authorization decision is faked.
+
+Six mutation histories cover admission, final grant and response persistence.
+Eight disclosure histories cover admission, final grant, exact ACK and Unknown
+persistence, with separate native gateway/recipient processes and opaque Unix
+wire capture. A committed grant with a lost reply never recreates a permit, even
+if actual send count is zero. Absent/reserved work requires native reauthorization
+to continue. Existing count/sum/lifetime rows remain byte-identical and terminal
+replay makes no new custody/provider calls or authentication/content frames.
+
+An aborted ACK transaction is followed by the existing native Unknown transition;
+an ACK that committed despite reply loss stays acknowledged. A recipient that
+received content can supply its genuine historical signed ACK later, including
+when its original ACK was deliberately withheld to exercise Unknown persistence.
+That action refines history and preserves charges without delivery. No saved ACK
+is invented when a lost grant sent no content. This is native transaction/transport
+evidence under fixture keys, clocks and same-UID local process policy; file faults,
+server/power loss, hostile OS containment and live-provider qualification remain
+separate work.
+
+## Native file-error parity across typed protocols
+
+The existing acquisition file-error selector now accepts an exact operation key
+and typed native stage path. It still forwards each native decision unchanged and
+arms only the maintained adapter's test-only syscall-error site. Fourteen mutation
+and eighteen disclosure schedules cover batch data sync/publication, every target
+install, all three admission directory-sync points, and data-sync/replace/directory
+sync errors at grant and protocol outcome persistence. Failed syscall returns are
+simulated; actual preceding writes, writer fences, rollback recovery, native
+authorization and transport execute unchanged.
+
+Mutation atomically retains or removes attempt/count/amount-sum rows and has no
+disclosure lifetime allocation. Disclosure preserves its actual source and charges
+while treating ACK and Unknown as separate facts. A native recipient in another
+OS process may deliberately withhold its ACK after exact content receipt; its
+genuine signed ACK is retained for historical refinement. An opaque Unix relay
+measures all gateway content frames. Lost grants never recreate permits or send
+content, even with zero prior delivery. Absent/reserved work requires native
+reauthorization; terminal replay and historical ACK refinement perform no new I/O
+or charging. Existing charge bytes remain intact. These test-only additions do
+not change a shipping owner, and establish neither actual kernel/device EIO,
+power-loss/whole-store rollback safety nor hostile OS/provider qualification.
diff --git a/docs/assurance/PROOF_COVERAGE.md b/docs/assurance/PROOF_COVERAGE.md
index 5b207aeb5..ec97ff428 100644
--- a/docs/assurance/PROOF_COVERAGE.md
+++ b/docs/assurance/PROOF_COVERAGE.md
@@ -7,7 +7,7 @@
At the commit that contains this document, 110 of 1,031 decision-scope functions (10.7%) and 1,289 of 11,353 code lines (11.4%) are translated to Lean and reached from audited theorem statements, and 15 of 82 kernel check sites map only to such functions. The rest of the decision path, including the staged verifier control flow, the composition evaluator, the registry handlers, the codec, and the signature suites, is tested, not proved.
- Revision: the commit that contains this document and `formal/proof-coverage-v1.json`; a release binds it in its assurance evidence.
-- Measurement inputs: sha256 `56f5f8df74c31a2554725ec967451d36bfa6e7c4a49004b3c83bb5cd32c35f52` over the 89 files listed in `formal/proof-coverage-v1.json`.
+- Measurement inputs: sha256 `ac052be07bf5d42e8b136ee8c1e0cf94f0cc3253f24c787165acc5b8607e3781` over the 89 files listed in `formal/proof-coverage-v1.json`.
- Tool: `cargo xtask formal coverage`, version 1.
- Toolchain pins: `formal/lean-toolchain` = `leanprover/lean4:v4.31.0`; `formal/translation-toolchain.lock` sha256 `0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53`.
- Role table: `formal/coverage-scope-v1.toml`, sha256 `053f0be216b0003768d422d9ccc92a692726488874284fa80615657b30a495b5`. Figures computed under a different role table are not comparable.
@@ -52,7 +52,7 @@ Files in scoped crates with a non-decision role (every other scoped file is deci
| --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
| decision | 1,031 | 110 (10.7%) | 56 (5.4%) | 110 (10.7%) | 11,353 | 1,289 (11.4%) | 856 (7.5%) | 1,289 (11.4%) |
| verifier-core | 785 | 86 (11.0%) | 41 (5.2%) | 86 (11.0%) | 8,802 | 907 (10.3%) | 558 (6.3%) | 907 (10.3%) |
-| translated-crates | 1,138 | 210 (18.5%) | 70 (6.2%) | 210 (18.5%) | 10,662 | 2,508 (23.5%) | 1,024 (9.6%) | 2,508 (23.5%) |
+| translated-crates | 1,146 | 213 (18.6%) | 70 (6.1%) | 213 (18.6%) | 10,738 | 2,497 (23.3%) | 1,027 (9.6%) | 2,497 (23.3%) |
| verifier-closure | 1,083 | 86 (7.9%) | 41 (3.8%) | 86 (7.9%) | 13,762 | 907 (6.6%) | 558 (4.1%) | 907 (6.6%) |
| decision-nontrivial | 443 | 81 (18.3%) | 44 (9.9%) | 81 (18.3%) | 9,589 | 1,202 (12.5%) | 820 (8.6%) | 1,202 (12.5%) |
@@ -186,7 +186,7 @@ None: every function Aeneas translates is in the refined closure.
`cargo xtask formal mutations` runs study 006's three campaigns: S1 applies each case of `formal/refinement-mutations-v1.json`, S2 mutates comparison, Boolean, and result-position operators in the model's `def` bodies, and S3 replaces each translated function's body with `fail .panic`. A mutant is killed when the build of the audited targets fails, and stillborn when its first error lies in the mutated definition itself. The first error (the earliest in the mutated file, else the least location among the failing modules, so that parallel builds record the same one) and its declaration are recorded; an S1 mutant whose build reports no error in its case's named witness is reported as killed elsewhere. Every survivor must be fixed or explained before a release; pull requests are not blocked.
-**Some results predate the current formal sources** (current formal inputs `f250c011aa84a978f1f65e91ead3ed8aee7e3b6502e7d80868a583381df305db`); re-run the stale campaigns before a release.
+**Some results predate the current formal sources** (current formal inputs `9f655722b5ba3e4322f25ffd60c37fa9149373099562a28a8faa08f3a9848169`); re-run the stale campaigns before a release.
| Campaign | Mutants | Killed | Of which killed elsewhere | Survived | Stillborn | Current |
| --- | ---: | ---: | ---: | ---: | ---: | --- |
diff --git a/docs/operations/GATEWAY_PRODUCTION_RUNBOOK.md b/docs/operations/GATEWAY_PRODUCTION_RUNBOOK.md
index c42db3210..c60bd4c00 100644
--- a/docs/operations/GATEWAY_PRODUCTION_RUNBOOK.md
+++ b/docs/operations/GATEWAY_PRODUCTION_RUNBOOK.md
@@ -42,6 +42,46 @@ Active or disabled connections keep the normal floor requirement. If no
connection committed, retain the journal as an orphan cleanup obligation;
never infer deletion from a failed install command.
+## First installation with credential checks
+
+For a recipe declaring credential reads, `install` retains private installation
+inputs and writes `candidate-installation.cbor` and `candidate-installation.json`
+under the state directory. It stages the secret, cleanup note and exact host
+floor, but does not publish a connection or contact the provider. Continue with
+steps 2–5 of the candidate workflow below, using that proposal. Serving ordinary
+operations requires successful publication.
+
+If installation stops after retaining the draft, resume explicitly under
+operator identity with the same secret:
+
+```sh
+credential-source | auths-gateway candidate-stage-installation \
+ --state-dir /private/installation --credential-stdin \
+ --proposal-out /private/resumed-candidate.json
+```
+
+The output must be a new absolute owner-private file. Preserve an existing
+proposal; do not overwrite it to conceal a failed or ambiguous attempt. Restaging
+reuses the draft's exact connection identity and timestamps, confirms an existing
+matching secret or stores an absent one, and refuses conflicting custody/floors.
+A missing runtime floor is a refusal, never implicit repair. This explicit
+operator action may initialize the unpublished installation floor; it cannot
+stage after a connection has been published. A failed stage can retain custody
+and cleanup notes without a connection; retain that orphan cleanup obligation.
+
+Development provider doubles are selected only on `candidate-acquire
+--loopback-provider PORT` in a feature-enabled development build. `install`
+no longer accepts a provider transport option. Recipes without credential reads
+keep the offline installation flow.
+
+The local `candidate-process-journey.json` artifact exercises these commands
+through separate production CLI processes and a counting HTTP double, including
+active/disabled rotation, replay, failed credential predicates and response loss.
+Its reviewer uses public test keys; it is not a live-provider, human-presence,
+production-custody or PostgreSQL qualification. In a production installation,
+the required signed qualification inputs must permit the candidate reads before
+acquisition. Staging alone does not bypass that gate.
+
## Provider-secret rotation
For the web-identity reference, the operator configuration must explicitly
@@ -51,11 +91,42 @@ the write-only operator identity. The operator process reads with the former
and creates/deletes with the latter; the serving process keeps only its
read-only identity. Neither role needs broader permissions.
-Use the operator identity and `rotate-prepare --operator-process
---credential-stdin --state-dir ...`, piping the new provider secret. It runs
-the recipe's fixed credential checks and stores a successor without publishing
-it. Retain only the returned reference commitment. Then use `rotate-commit
---operator-process --state-dir ... --commitment ...` to atomically publish it.
+For a recipe with credential reads, use the approved candidate workflow:
+
+1. Under operator identity, pipe the secret to `candidate-stage-rotation
+ --credential-stdin --state-dir ... --proposal-out /private/candidate.json`.
+ The absolute output must be new. Staging writes cleanup notes and custody,
+ preserves the live record, and performs no credential lease or provider read.
+2. Run `effect-candidate-call --state-dir ... --candidate /private/candidate.json
+ --operation ... --call-out /private/call.json`. Use the normal native authoring
+ flow to obtain proof/action for this exact call. The proposal is private data,
+ not authorization.
+3. Run `effect-review --candidate /private/candidate.json --state-dir ...
+ --proof ... --action ... --not-before ... --expires-at ... --request-out ...`.
+ Review the actual requests and connection/predecessor, obtain the reviewer's
+ signature, and assemble the exact envelope with `effect-approval`.
+4. Under runtime read identity, run `candidate-acquire --state-dir ...
+ --candidate /private/candidate.json --proof ... --action ... --approval ...`.
+ It returns retained response/unknown facts. An unknown result does not permit
+ blind retry or publication.
+5. Under operator identity, run `candidate-publish` with the same state directory,
+ proposal, proof, action and approval. The connection changes only if fresh
+ native authorization and every retained credential predicate pass under the
+ writer fence. Evidence expires 30 seconds after its original grant, including
+ earlier children. `AlreadyCurrent` reports exact state without attributing a
+ prior publication to this invocation.
+
+Keep the exact private proposal across restart. A different secret at an already
+staged generation is refused; staging never deletes it implicitly. Local file
+custody serializes independent processes with a stable lock and refreshes state;
+lock contention returns unavailable without a write. The current read-bearing
+rotation API refuses `rotate`, `rotate-prepare` and commitment-only
+`rotate-commit` with `gateway.admin.candidate-approval-required`. Those commands
+remain the offline operator flow for recipes declaring no credential reads.
+Joining an already-published connection confirms the exact committed secret
+without acquiring new provider evidence. Full installed provider/custody
+qualification remains unfinished AP074 work.
+
Check status on both hosts; each must see the same shared generation and hold
the exact committed credential generation. The old generation remains for at
least the fixed 20-second retirement delay. Run `credential-collect --state-dir
@@ -151,17 +222,30 @@ and rebuild authority/connection state under owner review; do not serve old
operations as fresh. A newly created host without a floor cannot prove that a
database is current; recover its floors before starting it.
-For each response-recorded or unknown operation, use `reobserve --state-dir
-... --operation-id ...` once. It uses the stored plan and a fresh read-only
-check under the declared capability. Reobserve requires an enabled connection
-because it leases a credential; a disabled or revoked connection refuses even
-this read. Only when the restored state passes the continuity and retained-floor
-checks, keep app ingress blocked on both hosts, require every other readiness
-check to pass, enable and rerun full doctor before reconciling. Do not enable
-a quarantined restore to work around a refusal. Confirmed evidence may advance the
-record; missing or delayed evidence stays unknown. Reobserve again after
-visibility returns; never issue a new provider write as recovery. Keep the
-opaque operation digest and stage only in support evidence.
+A mutation submission stops after recording its response; it does not perform
+an automatic provider observation. A `response-recorded` result is not a
+confirmation of the intended provider state.
+
+For a response-recorded or unknown operation, author a distinct bounded read
+using `effect-recovery-call`, authorize it through the native MCP flow, review
+it with `effect-review --recovery-of`, and sign its exact approval. Then use
+`recover --state-dir ... --source-operation ... --proof ... --action ...
+--approval ...`. The old unapproved `reobserve` command is refused. Each read
+reserves capacity permanently from the source's 16-attempt count/byte limit.
+Changed credentials and unimplemented ancillary/read-policy bounds fail closed.
+
+A new read requires an enabled connection. Only when restored state passes the
+continuity and retained-floor checks, keep app ingress blocked on both hosts,
+require every other readiness check to pass, enable and rerun full doctor before
+reconciling. Do not enable a quarantined restore to work around a refusal.
+`recovery-record --source-operation ... --acquisition-operation ...` can record
+fresh evidence already acquired; it performs no provider or credential I/O and
+does not borrow the writer's authority. New evidence must be no older than 30
+seconds at the journal fence. Stale, missing or inconclusive evidence leaves the
+source unresolved and charged. A repeated recorded fact is history with its
+original timestamp. No recovery command authorizes another mutation send.
+Retain only native redacted reports in support evidence. The complete installed
+AP074 recovery journey and fault evidence remain under implementation.
## Rolling upgrade
diff --git a/docs/operations/GATEWAY_TRUST_AND_GIT_SIGNING_RUNBOOK.md b/docs/operations/GATEWAY_TRUST_AND_GIT_SIGNING_RUNBOOK.md
index b5c69dffa..ed80bd8a6 100644
--- a/docs/operations/GATEWAY_TRUST_AND_GIT_SIGNING_RUNBOOK.md
+++ b/docs/operations/GATEWAY_TRUST_AND_GIT_SIGNING_RUNBOOK.md
@@ -18,8 +18,103 @@ A production deployment keeps three principals apart.
| Operator | The gateway host, its admin socket, and the provider credentials | A signed operator attestation at install |
| Observer | The key that signs what the gateway saw, in qualified custody | Observer anchors of the trusted context |
+AP074 installations require an explicit effect policy. Generate its canonical
+bytes with native code and the human approver's public Ed25519 key:
+
+```text
+auths-gateway effect-policy --partition --epoch 1 \
+ --approver-public-key <64-hex-public-key> > effect-policy.json
+```
+
+The policy contains no signing key. Protect it as operator configuration. Never
+choose a new epoch merely because a journal is missing. The same policy must be
+passed to `operator-request` and `install`; the operator signature binds its
+partition, epoch and approver. Installation schema `/6` and operator attestation
+schema `/2` reject the obsolete formats; reinstall disposable prelaunch state.
+This configuration is a foundation for exact-effect approval, not evidence that
+transactional entry fencing has been completed.
+
+The application submission schema is now `auths.gateway-submit/2`; it requires
+canonical native exact-effect approval in addition to proof and action. The
+commissioning path requires the same approval. Resource-authority approvals in
+the proof remain separate requirements and cannot substitute for it.
+
+The private operator flow is:
+
+```text
+auths-gateway effect-review --state-dir --proof proof.cbor --action action.cbor \
+ --not-before --expires-at --request-out
+auths-gateway effect-approval --request --signature-hex
+auths-gateway submit --app-socket --proof proof.cbor --action action.cbor --approval
+```
+
+`effect-review` emits private JSON containing the native review, signing request
+and `signing_message_hex`. The external approver signs those exact decoded
+32 message bytes with Ed25519; do not independently hash or reconstruct them.
+Keep review/request/approval files on the authorized private channel. The
+`effect-approval` command only assembles canonical data; submission independently
+checks the installed signer, effect binding and validity interval. Missing,
+obsolete or invalid approvals are refused before credential leasing.
+
+For an independently authorized recovery read, first obtain the native private
+call from an observable retained source, then authorize that exact call through
+the native MCP authoring flow. Use a new logical operation ID for the read:
+
+```text
+auths-gateway effect-recovery-call --state-dir \
+ --source-operation --recovery-operation
+auths-gateway effect-review --state-dir --recovery-of \
+ --proof read-proof.cbor --action read-action.cbor --not-before \
+ --expires-at --request-out
+```
+
+These are private authoring/review operations. They perform no credential lease,
+provider call or admission. The writer's proof or exact approval cannot serve as
+the read's authority. Source and current connection/generation must match in this
+compiler; changed credentials and managed read-policy bounds currently refuse.
+A compiled read cannot be sent through mutation `submit`. The native runtime now
+has a separate recovery acquisition entry point with permanent count/byte
+allocation and a durable one-use grant. It refuses recipes needing additional
+credential reads until those reads are accounted for. The commands above still
+perform authoring/review only. After signing the exact review request and
+assembling `read-approval.json`, the owner-only operator socket accepts:
+
+```text
+auths-gateway recover --state-dir --source-operation \
+ --proof read-proof.cbor --action read-action.cbor --approval read-approval.json
+auths-gateway recovery-record --state-dir \
+ --source-operation --acquisition-operation
+```
+
+Mutation submission records the response and stops without a provider read-back.
+It cannot report newly observed provider state from the write response alone.
+
+`recover` performs the independently authorized read and then attempts evidence
+refinement. `recovery-record` only uses an already retained acquisition response;
+it performs no credential or provider I/O. The original writer's expired proof
+is never reused as read authority. Freshness is checked under the journal fence:
+evidence older than 30 seconds, or with a future receipt time, cannot newly
+refine the source. A previously recorded fact remains history and does not get
+a new time on replay. Unknown sends remain unknown without supported linked
+evidence; a state match alone does not prove the original write occurred.
+
+Admin request/response `/2` removes the unapproved `reobserve` command. Native
+recovery report `/1` separates read and mutation facts and excludes provider
+content, locators, echo and private commitments. Installed transport/fault
+coverage, the authenticated approval frontend, ancillary acquisition cutovers
+and full acceptance evidence remain open AP074 requirements.
+
+Compiler `/8` exposes native candidate-check compilation for an exact staged
+first-install or rotation descriptor. The corresponding engine acquisition API
+now uses journal `/15` and entry audit `/3`: predecessor/absence, current native
+authority and all earlier check evidence are fenced before reads. It leases only
+the exact already-staged credential and returns redacted acquisition facts.
+This API does not publish credentials. The onboarding/rotation CLI staging and
+publication cutover is still pending; a recorded HTTP response alone does not
+mean that the final check passed or that a rotation is authorized.
+
A production installation names its operator only through a signed
-`auths.gateway-operator-attestation/1` statement. `operator-request` prints
+`auths.gateway-operator-attestation/2` statement. `operator-request` prints
the statement for the installation and the preimage the operator's own
signer signs; the operator assembles the statement, the base64url signature,
and at most four control-evidence objects into a file of at most 16 KiB,
@@ -27,10 +122,10 @@ readable only by its owner:
```text
auths-gateway operator-request --recipe recipe.json --profile-lock profile.lock.json \
- --trusted-context trusted.context.cbor --provider --alias \
+ --trusted-context trusted.context.cbor --effect-policy effect-policy.json --provider --alias \
--deployment production --operator-principal \
--principal-method did-key-v1 --verification-method --signature-suite ed25519-v1
-auths-gateway install --deployment production --operator-attestation operator-attestation.json ...
+auths-gateway install --deployment production --effect-policy effect-policy.json --operator-attestation operator-attestation.json ...
```
`install` and every `serve` start verify the attestation as the kernel
@@ -90,16 +185,26 @@ held which key, is operator evidence and is not produced by this code.
`AUTHS_POSTGRES_URL`, `AUTHS_POSTGRES_CA_PEM`, and
`AUTHS_POSTGRES_SERVER_NAME`. Connections are TLS-only with
certificate and server-name verification. The schema is
- `auths.lifecycle.postgresql/6`; it installs only into an empty database.
- A database created at schema 4 or earlier is disposable prelaunch state:
+ `auths.lifecycle.postgresql/8`; it installs only into an empty database.
+ A database created with an obsolete schema is disposable prelaunch state:
recreate it. Gateway claims are stored as attempt record
- `auths.gateway-attempt/3`; a store holding `/2` records is refused as
- corrupt, so the file store's attempts directory is recreated too.
+ `auths.gateway-attempt/16`; a store holding obsolete records is refused as
+ corrupt, so the file store's attempts directory is recreated too. The journal
+ keeps typed mutation/acquisition facts under one header. Recovery allocations
+ are permanent source-lifetime count/byte reservations and are never swept.
+ Support bundle `/2` reports protocol and stage separately.
- **Several gateway processes** may serve from the same database. Each
logical operation is claimed once, by an insert-once row; stage changes are
compare-and-swap on the exact stored record. A second process that loses a
- race records nothing and answers `gateway.attempt.replay`. Gateway claims
- do not take the lifecycle store's singleton contract-row lock.
+ grant race records nothing and loads the matching journal state. An equal
+ binding with a prior grant or terminal outcome returns history without a
+ credential lease or provider read. A reserved operation can progress only
+ through fresh authorization and the fenced single-winner grant. Changed
+ bindings return `gateway.effect.operation-conflict`. An unavailable journal
+ reports `gateway.attempt.state-unavailable`, never proof of non-entry.
+ PostgreSQL writers share the table-scoped advisory transaction fence;
+ file writers share the bounded writer lock. The lifecycle store's singleton
+ contract-row lock is not the gateway's fence.
- **Spend limits.** A grant's bounded policy (evaluator
`auths.gateway.argument-ceiling-window-count/2`) counts per link subject in
fixed, epoch-aligned windows, in the attempt store: every gateway process
@@ -140,7 +245,7 @@ held which key, is operator evidence and is not produced by this code.
- **Sockets.** `serve` keeps separate capacity for the app socket
(`--app-capacity`, 1–1 024, default 64) and the owner-only `admin.sock`
(4), so `auths-gateway disable`, `enable`, `revoke`, `rotate`, `status`,
- and `reobserve` still connect while the application holds or refills the
+ and `recovery-record` still connect while the application holds or refills the
app socket. A connection past either socket's capacity is closed at accept
without a response. An admin change commits without waiting for any
submission or provider call, then waits at most 20 seconds for this
@@ -157,7 +262,7 @@ held which key, is operator evidence and is not produced by this code.
must already exist, equal its canonical path (on macOS `/tmp` resolves to
`/private/tmp`), be owned by the gateway's UID, and grant no group or
other permission (mode 0700 or narrower); `rotate` sends the credential
- over this socket. Give `disable`, `enable`, `status`, `reobserve`,
+ over this socket. Give `disable`, `enable`, `status`, `recover`, `recovery-record`,
`revoke`, `rotate`, `operator-attest`, and `doctor` the same
`--admin-socket`; without it they look for `/admin.sock` and
report `gateway.admin.socket-unavailable` naming that path. The admin
@@ -194,7 +299,7 @@ custody is suspect.
rotation. Do not disable the old key yet.
2. Record the new key's anchor facts: principal, `raw-key-v1`, verification
method, `p256-sha256-v1`, both schemas (`auths.gateway-readback/1`,
- `auths.gateway-outcome/2`), and both subject namespaces. For a
+ `auths.gateway-outcome/3`), and both subject namespaces. For a
development seed, `auths-gateway observer-show` prints them together with
`observer_custody`.
3. Update the trust as in §4, replacing the principal of the existing
@@ -267,3 +372,76 @@ verifier reads.
Git trust pins a single root anchor today; the M-of-N composition in §1
applies to gateway trust.
+
+
+## Native content controls under AP074
+
+A content-capable installation accepts `content-acquire --app-socket ... --proof
+... --action ... --approval ...`. Its result is redacted acquisition history;
+it does not release provider content. `effect-review --content` reviews the
+native signed read action. Release is separately authorized and reviewed with
+`effect-review --disclosure --binding-out `.
+The proposal is the closed native `AppDisclosureTarget`: source operation,
+recipient public key, exact UID/channel policy and delivery instance. Use native
+call authoring as follows; the authenticated UI journey remains tracked by AP074.
+
+1. Save only the declared resource arguments in an owner-private JSON object.
+ Run `effect-content-call --state-dir ... --resource-arguments /private/resource.json
+ --operation ... --call-out /private/read-call.json`. The output path must be
+ new and absolute. The native parser rejects duplicates, extra fields and
+ input above 65,536 bytes. Obtain native proof/action for the returned exact
+ call, then review with `effect-review --content` and obtain exact approval.
+2. Acquire through `content-acquire`. A recorded provider response is not a
+ delivery and does not authorize release.
+3. Save the private native recipient proposal. Run `effect-disclosure-call
+ --state-dir ... --target /private/target.json --operation ...
+ --call-out /private/disclosure-call.json`. This loads retained provenance,
+ not provider content from the network. Obtain independent proof/action and
+ review with `effect-review --disclosure /private/target.json --binding-out
+ /private/expected-binding.json`, then obtain exact approval.
+
+Both authoring commands preserve native canonical calls; they create no proof,
+approval, reservation, credential lease or transport permission. Resource data
+is not a caller-controlled URL/body/limit override.
+
+`disclosure-receive` requires the trusted expected-binding file, exact byte count,
+proposal and independent proof/action/approval. Pipe the recipient's raw 32-byte
+Ed25519 seed through non-interactive stdin with `--recipient-seed-stdin`; never
+put it in command arguments. Use new absolute `--content-out` and
+`--acknowledgement-out` paths. The native client checks the pinned gateway UID
+before sending private authority. It writes private content and native ACK files
+before transmitting the ACK. Its `recipient-ack-sent` result describes the client
+send, not durable gateway acknowledgement.
+
+Use `content-history --app-socket ... --operation ...` to inspect native retained
+facts. If a received ACK was lost before persistence, `disclosure-acknowledge
+--app-socket ... --operation ... --acknowledgement ` records
+that exact evidence without redelivery. A disconnect, unknown outcome or retained
+grant must not trigger an automatic repeat. Python/TypeScript expose matching
+acquisition/history/ACK control methods returning native result bytes. Recipient
+handshake, binding/ACK interpretation and signing remain in Rust.
+
+
+The `content-process-journey.json` artifact runs the real local daemon, authoring,
+review and recipient CLI with a counting HTTP double. It exercises read/release,
+ACK-loss crash/reopen/no-redelivery/late ACK, an exact write and independent
+observation, plus separate-process disable before release. Public fixed keys
+simulate signing. This establishes neither human presence, installed language
+binding integration, hostile UID/egress isolation nor PostgreSQL durability.
+
+
+A standalone content installation can omit `write` from recipe source `/3`.
+Its source and profile lock must name `auths_gateway_content_read_v1`, and the
+source must declare `content`. Include only binding/resource/account fields;
+mutation observation, echo, pre-entry, preconditions, relative ceilings and spend
+bounds are refused. The native review explicitly shows no mutation method or
+recovery capability, an empty write path and zero body allowance, alongside the
+content GET declaration. Do not interpret its null method as an unspecified
+method or substitute a default write. `RecipeReview::method` and mutation
+`recovery` return `Option` in Rust; JSON consumers receive native null values.
+
+The generated `bindings/fixtures/gateway/content-only/` files are a synthetic
+local recipe example. `read-only-process-journey.json` proves real installation,
+read-only authority, mutation refusal, one GET and separate recipient disclosure.
+It is not a Google integration or a replacement for AP070's pending metadata,
+credential-custody and Python state-owner cutover.
diff --git a/docs/plans/0074-assurance-map.md b/docs/plans/0074-assurance-map.md
new file mode 100644
index 000000000..0b894a1a1
--- /dev/null
+++ b/docs/plans/0074-assurance-map.md
@@ -0,0 +1,70 @@
+# AP074 assurance boundary and evidence map
+
+This maps AP074 §9's six conditional properties to shipping source and required
+runtime evidence. It is a producer/assumption map, not an execution receipt or a
+whole-system proof. Every final claim still needs the corresponding exact-source
+hosted artifacts. The acceptance ledger records that execution separately.
+
+## Pure proof surface and runtime obligations
+
+| Property | Shipping predicates and native owner | Evidence outside the pure proof |
+| --- | --- | --- |
+| Authority preservation | `effect::ApprovedEffect::validate_for_entry` recompiles the original native proof/action and compares the complete binding. `entry::EntryValidation::validate` reads connection/configuration dependencies and samples current time under the adapter fence before invoking it. `auths_gateway_kernel::order::next_step` orders the abstract validation/reservation/grant/send events. | Native signatures, trust installation, exact approval and clock authenticity are runtime obligations. Native dependency, mutation-entry and disclosure-entry corpora exercise independent authority/approval/observation expiry, changed threshold selection and withdrawal on both stores. Boolean ordering premises alone do not prove cryptographic authority. |
+| Effect preservation | `auths_gateway_kernel::construct::{construct_write,construct_action_read,construct_credential_read}` implement the pure closed request leaves. `effect::BoundEffect` retains sealed native authority, the immutable binding and compiler result. `transition::valid_transition` requires unchanged fixed bytes and protocol/plan. | Reviewed recipe meaning, canonical codec correctness and credential insertion are explicit boundaries. The native effect/review corpus, two-recipe independent differential and installed owner journey compare the actual request against the exact approved review. TLS/provider behavior remains external. |
+| Single-grant safety | `transition::{stage_transition_allowed,entry_kept,valid_transition}` permit an entry addition only at the reservation-to-grant edge and retain it afterward. `order::next_step` reaches send only after the new grant acknowledgement. Native mutation/acquisition/disclosure grant methods return consumed permits only after successful fenced persistence. | File/PostgreSQL atomicity and truthful durability acknowledgement are assumptions checked by independent-process races, process crashes, file failures, partial SQL batches and real lost COMMIT replies. Generated concurrent native histories add an independent set-of-entered-operations oracle. No persisted byte string recreates a permit. |
+| Capacity conservation | Native `store::{reserve_with_claim,frontier,cumulative}` and the typed acquisition/disclosure allocation paths own reservations in the same admission batch. Final entry validates retained counters/allocation rather than charging again. Kernel transition fixed-byte equality preserves admitted charge facts. | Arithmetic/policy predicates do not prove database atomicity or actual occupancy. Native lifetime/fixed-window count and sum races, restart/retention and fault corpora check actual permanent/window rows on both adapters. Unknown stays charged; elapsed time, absent evidence or administrative annotation supplies no refund fact. |
+| Revocation ordering | `entry::EntryValidation::validate`, `connection` validation and `configuration::SharedConfiguration::entry_snapshot` use snapshots held under the same fence as the final attempt replacement. Connection/configuration writers participate in that fence. | File `flock` is single-host; PostgreSQL uses the maintained transaction/advisory-lock contract across hosts. Native worker/operator schedules order withdrawal before versus after grants for mutation, acquisition and disclosure, including waits and physically distinct dependency records. Earlier committed grants retain their documented semantics. |
+| Evidence soundness | `transition::{protocol_consistent,response_kept,acknowledgement_kept,valid_transition}` keep distinct protocol facts and immutable evidence. Native recipient authentication/ACK validation and recovery evidence-link validation establish the facts before projection. `observer` separates signed public history from full private audit; the auditor rechecks complete facts. | Signatures, recipient/channel/source binding, bounded HTTP evidence and provider linkage are not proved by nonempty model byte fields. Real recipient ACK substitution/loss, independent recovery, privacy and installed-interface tests supply that evidence. State equality, a response, an ACK and declared provider linkage remain distinct claims. |
+
+The native adapters call the shared `store::valid_transition`, which invokes the
+translated `auths_gateway_kernel::transition::valid_transition` over both typed
+record projections **and** validates the complete new protocol record. The model
+does not replace canonical record validation or reinterpret provider fields.
+
+## Formal source link
+
+The authoritative extraction inventory is
+[`qualification.toml`](../../formal/qualification/aeneas/qualification.toml).
+It names the actual shipping gateway-kernel files and selected functions,
+including construction, submission order, transition/presence, recovery and
+ratio leaves. The qualification reproduces translation twice under pinned
+Charon/Aeneas toolchains. Generated Lean files and
+[`Gateway.lean`](../../formal/qualification/aeneas/cases/Gateway.lean) feed the
+reviewed refinement claims in
+[`assurance-manifest-v1.toml`](../../formal/assurance-manifest-v1.toml).
+
+The compiled assurance audit checks statement identities and transitive axioms;
+Kani's record-table harness checks bounded representations, including every
+protocol/stage pair and illegal edge under its byte bounds. Sequential generated
+file-store properties and actual durable process histories complement this pure
+surface. The existing formal mutation inventory includes gateway translations in
+campaign s3. Its source freshness and adequacy must pass the existing release
+gate; its presence in the repository is not a fresh execution result.
+
+Codecs, cryptography, full verifier control flow, store implementation, clocks,
+networking, recipient behavior and provider behavior remain outside the pure
+gateway proof. Full API/record validation, current native authority and concrete
+adapter/transport evidence are mandatory premises, never optional exclusions
+that waive an AP074 security invariant.
+
+## Operational limits of the claims
+
+File durability assumes acknowledged sync/rename and private storage on one
+trusted host. PostgreSQL durability assumes the maintained TLS/isolation/commit
+configuration and participating writers. Both require current epoch/configuration
+and declared retention/rollback floors. A malicious operator, malicious host,
+unsupported whole-store rollback and independent external provider writers are
+not prevented by Rust types or these tests.
+
+The installed Linux inventory checks the actual unprivileged app account against
+managed sockets, credentials, private files and every application/CLI route.
+AP070 remains a separately application-owned reference with its own narrower
+claims. It cannot receive managed grants, permits or credentials or supply a
+second decision for the same managed operation. Its receipts are not native
+entry, recovery or qualified execution evidence.
+
+The [cost evidence](0074-native-cost-evidence.md) and
+[generated concurrent histories](0074-generated-histories.md) describe additional
+producers and their limitations. Neither a producer definition nor a successful
+artifact constructor closes A01–A22; final exact-revision verification remains
+required.
diff --git a/docs/plans/0074-ci-repair-46989b7d.md b/docs/plans/0074-ci-repair-46989b7d.md
new file mode 100644
index 000000000..9f95e6e74
--- /dev/null
+++ b/docs/plans/0074-ci-repair-46989b7d.md
@@ -0,0 +1,25 @@
+# AP074 hosted repair: 46989b7d
+
+Published source `46989b7db512f382b184125e29047984f5d9607d`; PR merge producer `f4d421821704c809dfede654f7fbc8eb66e8a95d`, both verified tree `3bcc7b399b436ab67027732001e9eae7d0fd15d5`. Merge parents: `e2d10a74fa0be5e6bfe2b9f698632c835cd7b11b` and46989. Draft PR213 unchanged.
+
+Repair commit `bbb391a9afc11e04654d01ce8567c4b06897da59`, isolated clone `/tmp/auths-ap074-ci-resumed-20261105`. Five source files, no shared closure/freeze/ledger changes, no main/AP070/AP073 edits. Parent must fetch independent clone before integration. No repair publication without exact integrated-head authorization.
+
+## All current source roots
+
+- [Python source contracts126820242109](https://github.com/auths-dev/auths-proof/actions/runs/41900200180/job/126820242109):438passed/11skipped/1failed. The audit argument-forwarding stand-in fixture created public bundle.json, while the correct operator default is private-bundle.json. Repair creates the private stand-in fixture, poisons the public inventory, explicitly asserts private path forwarding, and retains exact refusal/pin/option/path-escape assertions. No fake observation used as managed execution evidence. Focused test passes1.44s; installed _enforcement_init/_profile_cli/operate/common bytes verified equal current sources (local Python3.13).
+- [Repository preflight126822586483](https://github.com/auths-dev/auths-proof/actions/runs/41900199852/job/126822586483):rustfmt requires sorted cost_tests module registration. Running the subsequent focused gateway Clippy portion of the same failed preflight found two latent diagnostics: timestamp digit separators in native_result_wire_cases.rs and explicit unit pattern in native_cost_store.rs. Both mechanical repairs preserve generated bytes/semantics. `cargo fmt --all --check` passes; `cargo clippy --locked -p auths-gateway --all-targets --all-features -- -D warnings` passes11.24s. No lint allowances.
+- [PostgreSQL126820236050](https://github.com/auths-dev/auths-proof/actions/runs/41900199751/job/126820236050):all48ordinary selectors and both new cost selectors actually passed. Artifact upload failed because relative AUTHS_AP074_COST_DIRECTORY resolves under the Cargo crate cwd, while upload expects workspace target. Workflow now uses `${{ github.workspace }}/target/ap074-native-cost`. No completed native suite repeated locally. Cost metric JSON remains unavailable until repaired hosted upload; test durations alone are not path-latency measurements. Independent information comparison and restore were skipped afterward.
+- All eight main failed rollups individually inspected: implementation RUN_RESULT=skipped after preflight failure, not eight extra source defects. Formal-fast remains active at this report checkpoint. No healthy workflow restarted/cancelled and no platform coverage removed.
+
+Exact staged Gitleaks8.28.0:704bytes/0findings; diff check clean. Per-command unsigned local commit under parent's established convention; global signing config untouched.
+
+## Current positive evidence and limits
+
+- Native log audited against workflow:48ordinary selectors and2cost selectors each exactly once,1passed/0failed/0ignored. File cost3.88s, TLS PostgreSQL cost4.88s. These are test durations, not reconstructed metric values.
+- Uploaded PG artifact13357901657 ZIP SHA256 `e2159042bed2e902dd9ab98bff489d4043cce31db80ae0534fb99c09cdeb6d6c` verified. Despite artifact name, it contains only native log, no information comparison report. Do not count22pairs or restore as passed for46989.
+- Native mutation generator and independent28request/26response comparison now pass hosted. Artifact13358171429 ZIP `d7165e53a9601f741f1be558553d109e8d5603c7166deeb33210bb1c9d43fcef` verified; seven source hashes match current sources. Python pytest reached438passes after this gate, with only unrelated audit stand-in failing.
+- Previously failing real Python north-star126820238519 and scaffold126820238005 now pass on hosted source46989; npm north-star126820238099 also passed. All current installed demo jobs pass.
+- Airtable artifact13359390742 ZIP `a1521b8cf5808858f0cc787252342fe1d175a4e94abbfd4cc5076b879c03ce46` verified:exactsource46989,36simulation cases/privateaudit13verified0inconsistent; all positive owner-browser flags true, productionfalse. Installed Python/TS acquisition, separate disclosure/ACK, private0600content, replay/no extra execution, loss/restart controls retained.
+- Stripe artifact13359295606 ZIP `6c676f969afa8c36d2f56f62ad13813c81f2c07ba426c550dca2e46006bcbbfa` verified:exactsource46989,23simulation cases/privateaudit6verified0inconsistent, productionfalse.
+
+Final source qualification remains open. Required implementations are blocked by formatting prerequisite; cost artifact upload/information comparison/restore must run on repaired head. macOS wheel producer/git protocol were still queued at checkpoint, so their dependent matrix cannot be inferred. Preserve formal-fast and finish terminal cohort before next authorized publication. Parent's newer generated-history work is not part of this source or evidence.
diff --git a/docs/plans/0074-ci-repair-7807d04b.md b/docs/plans/0074-ci-repair-7807d04b.md
new file mode 100644
index 000000000..174b2dd88
--- /dev/null
+++ b/docs/plans/0074-ci-repair-7807d04b.md
@@ -0,0 +1,47 @@
+# AP074 CI repair: 7807d04b
+
+Published source `7807d04bfad2a2d2dba7b8f842e27fd742a9a8ad`, draft PR213. Producer merge `81a5e41c81807ffbf08de3276fdded25e3e62129` has the same verified tree `9e97fc5a871cae92cecfff97eb8cb9d8d51ba142`. Merge parents: `e2d10a74fa0be5e6bfe2b9f698632c835cd7b11b` and source7807.
+
+This is a repair report, not final qualification. Snapshot:48success/15failure/17skipped/1cancelled; ten macOS jobs queued and formal-fast126807398330 running. Preserve healthy jobs. Authoritative implementation/formal/dependency/compliance/security jobs skipped after preflight failure, so their required rollups correctly failed. No retry/cancellation/coverage removal/live provider action/automatic human consent.
+
+## Root matrix
+
+| Actual failed jobs | Cause and complete repair |
+| --- | --- |
+| Recipe41895808306:126804868144 simulation,126807854414 Stripe candidate | `qualification.harness.reviewed-plan-drift`: ADR0014 changed outcome-v2→v3. Existing generator refreshes only Stripe decision-record digest and transitive contract digest. All exact source checks retained; downstream simulation unexecuted. |
+| Python41895808782:126804869433 source contracts | Native mutation differential stale: observed trust changes commitment→effect-reference, changing ten effect IDs. Actual hosted assertion decoded; regenerated native artifact differs only in ten effect_id values. All decisions/request bytes/response cases unchanged. Independent28request/26response report regenerated. |
+| SDK41895808373:126804867858 Python and126804867445 npm north-star | Operator audit export ran after deliberate rollback erased journal records. Both now export private signed historical snapshot before rollback and audit it afterward. Separate post-rollback export requires only truly retained/resubmitted refund-1. No missing evidence synthesized; every original audit/provider/tamper assertion preserved. |
+| SDK41895808373:126804867717 scaffold | New audit-export omitted configured run/admin.sock. Selftest and operator README now supply explicit --admin-socket; native permission gate unchanged. |
+| CI41895808417:126807398306 preflight | Strict Clippy method closure in private audit signing. Focused reproduction then found eight latent test diagnostics: cloned reference slice, two overlong functions, nested if, three use-after-statement diagnostics, owned String parameter. Helper extraction/borrowing/import placement preserves all assertions, without new allowances. |
+
+Job URLs use `https://github.com/auths-dev/auths-proof/actions/runs/RUN/job/JOB`. All eight failed downstream main rollups inspected: implementation result skipped, not additional source failures.
+
+## Focused local reproduction
+
+Independent isolated clone `/tmp/auths-ap074-ci-resumed-20261105`; main/AP070/AP073 untouched.
+
+- Authoritative `python3 qualification/reference/generate_families.py` refreshes two Stripe generated files.
+- `/tmp/ap074-ci-7807-venv/bin/python -m unittest discover -s qualification/reference/tests`:104tests,2conditional skips,success0.608s (local3.13; hosted3.12). System3.10 lacks tomllib; uv3.12 has unrelated venv-loader environment failure. No repository workaround for either.
+- Native exact generator with `AUTHS_UPDATE_FIXTURES=1 cargo test --locked -p auths-gateway --lib --features loopback-provider,testkit-harness observed_tests::effect_native_tests::differential_tests::native_independent_mutation_differential_fixture_is_current -- --exact`:1passed/0ignored0.10s. Unchanged check without update:1passed/0ignored0.11s.
+- `python3.13 qualification/reference/ap074_differential.py --update`, then without update:2families/28requests/26responses.
+- `cargo clippy --locked -p auths-gateway --all-targets --all-features -- -D warnings`:success14.53s after collecting latent diagnostics; no broad workspace rerun.
+- Isolated wheel built with locked maturin1.9.6/python-extension and installed into private venv. Native binary copied read-only from main's7807-built executable, parent-confirmed provenance. Local reproduction, not hosted artifact qualification.
+- Full Python and TS north-star pass: each2verified/7refused/3unverified/0inconsistent, original provider/idempotency/tamper controls, restored audit operations exactly[refund-1]. TS locally packed source uses verified7807 hosted WASM artifact13356521420 ZIP SHA256 `3f3546c5387310894f39a7f787a6160dbd9e984de256b5dd812b2e6f3e7b29ab`.
+- Full generated scaffold from installed wheel passes setup/install/start/selftest/stop:2verified/3refused/3unverified/0inconsistent,8tamper detections,2provider writes, credential scan empty, original one-JSON assertions. Transient processes stopped.
+- cargo fmt and staged diff checks pass. Parent explicitly requested exact staged Gitleaks8.28.0 before child commit; result appended below.
+
+Logs: `/tmp/ap074-ci-7807-{reference,clippy,differential,differential-check,journey,tsjourney,scaffold}.log`; two journey JSON summaries; `/tmp/ap074c7sc/audit/selftest.json`; `/tmp/ap074-ci-7807-mutation-report.json`.
+
+## Hosted positive evidence at source7807
+
+- PG41895808323/job126804867562 success: all48 exact selectors audited from downloaded log, each once/1passed/0failed/0ignored. Actual22information pairs, matching native-log digest. Artifact13356796759 ZIP SHA256 `5a8bc47f697b8048b321c018c427e91bb20227203f3baf5214bc5eba995f5fb2` verified. Producer is merge81a5 above, not relabeled as source execution.
+- Linux isolation41895808583/job126804868451: old2tests pass; new distinctUID1passed/0ignored3.49s. Artifact13356001317 ZIP `d27f4016e3b36a163a47076a4a59fec995e8dc20366cbb976e5e4edf45ea1bee` verified. Exactsource7807;33app/12admin/53CLI/5privateCLI.45then49privatefiles inaccessible; canary exclusion/native recipientACK/no extra custody-provider/source retention true. Operator audit file0600,oneverifiedentry,overwrite refusal. No hostileOS/root/provider qualification claim.
+- Airtable41895808300/job126810262944:success. Artifact13355937306 ZIP `de504857e77a157a2834349508b7c138b3d5fb9547679129c3ebfc37a8967899` verified.36simulationcases/privateaudit13verified0inconsistent. Owner browser all positive flags true, exactsource7807:installedPython+TS acquisition, separate disclosure/nativeACK,0600recipientfile,no replaydelivery,loss/restart no reentry. Browser-simulated consent only;productionfalse.
+- Stripe41895808302/job126810393629:success. Artifact13356986494 ZIP `1b02b10327f2eeadcdba7b89acb165783b6b629817f06c5c13a976c129540858` verified.23simulationcases/privateaudit6verified0inconsistent;productionfalse.
+- Operator41895808719,gitprotocol41895808393,branch41895808466,independentGoogle41895808299 passed. Custody-live41895808322 conditional skip supplies no evidence.
+
+## Remaining
+
+Exact staged Gitleaks8.28.0 passed:12,545bytes,0findings. Official release binary verified against published checksum; `/tmp/ap074-ci-gitleaks-8.28.0/gitleaks`. Per-command unsigned child commit authorized by parent; global signing configuration unchanged and no signing key accessed.
+
+Preserve running formal-fast126807398330. Authoritative implementation, translated/proved qualification and required rollups remain unexecuted/failed prerequisites. Ten macOS SDK/harness jobs queued. Repaired source needs next integrated hosted cohort. Parent owns combined gateway closure/release identities/compliance/ledger. No repair publication until exact parent authorization.
diff --git a/docs/plans/0074-ci-repair-ab980d10.md b/docs/plans/0074-ci-repair-ab980d10.md
new file mode 100644
index 000000000..bf6c913dc
--- /dev/null
+++ b/docs/plans/0074-ci-repair-ab980d10.md
@@ -0,0 +1,57 @@
+# AP074 holistic CI repair — ab980d10
+
+Published source: `ab980d1028927ef7bb08159ac489153d767993d5`, PR #213.
+Child repair: `31f71657190af6f5ad8348cdaa6db4dbd2742328`, one coherent seven-file commit, not pushed by child. Main integrates separately. Main/AP073/AP070 source and shared metadata/index were not modified by this investigation.
+
+## Root matrix and repairs
+
+| Root | Actual evidence | Repair |
+| --- | --- | --- |
+| Fresh owner container lacks counting ledger before its first provider request | [Airtable installed job123776995604](https://github.com/auths-dev/auths-proof/actions/runs/40967090762/job/123776995604): owner_browser.py calls provider_counts(), strict docker cat fails because provider.jsonl does not exist | Create the owned fixture ledger exclusively at startup with mode0600; refuse an existing path, never truncate retained counts. Counter readers remain strict. Add a regression proving initial empty evidence, mode, append and restart refusal; invoke it in the package workflow. |
+| Managed recipe declares UTF-8/text/plain but actual Airtable-shaped response is application/json | Full installed owner reproduction reaches Python acquisition and native history rejects it with gateway.content.response-media-type | Use the existing native json-string projection at /fields/PrivateNote; regenerate the native review and document the narrower projection. Keep native media-type rejection and all bounds. Recipient assertion requires exact sentinel bytes rather than interpreting the entire record as content. |
+| Second acquisition has the same rendered outcome label as the first | Reproduction reaches both native acquisitions but reads the first rendered Python result before the TypeScript response completes | Wait for the submit control to re-enable; require exact current operation_id and language. Keep outcome, counters, withheld-content and full native literal-review assertions. |
+| Disclosure review assertion expects a nonexistent native.protocol field | Reproduction reaches separately approved disclosure review and raises KeyError | Require actual binding.protocol.kind=disclosure, native versioned disclosure schema, exact source operation and native/semantic disclosure equality. Native renderer output remains byte-for-byte unchanged. |
+
+## Exact installed reproduction
+
+Original package artifact `12939290801`, run `40967090762`, has ZIP SHA-256
+`a2032da2971a0edcdcf118271b8a5eb6c9bb73cce45e068940583bbc8a6ad8e4`.
+The ZIP, nested tar checksum, source_commit=ab980d10 and all41 manifest file hashes were verified before overlaying repairs. Package directory:
+`/tmp/ap074-ci-ab980-airtable-package/airtable-gateway-demo`.
+
+The Docker reproduction uses that package's original Linux x86_64 shipping and simulation gateways, wheel, npm/WASM artifacts and compiled TypeScript client. Only repaired demo payloads and browser assertions are overlaid. Host browser is Playwright1.62.0 with its installed Chromium; test driver changes only owned container name/output paths beyond the committed assertion repairs. This is local patched-payload evidence, not hosted success for the new commit.
+
+Actual native review generation uses the original hosted shipping binary:
+
+```text
+auths-gateway review --recipe managed-recipe.json --profile-lock profile.lock.json
+```
+
+New managed recipe digest: `88f9e64b3f82ff6d369a31082d9e2b00890b0d676bb1c0c8e3610e961b1769eb`.
+
+Results:
+
+- `python3 demos/airtable-gateway/tests/test_provider_ledger.py`:1passed.
+- Source-free Docker `--selftest`:36cases passed, production_qualification=false. Log `/tmp/ap074-ci-ab980-selftest.log`.
+- Standard installed desktop/mobile browser:passed, log `/tmp/ap074-ci-ab980-browser.log`.
+- Full installed owner browser:passed, log `/tmp/ap074-ci-ab980-owner-browser-disclosure.log` and artifact `/tmp/ap074-ci-ab980-owner-evidence/owner-journey.json`.
+- The owner artifact asserts native literal renderer equality, withheld approval before consent, decline/substitution refusals, private operator files, separate owner origin, installed Python and TypeScript acquisition, separate disclosure consent, retained native ACK, exact0600 recipient bytes, no public HTTP content, no replay redelivery/refetch, same agent actor, exact write, independent recovery, unknown response loss and fresh-process replay without new entry.
+- Both owned browser containers were stopped and removed. The disposable selftest container was automatically removed. Parent and unrelated PostgreSQL containers were untouched.
+
+No native outcome, security deadline, media-type check, counter assertion, signature or consent requirement was weakened. No live-provider or human-attention qualification claim follows. No redundant broad local suites, secret scan, workflow reruns or cancellations were performed.
+
+## Current hosted evidence
+
+PG/source-contract jobs checked out PR merge `de85f0ffc03f6bc29244a2c69c8d74c071da453b`. Merge and published source have identical tree `3a4b23482a6a5fd7efc87692233405409fa44002`, fetched and compared.
+
+[PG123774214713](https://github.com/auths-dev/auths-proof/actions/runs/40967090703/job/123774214713) passed all26 explicit native selectors exactly once, each1passed/0failed/0ignored, including both new epoch cases. Then installed-SDK information comparison22pairs and physical/WAL restore passed; restore retains replay_continuity_claim=false. [Python source123774214528](https://github.com/auths-dev/auths-proof/actions/runs/40967090693/job/123774214528) passed native mutation generator plus28request/26response comparisons.
+
+Downloaded artifact ZIPs and checksums verified:
+
+- Information/native-log12938568979: `c1be7c1114d6bbe8ec39214d05f849aea981bfa97e0a97e45c6cd32197fc9f55`.
+- Mutation12937534434: `0aeef604c259ff01e79c131ddbe7ceddd8bad7da4443e7575d8f624902c71781`.
+
+Per-selector evidence: `/tmp/ap074-ci-ab980-native-executions.json`.
+Reports: `/tmp/ap074-ci-ab980-information-report.json`, `/tmp/ap074-ci-ab980-mutation-report.json`.
+
+At repair handoff the only observed hosted failure is the Airtable owner journey. Main authoritative implementation, Kani and authoritative Lean are still running; final qualification is not yet established. Continued holistic observation will produce a terminal cohort report. Parent owns metadata regeneration and coordinated publication.
diff --git a/docs/plans/0074-differential-contracts.md b/docs/plans/0074-differential-contracts.md
new file mode 100644
index 000000000..f32435fb8
--- /dev/null
+++ b/docs/plans/0074-differential-contracts.md
@@ -0,0 +1,116 @@
+# AP074 independent differential contracts
+
+A19 requires both mutation and acquisition/disclosure comparisons. This file
+records the exact overlap and differences; neither independent reference becomes
+a managed execution path. The acceptance row remains partial until all required
+comparisons and exact-head hosted execution exist.
+
+## Mutation corpus
+
+`native_differential_tests.rs` compiles the maintained Airtable record-update and
+Stripe platform-refund recipes. The Airtable fixed base/table substitution is
+the same reviewed, pure expansion permitted by the independent reference. No
+network request is sent to those resources. Each request case has an actually
+signed native grant/action/proof, ordinary verification, and either a refusal or
+an immutable bound effect. Successful cases also obtain and revalidate a real
+exact-effect approval signed by public fixed test keys. There is no admission,
+entry permit, credential, provider execution or human-presence claim.
+
+The retained native corpus has 28 request cases and 26 response cases across
+the two families. `qualification/reference/ap074_differential.py` calls the
+unchanged `airtable_record.py`, `stripe_platform.py`, `fresh_evidence.py` and
+`provider_readback.py`; the references never import the gateway implementation
+or this comparator. Every case name/count is required, including duplicate and
+omission refusal. A negative comparator test mutates request bytes, native
+decisions, case inventory, observations, recovery targets and resource scope.
+
+| Contract | Comparison and limits |
+| --- | --- |
+| Request equality | Exact method, URL, body bytes, content type, ordered noncredential headers, echo and idempotency key agree for all shared positive inputs. |
+| Shared refusal | Missing/extra fields, wrong namespace/recipe, operation limits, invalid values and signed amount-ceiling violations refuse on both sides. Refusal codes are retained separately: reference input validation and native authorization occur at different stages. |
+| Resource scope | The independent qualification pool refuses an otherwise well-shaped unowned resource. A native grant/recipe can authorize that resource without knowing the qualification pool. These expected differences are required assertions, not accepted unexplained drift. |
+| Provider guard | Native compilation is not the separately fenced fresh provider-read decision. Independent Stripe guards separately demonstrate the exact 50 percent ceiling and currency binding; this corpus does not claim to execute the native live guard. |
+| Observation | Exact expected value plus exact declared echo links the generic native observation. Wrong value/echo and unavailable JSON retain distinct facts. Duplicate members, malformed/nonfinite JSON and over-limit responses never become a definitive observation. |
+| Byte limits | Complete valid JSON plus trailing whitespace remains accepted at the native 16,384-byte observation limit and becomes unavailable at 16,385. The independent full-domain witness has a different 65,536-byte limit, so it accepts that larger valid native-refused response; both refuse 65,537 bytes. These bound differences are explicit. Large padding is represented by its bounded count in the corpus and materialized before actual native/reference calls. |
+| Full domain witness | Independent references additionally check the record identity or refund payment, status, currency and live-mode facts. A generic recipe that projects one value and echo does not acquire those extra domain predicates. Missing echo can be native state equality while the full independent witness refuses. |
+| Refund identity | The independent Stripe reference discovers a refund by its action-bound echo. A different safe refund ID alone does not invalidate that witness. It never trusts an application-supplied refund locator. |
+| Recovery capability | Airtable has a verified record locator before any write response. Stripe's native response-derived locator is unavailable without the recorded response; a declared idempotency key cannot recreate it or confer a new entry permit. The independent reference can instead find exactly one action-bound refund in its own bounded listing. That discovery is an intentionally different test-only capability, not a native recovery grant. |
+| Path closure | Native response locator extraction refuses a bare dot segment. A slash inside a value stays percent-encoded within one closed segment. The reference's refund identifier grammar is narrower. |
+
+The source-contract Python workflow first executes the native generator with
+fixture updates disabled, then executes the independent comparison and retains
+both artifacts. The report hashes the exact native corpus and relevant source
+files. A checked-in report alone is not fresh execution evidence. Required
+hosted results must identify the actual tested source/PR-merge revision and run.
+
+## Acquisition/disclosure comparison (hosted execution on 74ce9f75)
+
+AP070 remains an independently implemented application-owned Python/SQLite
+reference. `ap074_information_differential.py` executes its unchanged admission,
+release and runtime-reopen code on a precisely named post-authorization contract.
+Its synthetic `VerifiedRead` inputs use the existing reference unit-test boundary;
+they do not prove authorization equivalence. Challenge body digests still call
+the actual native SDK built from the same checkout; hashing is not stubbed or
+reimplemented. No reference source, provider credential or managed permit is
+changed or supplied.
+
+The authored corpus has three separate count/byte/both capacity races, with a
+prefix admission and two independent SQLite worker processes contending for the
+last unit. Closing in-flight work retains all conservative call, content,
+provider-byte and two-GET attempt reservations. Four release gates cover current
+state, task revocation, connection revocation and task expiry. Reopening refuses
+the old claim and repeated release without additional charges. A further
+positive release followed by connection revocation is paired with native
+historical replay; retained facts cannot send again. A real SIGKILL after the
+reference persists its first provider-attempt start proves that runtime reopen
+marks the claim abandoned, frees only in-flight capacity and preserves charges.
+No Google I/O is performed.
+
+The PostgreSQL lifecycle job retains the output of its existing native selectors
+and then runs this comparator using the same checkout's installed native Python
+SDK. It requires exactly one passing, nonignored execution for each of eight
+file/PostgreSQL capacity, information replay, acquisition crash and disclosure
+crash selectors before reading their natively regenerated-and-checked artifacts.
+Missing, duplicated, ignored and zero-pass results refuse comparison; parser
+controls use explicitly synthetic logs and make no runtime claim themselves.
+The report records tested commit/run, native log/artifact hashes, unchanged
+reference source hashes and 22 paired comparisons. This avoids rerunning the
+native matrices merely to perform the independent comparison. On published `74ce9f753faa6bf3f1cb065bc369f655e6c0b236`,
+[PG job123745460518](https://github.com/auths-dev/auths-proof/actions/runs/40958689260/job/123745460518)
+passed the required native selectors and all 22 information comparisons. Its
+report identifies actual PR merge `df0cc98e6eae440db67038c4f552b74e1d64b90f`,
+whose tree matches the published source. Artifact12934958430 was downloaded and
+its ZIP SHA256 verified as
+`c7af0195a0a6645e2cc6fa58f6b0ada451b46692c98649edfac35d4b551b08c0`.
+All five recorded reference/comparator source hashes still match this checkout.
+Source-only reference job123745461358 passed 104 tests including the parser
+refusal controls. Python source job123745459370 separately passed native mutation
+construction and the 28/26 independent comparison; artifact12933979778 has ZIP
+SHA256 `3437045757e96a2f1020cb729935e280a2ad8758c125646608f9f766ec1143e0`.
+Its seven recorded generator/reference/comparator source hashes also match this
+checkout. These observations establish execution on that head; subsequent source
+changes and the final completed PR head require their own hosted confirmation.
+
+Comparisons must preserve these differences:
+
+- Native acquisition response and separate recipient ACK are distinct facts.
+ AP070's transactional release commitment has no authenticated recipient ACK
+ protocol and cannot be relabeled as one.
+- Native replay may return retained historical facts. AP070 can refuse a
+ repeated challenge. The shared property is no new external entry or charge;
+ equality of public result codes is not required across those contracts.
+- AP070 task-lifetime content ceilings and metadata-plus-content provider byte
+ reservations differ from native fixed-window response ceilings and separate
+ disclosure byte allocations. Preserve raw values and named units; do not
+ force unlike quantities into equal totals.
+- A native durable multi-process/adapter guarantee cannot be inferred from
+ AP070's application runtime lock or SQLite demonstration.
+
+Native current-authority gates and AP070 task expiry/revision are compared only
+on the shared fail-closed state property; their authority models are different.
+The executed corpus covers the named A19 overlap: two mutation recipes with
+distinct locator/recovery capabilities, plus acquisition/disclosure state,
+capacity, crash and replay contracts. Differences above are required assertions,
+not unexplained drift. Final-head confirmation remains mandatory. Broader native
+guard/fault schedules and installed journeys remain separately required by their
+acceptance rows. No managed guarantee is claimed for AP070.
diff --git a/docs/plans/0074-final-acceptance-audit.md b/docs/plans/0074-final-acceptance-audit.md
new file mode 100644
index 000000000..9a2f24a80
--- /dev/null
+++ b/docs/plans/0074-final-acceptance-audit.md
@@ -0,0 +1,205 @@
+# AP074 final acceptance audit
+
+This is the current requirement-to-producer map, superseding stale outstanding
+work descriptions in historical checkpoints. It does not supersede the spec,
+remove a test or establish that authored checks passed. The final coding batch
+is implemented and its artifacts are constructed. Hosted source
+`eb44d8f7082db6ed9ab46c1bf5319e8e736d7b28` has actual native adapter,
+installed, containment, differential and cost evidence. Its broad implementation
+gate exposed two obsolete committed fuzz seeds; the native-generator repair is
+constructed and awaits hosted confirmation. Formal qualification and the final
+repaired revision's applicable gates remain open. **This is not completion.**
+
+Native test files below are in `product/runtime/auths-gateway/src/`. The
+PostgreSQL lifecycle workflow explicitly runs both durable adapters, including
+ignored PostgreSQL selectors, and retains real execution logs. A selector must
+execute once with one pass and zero ignored; a skipped or zero-match invocation
+is not evidence. Application-owned AP070 remains independent. AP073 is untouched.
+
+## Current executed evidence and remaining gates
+
+The [hosted evidence report](0074-hosted-evidence-eb44d8f7.md) preserves exact
+source, merge producer, tree, closures, authenticated artifact hashes and limits.
+Source `eb44d8f7` and tested merge producer `8335af6a` share tree
+`195e1ae19c0491af1c7e8fdd6d0c63b87655fe89`. Neither may be relabeled as the
+subsequent repaired head.
+
+| Requirements | Actual evidence on source eb44d8f7 | Remaining acceptance gate |
+| --- | --- | --- |
+| P0/P1, E01–E05 | Native owner/consumer and canonical binding/review implementation; installed exact review, consent, request and disclosure controls; signed binding/review corpora execute in the required cohort | Final source/architecture/compliance and repaired-head confirmation |
+| P2/P3, R01–R10, A01–A17/A21 | All 63 required native selectors execute exactly once with one pass and zero failures/ignored tests on file and TLS PostgreSQL; process, capacity, crash, partial-upload, ambiguous-commit, epoch/floor, recovery, ACK and 64-event cases | Final broad repaired-head gate and conditional formal/runtime audit |
+| A18 | Native journal/frame/binding limits and shared 209-vector binding corpus execute; broad gateway library passes 302 tests, zero failures and 63 ignored; separately, all 63 required native selectors and 30 additional PostgreSQL conformance/hostile tests pass | Two stale positive fuzz seeds require native-constructor repair, obsolete refusal controls and repaired hosted corpus confirmation |
+| A19 | Two mutation recipes: 28 request and 26 response comparisons; acquisition/disclosure: 22 pairs after eight named native histories; all 12 reference source hashes match | Repaired-head confirmation with references still independent |
+| A20 | Four-seed concurrent histories per adapter: 32 positive and four negative rounds, 72 workers, independent HTTP/transport-count and historical-byte oracles | Clean source-linked Lean/329-claim audit, two translation reproductions and Kani49 passed; final formal aggregation passed; consumer qualification remains pending at publication |
+| A22 | Installed Airtable: 36 cases/13 verified private audits; Stripe: 23/6, zero inconsistent/unverified. Linux distinct-UID: 33 application cases, 12 admin commands, 53 CLI routes, five private CLI refusals and actual credential/provider/file/privacy controls | Final repaired-head installed/platform confirmation |
+| P4 cost and bounded work | Eight samples per path/backend with actual verification, compilation, approval, callback, adapter, inventory, snapshot-size and Linux RSS distributions | Preserve actual producer and limits; these fixtures establish no production latency or heap SLO |
+| Lean reuse | Complete-input/provenance refusal controls; fast and clean authoritative Lean bootstrap passed with authenticated complete-input artifact | Observed reuse by both consumer Lean jobs while current Rust/source checks still run, pending at publication |
+
+All twelve applicable workflows outside main CI passed on this source. The
+conditional live-custody workflow was skipped and provides no live qualification.
+The failed implementation job and derivative rollup remain failures until the
+complete corpus repair passes; narrow local repair results do not replace the
+hosted gate. Completion still requires every row below and every applicable
+final-head gate.
+
+## Architecture and phase deliverables
+
+| Requirement | Authoritative implementation/deliverable | Required closure evidence |
+| --- | --- | --- |
+| P0 | Spec ownership map, `docs/adr/0017-bound-effect-and-recovery-ownership.md`, `docs/research/domains/abstraction-cases/0009-bound-effect-and-recovery.md`, `0074-differential-contracts.md`, maintained field/limit/state tables | Audit native managed owners and independent reference consumers; architecture/compliance gates |
+| P1 | `effect.rs`, `effect/`, native approval/private complete review; application protocol and thin Python/TypeScript wrappers | Native signed compilation/review/binding corpus plus installed exact approved request |
+| P2 | `store.rs`, typed protocols, `transition.rs`, `entry.rs`, `configuration.rs`; file/PostgreSQL adapters | Actual independent-process capacity, crash, dependency, retention, corrupt-state and durability-fault cohorts |
+| P3 | Native acquisition/disclosure, separate source/recipient/channel approval, one-use grants and exact ACK | Real HTTP/Unix transports, independent read/release charges, revocation, lost ACK and late evidence; installed read/release journey |
+| P4 | Second recipe, independent reference comparisons, installed journey, private/public audit, formal links, cost producers | All A rows and complete exact-head hosted gates; publication on draft PR213, no merge/deploy |
+
+## Effect and recovery invariants
+
+| Requirements | Native owner and independent evidence families |
+| --- | --- |
+| E01 | Sealed verified command and bound effect; `effect_native_tests.rs`, `native_mutation_replay_tests.rs`, installed owner journey compare actual requests |
+| E02 | Complete native binding/authority comparison; `effect_binding_boundary_tests.rs`, `effect_review_approval_tests.rs`, `native_scope_tests.rs`, real retained-input conflicts |
+| E03 | Shared operation key and durable binding conflict; native process capacity, retained replay and generated concurrent histories |
+| E04 | Native verifier, exact approval and fenced current validation remain separate; all three entry-order matrices exercise expiry/withdrawal after admission |
+| E05 | Native review/history projections, strict shared result corpus, public/private audit separation; installed Python/TypeScript/UI and sentinel/privacy cohorts |
+| R01 | Shared typed transition owner and transaction contract; architecture/source inventory and source-linked translated predicates |
+| R02–R03 | Acknowledged atomic admission and final grant precede permit construction; SIGKILL, sync/rename errors, SQL-prefix aborts and lost COMMIT replies on all protocols |
+| R04 | Count/sum/lifetime allocation rows are atomic with admission and retained through uncertainty; actual capacity/fault/cleanup and generated history oracles |
+| R05–R06 | Separate response, observation and ACK states; crash/response-loss/partial-upload/late-ACK cases preserve Unknown and charges without retry |
+| R07 | All three native entry-order cohorts race connection/configuration/epoch and clock changes under actual adapter fences |
+| R08 | Immutable prior record/evidence assertions; exact ACK substitutions and independent recovery history preserve source facts |
+| R09 | Independent recovery actor/read approval and bounded acquisition; `native_recovery_actor_tests.rs` and its limits/history children |
+| R10 | Strict one-format journal decoder, malformed/obsolete native replay corpus, retention tombstones, shared epoch and independently retained generation floors |
+
+The six compositional proof/runtime obligations and assumptions are mapped in
+`0074-assurance-map.md`. Ranges group requirements sharing producers; no invariant
+is waived by grouping. Pure proofs do not establish codecs, signatures, store
+durability, deployment isolation, clocks, recipients or external provider truth.
+
+## Executable acceptance rows
+
+| Row | Concrete producer families and expected observations |
+| --- | --- |
+| A01 | Native compilation/approval fixture, mutation replay and installed owner journey: same effect and exact request at review, approval and real send |
+| A02 | Independent binding and changed-review signed corpora; native scope/read/release/ACK substitutions: old approval refuses, fresh valid control succeeds or unsupported scope refuses before entry |
+| A03 | Native process/capacity, retained replay and generated concurrent histories: at most one entry; same-key changed binding conflicts; fresh-key old approval refuses |
+| A04 | Native acquisition, mutation and disclosure budget matrices on file/PostgreSQL: final count/byte/sum capacity admits exactly the permitted total and preserves all prior rows |
+| A05 | Native acquisition/mutation/disclosure crash matrices; guarded mutation adds 15 parent and 28 child schedules per adapter; partial-upload relay forwards exactly one approved body byte before SIGKILL, retains Unknown/charges and never resends |
+| A06 | All three native file-fault, PostgreSQL partial-batch and real COMMIT-reply-loss cohorts: failed/ambiguous acknowledgement returns no new permit; reopened state is authoritative |
+| A07 | Source acquisition then real disclosure worker/operator revocation race: refusal has zero wire bytes after authentication, while grant-first retains delivery semantics |
+| A08 | Disclosure crashes, lost ACK/commit replies and genuine late ACK: no redelivery, no refund; exact late evidence refines original history |
+| A09 | Native ACK corpus and separate gateway/recipient process corpus: re-signed identity/channel/content/source substitutions leave facts and charges unchanged |
+| A10 | Native candidate/recovery/partial-upload and installed projections: denied response, timeout, response loss, HTTP response, state match and declared linkage remain distinct |
+| A11 | Recovery limits/history: zero/multiple/unrelated/incomplete matches unresolved; exact supported link adds evidence without writes; exhausted calls/bytes/time, aged evidence and unavailable state retain charges/escalation |
+| A12 | Acquisition, mutation and disclosure entry-order matrices: independent operators race revocation, configuration and epoch before/after actual fenced grant; stale worker cannot create a new grant |
+| A13 | All three entry-order lock waits plus prerequisite freshness: independent approval, resource and observer expiry is sampled under the lock; original signed bytes/times remain unchanged |
+| A14 | Genuine threshold authoring and all three entry-order matrices: proof stays Authorized through another branch but changed selected authority cannot reuse old exact approval |
+| A15 | Independent writer/reader recovery and retained history: expired/revoked writer grants no recovery write; bounded separately authorized read preserves attribution and charges |
+| A16 | Native managed scope, shared-store partition/recipe tests and binding substitutions: no cross-task/recipe/partition alias; unsupported delegated participants refuse without another authority |
+| A17 | Public mutation privacy, sentinel output and installed inventory/audit: private facts stay in operator export; public effect/content references use hiding commitments; no secret/proof/body sentinels in public surfaces |
+| A18 | Native binding/review/ACK/frame limits, 209 shared Rust/Python/TypeScript result vectors and new strict journal corpora: canonical positives and malformed/duplicate/obsolete/limit negatives agree at the surfaces exposing each type |
+| A19 | `qualification/reference/ap074_differential.py` and `ap074_information_differential.py`: two maintained mutation recipes (28 requests/26 responses), 22 acquisition/disclosure comparisons, unchanged independent references and explicit claim differences |
+| A20 | Shipping gateway-kernel transition/order/recovery predicates, pinned Charon/Aeneas translation, compiled Lean audit, Kani, refinement/property/mutation suites and generated real concurrent native histories |
+| A21 | Native retention, epoch fence and independent floor: retained terminal/reserved facts survive expired slot cleanup; stale controls and database-only generation rollback cannot recreate entry; missing/malformed floors refuse |
+| A22 | Source-free installed demo owner journey, native SDK/content journey and Linux unprivileged complete interface inventory: consent/decline/substitution controls, read/release/write/observation/replay/unknown and truthful UI; no alternate managed credential/permit route |
+
+This composition does not require every negative through every UI and adapter
+combination. It does require each row's actual boundary, both durable adapters
+where stated, installed binding/UI integration and complete non-bypassability
+inventory. Native authority tests cannot substitute for installed integration;
+installed happy paths cannot substitute for native adverse concurrency/faults.
+
+## Verification and artifact checklist
+
+1. Finish all coding before any constructor or verification run, per AGENTS.md.
+2. Construct the new nine selectors plus the changed dependency/entry/epoch
+ fixtures and the five stale compilation constructors reported by the CI agent.
+ Collect every independent failure; repair one coherent batch without weakening
+ assertions. The four mutation prerequisites are account, two denied reads and
+ relative ceiling. The onboarding probe is separately candidate-authorized.
+3. Refresh gateway semantic closure, generated compliance/API identities and
+ required release metadata once after the source/fixture batch is complete.
+ Review generated changes; never copy decoded hosted assertion byte arrays.
+4. Run required verification on the completed source. The CI agent owns hosted
+ inspection and holistic repairs; retain exact source/tree/closure identities,
+ selector counts, adapter logs, differential reports, cost/history artifact
+ hashes, installed inventory and formal/release evidence.
+5. Resolve required `cargo xtask ci` component gates: formatting, architecture,
+ repository hygiene, build/tests/Clippy, core/MSRV/ABI, wire/compatibility,
+ bindings/packages/demos, WASM/fuzz, compliance and applicable release/formal
+ source freshness/adequacy. Inspect conditional skips for applicability.
+6. Audit every row against final exact-head evidence before claiming complete.
+ Hosted runner capacity cancellations remain missing evidence, not passing
+ platform coverage. No live provider credential, merge or deployment is needed.
+
+## Supported claim boundaries
+
+File storage is a trusted single-host durability/fencing boundary; PostgreSQL
+requires the maintained TLS/transaction configuration and participating writers.
+SIGKILL and modeled syscall faults do not prove power-loss behavior. The partial
+upload test observes application-level prefix delivery, not kernel backpressure
+or provider atomicity. Generation rollback detection requires independently
+retained floors and protects only the state those floors bind. An unchanged
+connection floor does not anchor rolled-back attempt/accounting rows. Physical
+WAL restore establishes before-target retention and after-target exclusion and
+explicitly reports `replay_continuity=false`. Whole-journal rollback and safe
+automatic managed resumption after arbitrary point-in-time restore are outside
+the guarantee. Browser automation proves the authenticated owner protocol and UI
+controls, not human attention. Local recipe/custody fixtures do not qualify a
+live provider. These are explicit premises, not excuses to omit acceptance tests.
+
+AP070 Google metadata/OAuth/productization remains optional future integration.
+Its independent Python/SQLite broker receives no managed permits/credentials and
+its receipts carry none of AP074's managed guarantees.
+
+## Local construction and repair execution
+
+All 22 planned artifact selectors executed. The first aggregate had 21 passes and
+one invalid test signing window. The repair aggregate and final binding constructor
+passed after correcting that window and using an existing registered header; the
+closed header registry was preserved. All 13 binding dimensions now execute even
+when another case fails, before the constructor reports the combined failures.
+All new file/TLS PostgreSQL crash, upload, journal, epoch and floor cases passed.
+
+The complete gateway library run executed 298 passing tests, three failures and
+63 ignored tests. Both application failures were stale generated fuzz/frame bytes
+transitively using the changed approval fixture; their authoritative constructors
+now pass. The other failure was `Unavailable` from a raw file-store race under
+concurrent local I/O. The unchanged exact file-store conformance selector passes
+in isolation. No production deadline, failure handling or assertion was relaxed.
+The final six-selector repair cohort passed, including both dependency adapters
+with epoch activation retaining the same approver, both frame constructors,
+file-store conformance and source closure. The full concurrent run is not relabeled
+as passing. The final hosted broad/adapter/platform cohort remains required.
+
+Gateway all-target/all-feature Clippy, workspace formatting and semantic freeze
+checks passed in the aggregate verification phase. The one subsequent source
+change isolates epoch from approver replacement in the acquisition test; its two
+adapter constructors passed. Source closure is refreshed after that change.
+Release metadata advances only the reported identity/public-SDK/public-surface
+subjects to 172/168/431, with freeze431. These local results construct and check
+the change; they do not close final exact-head hosted or external qualification.
+
+## Final source audit correction — event bound
+
+The P0 abstraction case declared64events per operation, while the native journal
+previously admitted any positive JSON-safe revision. The final correction shares
+one64-revision advancement check across all native protocol/grant/refinement
+writers and enforces it in decoding and read-evidence references. Exact64 retains
+terminal read-only history;65 is corrupt and cannot become a new grant. All
+checks are authored before artifact construction/verification. The source audit
+also reconciles the P0 table's stale five-second lock and1,024-entry inventory
+claims with the stricter maintained two-second deadlines and separately bounded
+inventory/expiry scans. The translated kernel and existing security tests remain
+unchanged; final construction and hosted verification of this correction remain
+pending.
+
+The complete revision-bound correction cohort passed six exact selectors:
+all legal/terminal arithmetic boundaries, file/TLS PostgreSQL retained information
+and mutation journals, and source closure. Every selector matched once with one
+pass and zero ignored tests. Exact64 histories produced no custody, provider or
+recipient entry;65 refused and preserved bytes. All-target/all-feature gateway
+Clippy passed (10.02s), and semantic metadata construction passed without further
+identity changes. Final correction closure is
+`71ff1dcdd7464474a0d1b12f661f7277d665775a6a192128bbbc5870ff9231eb`.
+The correction is committed for integration with the CI agent's combined actual
+977 repair roots; final hosted acceptance remains unverified until that cohort.
diff --git a/docs/plans/0074-final-coding-checklist.md b/docs/plans/0074-final-coding-checklist.md
new file mode 100644
index 000000000..61259f925
--- /dev/null
+++ b/docs/plans/0074-final-coding-checklist.md
@@ -0,0 +1,73 @@
+# AP074 final coding batch
+
+Execution is deferred until this complete coding batch is written, per the
+owner's instruction and AGENTS.md. Do not run individual new tests, artifact
+constructors, fixture generators or replacement CI cohorts while authoring it.
+The existing CI agent may finish observing the already-running 5acd62b0 cohort.
+
+Current checkpoint: the planned coding and constructor batches below are done.
+Source `eb44d8f7` passed all 63 required adapter selectors, installed journeys,
+distinct-UID containment and differential/cost/history workflows. Its broad
+implementation gate found two stale committed fuzz positives. Native constructors
+now generate the complete repair with explicit obsolete refusal controls; hosted
+confirmation remains required. The clean Lean producer now passed with its complete-input attestation. Combine
+this repair and evidence reconciliation in one follow-up publication to establish
+actual Lean reuse; consumer results remain pending at publication. Both unchecked final items remain open until its gates pass.
+
+## Checks to finish writing
+
+- [x] **Guarded Stripe and prerequisite crashes (A05/A06/A10/A13).** Extend the
+ existing native mutation crash harness with the maintained full Stripe recipe,
+ its credential-prefix/account/denied-read checks and relative ceiling.
+ The onboarding probe remains in the independent candidate protocol.
+ Cover all seven existing crash points and all eight reserved-continuation
+ refusals on file and TLS PostgreSQL. Assert exact writes, separate retained
+ prerequisite records, no repeated reads/charges and original signed times.
+- [x] **Partial provider upload (A05/A10).** Exercise a real local connection
+ that forwards only an approved request prefix and then disconnects. Persisted
+ grant/charges remain unknown; fresh process/replay sends nothing again. State
+ the actual socket/relay boundary without claiming kernel backpressure or
+ provider atomicity.
+- [x] **Independent binding substitutions (A02/A16).** Complete the native
+ signed corpus for actor, resource/body/header/destination, account, credential
+ generation, verifier configuration, partition/epoch and approval changes.
+ Retain existing recipient/channel/source/recipe/scope negatives. Each changed
+ binding refuses old exact approval or requires fresh native authorization;
+ unsupported scopes never become a second admission authority.
+- [x] **Concurrent epoch/configuration and rollback floors (A12/A21).** Use the
+ existing dependency/epoch harness to order activation against an old-epoch
+ grant on both adapters. Verify stale configuration/epoch cannot grant, current
+ controls can, retained records/charges survive and supported floors refuse
+ regression. Preserve the explicit whole-store rollback exclusion.
+- [x] **Malformed/obsolete journal records (A18/R10).** Add a shared corpus over
+ retained mutation/acquisition/disclosure records on both adapters: obsolete
+ schema, duplicate/noncanonical members, revision/type bounds, changed immutable
+ binding, inconsistent protocol/evidence and oversized state. Reopen refuses
+ without custody, provider/delivery entry or a regenerated permit.
+- [x] **Acceptance reconciliation.** Map P0–P4, E01–E05, R01–R10 and A01–A22 to
+ exact source/test/artifact producers and explicit assumptions. Resolve stale
+ historical "remaining" notes against current implementation. Missing behavior
+ stays open; existing evidence is not discarded merely because an old note is
+ stale. AP070 stays independent and AP073 stays untouched.
+
+## Final verification phase — only after coding is complete
+
+- [x] Construct all new/changed canonical fixtures together, collecting every
+ independent failure before repairing the complete batch.
+- [x] Regenerate the five stale native compilation constructors identified by
+ the CI agent, covering their fourteen transitive fixture files. Preserve all
+ assertions; use the authoritative constructors, not decoded CI byte arrays.
+- [x] Refresh source closures, compliance/API evidence and required semantic
+ identities once after the complete source/fixture changes.
+- [ ] Run the affected verification as a batch, collect failures, fix them
+ together, and publish the complete revision. The CI agent owns hosted checks.
+- [ ] Audit final exact-head native adapter/concurrency/crash/containment,
+ installed Python/TypeScript/UI, differential, cost, formal/refinement,
+ mutation/release and platform evidence. No skipped required gate or merely
+ authored producer counts as completed acceptance.
+
+Already written before this batch: public/private audit separation, strict
+209-vector result parity, four-path native cost measurement, generated native
+concurrent replay histories, six-property assurance map and five signed changed-
+review/thirteen incomplete-review cases. Retain these; do not expand their scope
+while the remaining checklist is being written.
diff --git a/docs/plans/0074-generated-histories.md b/docs/plans/0074-generated-histories.md
new file mode 100644
index 000000000..eb23dfd22
--- /dev/null
+++ b/docs/plans/0074-generated-histories.md
@@ -0,0 +1,53 @@
+# AP074 generated concurrent durable histories
+
+`native_recovery_actor_tests::generated_history_tests` adds generated concurrent
+mutation histories to the existing all-protocol predicate proofs, sequential
+property tests and native crash/capacity/dependency matrices. The oracle is an
+independent set of previously entered logical operations. It does not invoke the
+shipping reducer to decide its expected number of new external entries.
+
+Four deterministic seeds generate different first-admission orders over four
+native-approved effects plus subsequent mixed replay schedules. Each history
+has eight concurrent rounds: duplicate first admission, two distinct first
+admissions, mixed new/replayed admission, and five generated replay pairs.
+Two real OS workers independently reopen the maintained store, load the installed
+engine, synchronize on a bounded TCP start barrier, then submit the original
+native proof/action/exact approval. Actual writer-fence order is chosen by the OS.
+
+For every round the test compares the model's expected new entries with both
+the real local HTTP capture delta and actual native transport witnesses. It
+checks native journal stage and byte-for-byte retention of every earlier record.
+The source baseline is unchanged throughout. A final concurrent negative uses
+fresh native resource authority for another operation with an earlier exact
+approval. Both workers must refuse before credential lease, reservation or transport;
+the new operation must remain absent and all earlier records remain unchanged.
+
+Each backend therefore produces four histories, 32 positive concurrent rounds,
+four negative concurrent rounds, 72 generated worker processes, 16 new effect
+entries and four baseline writes. The fresh process for every event also checks
+that reopening cannot regenerate a permit from retained Unknown history.
+
+The PostgreSQL workflow explicitly executes the file and TLS PostgreSQL selectors
+and uploads `ap074-generated-histories--`. The two JSON artifacts
+include exact source/checkout revisions, gateway source closure, generated plans,
+actual native results and observed HTTP/native counters. They retain real
+scheduling outcomes rather than pretending concurrent observations have one
+canonical byte representation. They expose no private proof, review, approval,
+record body, response body or credential.
+
+This producer covers generated concurrent single-grant/replay histories. It
+does not independently prove every protocol transition, conservation of bounded
+count/sum capacity, crash durability, hostile-process containment or provider
+qualification. Those remain the responsibility of the existing source-linked
+kernel/refinement and file/PostgreSQL fault, capacity, entry, disclosure and
+independent-recovery evidence. No production decision, journal or retry owner is
+added. The CI agent verified both selectors on source `5acd62b0`, each exactly
+once with one pass and zero ignored, retaining ZIP artifact `13359018734`
+(SHA256 `2f3572c91035a9cd7a022d489dd38377fb38a278ed417bba3aba49705aef4295`).
+Both reports contain all four seeds, 72 workers per backend, actual HTTP/native
+counters and preserved replay bytes. This earlier-source evidence cannot qualify
+the final changed head or close A20 on its own.
+
+Installed trust, signer keys, provider credentials and clock are explicit synthetic
+fixtures. Ordinary native verification and exact signatures still execute; the
+cohort does not claim human presence or an authenticated production clock.
diff --git a/docs/plans/0074-hosted-evidence-5acd62b0.md b/docs/plans/0074-hosted-evidence-5acd62b0.md
new file mode 100644
index 000000000..64ea0f1b7
--- /dev/null
+++ b/docs/plans/0074-hosted-evidence-5acd62b0.md
@@ -0,0 +1,101 @@
+# AP074 hosted evidence: 5acd62b0
+
+Published source `5acd62b0cac44edd80956d7d16eaf15483240a38`, open draft PR213. Non-force publication verified against origin and PR. PR merge producer `4906b8bdef467c6367c35ab9776dbb500a0f03dd`; source and merge both tree `8909ce2164f39f1f87c0427b5f482e253887eb15`. Merge parents: `e2d10a74fa0be5e6bfe2b9f698632c835cd7b11b`, source5acd62b0.
+
+Outgoing commits: aa7cfc76 generated concurrency cohort;5acd62b0 holistic repair plus workspace-anchored cost/history artifacts and source metadata. Previous46989 cohort was fully terminal before publication; no healthy work cancelled/restarted. Prior report `/tmp/ap074-ci-46989-repair-report.md` and snapshot `/tmp/ap074-ci-46989-terminal-checks.json` retain all limitations.
+
+Current cohort is live, not qualified. Expected PostgreSQL execution is48ordinary native selectors plus2cost and2generated-history selectors. Actual metric/history JSON,22information pairs, restore, SDK platform matrix and authoritative formal gates remain required. Published repository semantic closure `c46116f44bcab92337597ef95fc499dd79488a214584135c60135fb82c44150a`, freeze429 unchanged.
+
+Run handles: CI41903640319, PG41903640108, Python41903640032, TS41903640364, SDKrecipes41903640292, recipe41903639997, Airtable41903639933, Stripe41903640179, isolation41903640206, operator41903639815, Google reference41903640081, Git41903640142, branch41903640057, custody41903640054 conditionally skipped. Skipped custody/live provider work is no evidence of qualification.
+
+CI ownership stays isolated in `/tmp/auths-ap074-ci-resumed-20261105`; no main/AP070/AP073 changes.
+
+## Verified execution checkpoint
+
+PostgreSQL job126832350307 succeeded through all stages. All52 exact selectors matched actual log once each,1passed/0failed/0ignored:48ordinary+2cost+2generated. Cost test durations file3.17s/PG3.81s; generated file2.02s/PG7.42s. Downloaded ZIPs verified: cost13358744149 SHA256c2f22d2ed1be9a81a6b9491fe06f88e551acb58c271970288c3fd99318e52825; histories13359018734 SHA2562f3572c91035a9cd7a022d489dd38377fb38a278ed417bba3aba49705aef4295; information13359384484 SHA25666ae2f66e06a09d99c338f3e44388b11951f7f02c5d2d64eb80a0c638abda5ee.
+
+Both cost files retain8samples for each of4paths, actual adapter timing/bytes/callbacks and Linux RSS. Inventory cohorts cover3..17mutation/recovery attempts and2..16acquisition/disclosure attempts, declared4096limit; this is small test-profile measurement, not4096worst-case or production latency. Both generated reports retain4seeds,8positive rounds and one negative pair per seed,72real worker processes/backend,4distinct entries per seed plus baseline. Actual HTTP entries equal independent model and native witness; replay bytes preserved; fresh operation with old approval absent/refused. All cost/history reports bind source5acd, merge4906 and closurec461.
+
+Information report has22pairs, five matching source hashes and run41903640108 attempt1; its source_commit labels actual merge4906, proven tree-equivalent to5acd. Installed native SDK/reference and physical backup/point-in-time restore both passed. Restore explicitly makes no rollback replay-continuity claim.
+
+Python source126832350777 passes439tests/11skips, focused extra1test,ruff,strict adapter mypy,public API snapshot/docs. Native mutation28request26response artifact13360215005 ZIP SHA2568087b02580f4e8364aa82ca1c4f970fedc5d2cf8d2b2bef9d6ba1ad9107a9121verified, all7source hashes match.
+
+Linux isolation126832351540 passed. Artifact13358043624 ZIP SHA2565e0693aa49406acfd3a9a4ed372f8d6c99034f021bcb6e76598c7a5beb9268f0verified: exactsource5acd,33app/12admin/53CLI/5privateCLI, distinctUID, all private/canary/native recipient controls, export0600/1verifiedentry/overwrite refusal. No hostileOS/root or liveprovider claim.
+
+Whole cohort remains live. No source failure observed at this checkpoint; formal/platform and installeddemo completion still required.
+
+
+## Native cost measurements
+
+Producer [PostgreSQL41903640108/job126832350307](https://github.com/auths-dev/auths-proof/actions/runs/41903640108/job/126832350307), attempt1. Downloaded artifact13358744149 ZIP SHA256 `c2f22d2ed1be9a81a6b9491fe06f88e551acb58c271970288c3fd99318e52825` verified. Raw files below preserve source5acd, merge4906 and gateway closurec461 identities above. Eight small test-profile fixtures per path per adapter; no production SLO or worst-case claim.
+
+All duration cells are **minimum / median / maximum milliseconds**, calculated directly from retained integer nanoseconds. Compilation includes native verification. Fenced callback and adapter per-call ranges pool the real execution callbacks/calls across eight samples; adapter totals sum elapsed calls within each execution sample. Adapter duration includes wait/load/decision/persistence/unlock; it is not isolated lock-hold time. Callback duration includes native validation and encoding, with no provider/custody I/O.
+
+| Backend/path | Verification | Compilation | Exact approval | Execution | Fenced callback, per call | Adapter, per call | Adapter total/sample |
+|---|---:|---:|---:|---:|---:|---:|---:|
+| file/mutation | 0.682 / 0.759 / 0.790 | 2.389 / 2.461 / 2.524 | 0.186 / 0.187 / 0.204 | 30.973 / 35.151 / 52.722 | 3.701 / 4.471 / 5.377 | 0.012 / 0.022 / 24.431 | 11.935 / 14.936 / 31.409 |
+| file/recovery | 0.691 / 0.722 / 0.800 | 6.650 / 6.764 / 6.888 | 0.175 / 0.177 / 0.193 | 62.791 / 63.183 / 82.912 | 7.612 / 10.348 / 18.844 | 0.011 / 0.027 / 19.982 | 33.828 / 34.604 / 44.607 |
+| file/acquisition | 0.746 / 0.757 / 0.772 | 3.602 / 3.664 / 3.779 | 0.182 / 0.189 / 0.199 | 47.814 / 54.784 / 65.110 | 5.845 / 7.281 / 9.007 | 0.012 / 0.022 / 15.214 | 18.172 / 18.794 / 25.689 |
+| file/disclosure | 0.716 / 0.753 / 0.782 | 3.216 / 3.323 / 3.369 | 0.187 / 0.189 / 0.200 | 68.578 / 70.135 / 102.108 | 10.345 / 14.997 / 20.797 | 0.011 / 0.025 / 36.314 | 33.468 / 34.663 / 58.944 |
+| postgres/mutation | 0.733 / 0.767 / 0.941 | 3.255 / 3.374 / 5.180 | 0.186 / 0.194 / 0.214 | 38.226 / 38.814 / 40.923 | 3.705 / 4.496 / 5.337 | 0.363 / 0.449 / 7.956 | 18.008 / 18.683 / 20.254 |
+| postgres/recovery | 0.694 / 0.714 / 0.902 | 8.021 / 8.125 / 10.153 | 0.174 / 0.181 / 0.202 | 76.172 / 78.377 / 88.508 | 7.647 / 10.447 / 12.775 | 0.363 / 0.484 / 18.087 | 47.319 / 48.335 / 59.375 |
+| postgres/acquisition | 0.731 / 0.769 / 1.950 | 4.489 / 4.547 / 5.954 | 0.182 / 0.185 / 0.206 | 52.586 / 68.293 / 83.657 | 5.900 / 7.351 / 8.877 | 0.325 / 0.454 / 11.710 | 25.783 / 31.355 / 36.357 |
+| postgres/disclosure | 0.709 / 0.748 / 0.775 | 4.151 / 4.177 / 4.728 | 0.191 / 0.192 / 0.211 | 77.029 / 78.328 / 85.115 | 10.530 / 14.824 / 21.274 | 0.376 / 0.476 / 24.747 | 41.644 / 42.753 / 45.029 |
+
+Snapshot bytes/records are maximum observed boundary values across compilation and execution, not allocator totals. RSS is Linux whole-process memory, including runtime/provider; high-water is cumulative for the test cohort, not an independent peak per path.
+
+| Backend/path | Execution adapter calls / callbacks | Max snapshot bytes | Max snapshot records | RSS KiB min/median/max | Cohort high-water KiB |
+|---|---:|---:|---:|---:|---:|
+| file/mutation | 88 / 16 | 4529 | 3 | 32984 / 33720.0 / 34048 | 34048 |
+| file/recovery | 104 / 24 | 12022 | 18 | 33324 / 33778.0 / 34084 | 34084 |
+| file/acquisition | 136 / 16 | 7332 | 4 | 34416 / 34970.0 / 35272 | 35272 |
+| file/disclosure | 96 / 16 | 16693 | 5 | 34708 / 35024.0 / 35336 | 35336 |
+| postgres/mutation | 88 / 16 | 4529 | 3 | 37860 / 38720.0 / 39036 | 39036 |
+| postgres/recovery | 104 / 24 | 12022 | 18 | 38308 / 38786.0 / 39080 | 39080 |
+| postgres/acquisition | 136 / 16 | 7332 | 4 | 41240 / 41652.0 / 41952 | 41952 |
+| postgres/disclosure | 96 / 16 | 16694 | 5 | 41460 / 41696.0 / 42016 | 42016 |
+
+| Raw artifact member | SHA256 of archived member |
+|---|---|
+| native-cost-file.json | `22f49ae6d0071091a0dbf09923440ab875bb64b4ae29a88550e3654c60083df0` |
+| native-cost-postgres.json | `777c78520adcc125a83a770e1f332015ed7ad2130ecb250e28aec34e96400842` |
+
+## Installed demos
+
+Both source-free installed demo jobs succeeded on5acd. Airtable artifact13360335706 ZIP SHA256 `583802fba557720b70fd3eb98ed419f329eb959428a9823a61db4d8747501686` verified:36simulation cases,private audit13verified/0inconsistent/0unverified; all owner-journey positive controls true, private recipient mode0600, native Python/TypeScript acquisition and independent disclosure/ACK, withheld consent/decline/substitution protection, exact write/read/replay/unknown/fresh-process controls. Browser-simulated authenticated owner and local provider only; human attention, production custody, hostile OS and live provider excluded.
+
+Stripe artifact13359404577 ZIP SHA256 `b81dde3506490e7c678dbf083bc18b602802b273e3f6db3038843c0631b3fbe0` verified:23simulation cases,private audit6verified/0inconsistent/0unverified. Both artifacts explicitly bind source5acd and production_qualification=false.
+
+## Actual failure and deferred final-batch repair
+
+[Authoritative implementation126837641476](https://github.com/auths-dev/auths-proof/actions/runs/41903640319/job/126837641476) failed the gateway library suite:290passed,5failed,59ignored. All five failures are native golden-byte comparisons, not refusals/assertions weakened by this triage. Dependent authoritative rollup126843264528 reflects implementation failure.
+
+Hosted assertion arrays were decoded read-only for forensic field comparison; retained fixture is the LEFT side, newly generated bytes the RIGHT. Files in `/tmp/ap074-ci-5acd-native_*-{retained-left,generated-right}.json` are forensic only and must not be copied into the repository as substitutes for authoritative generators. Shared root:7807 changed `observed_tests::outcome_requirement` from outcome/2 and commitment to outcome/3 and effect-reference. `NativeAuthority::verified` commits normalized installed trust, so the new context propagates to native_authority, effect IDs, approval signatures and source-linked recovery contracts. Candidate/content changes are only authority commitment and dependent identifiers. Prerequisite changes also include the dependent request commitment. Recovery source/action/projection changes follow the new source effect; reservation sorting changes with derived keys, while both unit/amount/ceiling/window tuples remain unchanged: acquisition-count1/16source-lifetime and response-bytes16384/262144source-lifetime.
+
+The five constructors below fan out to14fixture files (later comparisons were unreachable after first mismatch): content-compilation.json; effect-binding.json,effect-policy.json,effect-approval-request.json,effect-approval.json; recovery-compilation.json,recovery-entry.json,recovery-policy-budgets.json,recovery-refinement.json; prerequisite-compilation.json,prerequisite-entry.json; candidate-compilation.json,candidate-journal.json,candidate-staging.json. Preserve every negative/positive assertion and native identity binding.
+
+At the user's instruction, NO new local test/reproduction/generator run or hosted rerun was started. Authoritative regeneration and verification are deferred until the complete implementation batch is coded. For that final phase only, each exact selector uses `AUTHS_UPDATE_FIXTURES=1 cargo test --locked -p auths-gateway --features loopback-provider,testkit-harness --lib -- --exact SELECTOR`:
+
+- observed_tests::effect_native_tests::content_tests::native_content_compilation_fixture_is_current
+- observed_tests::effect_native_tests::native_compilation_approval_and_reverification_keep_exact_request
+- observed_tests::effect_native_tests::native_recovery_acquisition_fixture_is_current
+- observed_tests::effect_native_tests::prerequisite_tests::native_prerequisite_compilation_fixture_is_current
+- observed_tests::effect_native_tests::candidate_tests::native_candidate_compilation_fixture_is_current
+
+Then review all generated differences, regenerate required source/release metadata and perform the coordinated final checks. No new publication is authorized until parent coordinates complete-batch readiness.
+
+## Platform capacity limitation
+
+Installed CPython3.13 macOS job126837943621 and3.14 macOS job126837943795 were cancelled with zero steps. GitHub annotations explicitly report hosted runner acquisition failure after multiple attempts and macOS arm64 capacity limits. These are infrastructure casualties, not test success or a source defect. No rerun initiated and coverage remains mandatory in the final batch. All other completed platform jobs are retained; formal translation is still running at this checkpoint.
+
+## Final terminal result
+
+The superseded5acd cohort is terminal:96checks,80success,2failure,6cancelled,8conditional skips. The only source failure is the five native golden constructors described above, plus authoritative rollup126843264528. Parent reports their14transitive outputs were regenerated through native constructors in the completed977ae61f batch; that changed head requires its own hosted verification.
+
+Two cancellations are macOS3.13/3.14runner-acquisition failures with zero steps. The other four are authoritative Lean126858353087, formal evidence126869045258, formal-translation rollup126869048360 and CI-qualified126869048259. Lean was still running when the completed977ae61f head was published, then this superseded cohort was cancelled. No completed authoritative Lean/final formal qualification is claimed for5acd. This CI agent did not cancel/restart any job or publish the new977head.
+
+Fresh pinned translation126837641240 did complete successfully: two byte-identical reproductions, production closure5c243bbe3196b6f20772b2d90b397ae43d7545312f1fe07a3b05be795fa471e9, generated artifacts c09f533647b96a11cf4bc6b5afad4c6dc326913d1b36cb00437e690ea505c367. This was executed, not reused. Kani126837641513 completed all49harnesses (5+2+7+7+8+20), zero failures. Formal-fast126834539925 completed successfully; feedback evidence is distinct from authoritative Lean.
+
+Downloaded/verified formal ZIPs: translation qualification13361779128 SHA256f8c0c30d91c7067672b1e5c15f92bc259a684ce19b2cdeb2127ceb81202a0c44; translation phase13362601795 SHA2561e7e897b9c2e1464a6b0045e35d949e7ac2ccd47632e58704e99bc4720499989; Kani phase13362140373 SHA25669847b3f289d075a325a494b8e60ef93b194fed8579d10a6b9e8dca505d3acee; proof-fast13361690575 SHA256b0b3c41d00d0465bde9c201dd263a23771835e24af37f92dbeca3481ca664c7e. Phase manifests bind exact5acd source/run41903640319 attempt1, toolchain closurebaff1cabbf9c83ffca0f4bdb1b8dda685a5ac59638e95d6d12e57f5d5bda9633 and lock0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53.
+
+Final raw check snapshot: /tmp/ap074-ci-5acd-final-checks.json. Whole specification acceptance remains dependent on completed-source977ae61f hosted evidence; partial passes and conditional skips are not a waiver.
diff --git a/docs/plans/0074-hosted-evidence-74ce9f75.md b/docs/plans/0074-hosted-evidence-74ce9f75.md
new file mode 100644
index 000000000..e4387e1bf
--- /dev/null
+++ b/docs/plans/0074-hosted-evidence-74ce9f75.md
@@ -0,0 +1,151 @@
+# AP074 hosted evidence — 74ce9f75
+
+This report records execution for published source
+`74ce9f753faa6bf3f1cb065bc369f655e6c0b236` on draft PR #213. It does not
+establish completion of AP074 or qualify later source changes. The remaining
+source defect is the stale gateway semantic closure; its repair belongs to
+the subsequent epoch and metadata batch.
+
+## Producer identity
+
+The PostgreSQL, Python source-contract and external-consumer jobs checked out
+PR merge `df0cc98e6eae440db67038c4f552b74e1d64b90f`, merging the published
+head into `e2d10a74fa0be5e6bfe2b9f698632c835cd7b11b`. Both the merge and
+published source have Git tree
+`bdc10c7f90813049c0b973327f51e5adf3f4f622`, verified by fetching the merge
+and comparing the trees. Workflow source labels do not replace this actual
+producer identity.
+
+## Failure and repaired prior failures
+
+[Authoritative implementation job 123749843777](https://github.com/auths-dev/auths-proof/actions/runs/40958689238/job/123749843777)
+failed `semantic_closure::tests::semantic_closure_is_current`. The gateway
+slice finished with 269 passed, 1 failed and 43 ignored tests. The
+[authoritative rollup](https://github.com/auths-dev/auths-proof/actions/runs/40958689238/job/123754086228)
+failed for that same cause.
+
+A read-only comparison found stale manifest hashes for
+`effect_native_tests.rs`, `engine.rs`, `native_information_replay_tests.rs`,
+`scenario_tests.rs` and `store.rs`; the manifest also omitted
+`native_differential_tests.rs`, `native_retention_tests.rs` and
+`store_retention_tests.rs`. No entries had been removed. The native generator
+must regenerate the gateway manifest and compiled digest from the subsequent
+combined source, followed by a rebuilt non-update check. The unchanged formal
+source closure below is a distinct identity and does not satisfy this gate.
+
+Both earlier CI roots were verified repaired on this head:
+
+- [Repository preflight 123746949082](https://github.com/auths-dev/auths-proof/actions/runs/40958689238/job/123746949082)
+ passed strict workspace lint, including the native differential generator.
+- [PostgreSQL job 123745460518](https://github.com/auths-dev/auths-proof/actions/runs/40958689260/job/123745460518)
+ executed the repaired spawned reference worker, actual SIGKILL and unchanged
+ conservation assertions as part of the successful information comparison.
+
+## Native, differential and restore evidence
+
+The PostgreSQL workflow ran all 24 explicit native selectors exactly once.
+Each selector's actual output was followed by `1 passed; 0 failed; 0 ignored`.
+This comprises the CLI process-capacity case and 23 exact file/PostgreSQL
+library cases, including both nested retention fixtures. The uploaded log
+was downloaded and checked against the selectors in this revision's workflow.
+
+After installing this checkout's native Python SDK, the same job executed
+**22 independent information pairs across both adapters**. The report records
+source commit `df0cc98e6eae440db67038c4f552b74e1d64b90f`, run `40958689260`,
+attempt `1`. Its scope remains post-authorization application state: it does
+not establish AP070 managed authorization, recipient-ACK equivalence or live
+Google behavior.
+
+[Python source-contract job 123745459370](https://github.com/auths-dev/auths-proof/actions/runs/40958688998/job/123745459370)
+executed the exact native mutation generator and the independent comparator:
+**28 requests, 26 responses and 2 families**. This retains the documented
+boundary between native compilation/classification and independent domain
+references; it is not provider execution qualification.
+
+[Reference simulation job 123745461358](https://github.com/auths-dev/auths-proof/actions/runs/40958688981/job/123745461358)
+passed **104 source-only tests**, including mutation-contract negatives and
+information-log parser controls. Synthetic malformed logs prove rejection
+behavior; they are not native execution evidence.
+
+Both differential artifacts were downloaded and their ZIP SHA-256 values
+verified against GitHub:
+
+| Artifact | ID | ZIP SHA-256 |
+| --- | --- | --- |
+| Information report and native log, run 40958689260 attempt 1 | 12934958430 | `c7af0195a0a6645e2cc6fa58f6b0ada451b46692c98649edfac35d4b551b08c0` |
+| Mutation report and native corpus, run 40958688998 attempt 1 | 12933979778 | `3437045757e96a2f1020cb729935e280a2ad8758c125646608f9f766ec1143e0` |
+
+The PostgreSQL job then passed the physical-backup/WAL point-in-time restore
+exercise: the pre-target value was retained and the post-target value was
+excluded. Its report explicitly records `replay_continuity_claim: false`;
+restore success does not establish replay continuity.
+
+## Formal evidence
+
+[Translation job 123749843927](https://github.com/auths-dev/auths-proof/actions/runs/40958689238/job/123749843927)
+passed validated reuse from successful source `d9618a56`, run `40942189152`,
+attempt `1`. Current phase artifact `12935932042` was downloaded and verified
+with ZIP SHA-256
+`0a7c6b905104cd086c5aa247322d28f8764a73c09ca22ddc032636aa685bc76d`.
+It records `execution: reused-pr-run`, `reproduction_count: 2` and these
+identities:
+
+| Identity | SHA-256 |
+| --- | --- |
+| Formal source closure | `a002dc42725b1e3e5234f8ba11eea3f8ba8141e2e64f052d950182d4f93a2825` |
+| Planned translation closure | `a5a98c9c76bde8f03021fd572008961e2c425d11a8eb9ba2080b270b4a65c519` |
+| Toolchain closure | `baff1cabbf9c83ffca0f4bdb1b8dda685a5ac59638e95d6d12e57f5d5bda9633` |
+| Evidence closure | `cbc4400d316703bd0a0af4136f65c580047835a03144353343d051214233269f` |
+
+[Authoritative Lean job 123756976236](https://github.com/auths-dev/auths-proof/actions/runs/40958689238/job/123756976236)
+passed its clean rebuild: 329 compiled claims and reviewed transitive axioms,
+4 reviewed external links with 0 compiled external axioms, and 4 Rust
+refinement tests. [Kani job 123749843917](https://github.com/auths-dev/auths-proof/actions/runs/40958689238/job/123749843917)
+recorded 49 successful verifications. The separate formal-fast feedback job
+also passed; it is not qualification evidence.
+
+[Final formal evidence job 123766644169](https://github.com/auths-dev/auths-proof/actions/runs/40958689238/job/123766644169)
+passed exact-revision aggregation of translation, clean Lean and Kani.
+[Formal-translation rollup 123769550016](https://github.com/auths-dev/auths-proof/actions/runs/40958689238/job/123769550016)
+also passed. Final evidence artifact `12938720181` was downloaded and verified
+with ZIP SHA-256
+`9611da85b53cfdac51ee1c90a2c9667157845962daf82189865dc3abfaf9cd0f`.
+
+## Terminal result
+
+All **96 checks** are terminal: **85 passed, 3 failed and 8 conditionally
+skipped**. The three failures are the authoritative implementation, its
+rollup and [CI qualified 123769584918](https://github.com/auths-dev/auths-proof/actions/runs/40958689238/job/123769584918).
+The final qualified gate records `AUTHORITATIVE: failure`; its formal,
+compliance, dependency, secret and required live-reference inputs succeeded.
+These three failures share the single stale gateway-closure root above.
+
+The **14 workflows** ended with **12 passed, 1 failed (CI) and 1 conditional
+skip (custody live)**. Python's complete supported interpreter/platform
+matrix and installed native harnesses passed, as did packed TypeScript and
+both installed demos. No platform was removed or retried.
+
+## Workflow inventory
+
+| Workflow | Run | Result |
+| --- | --- | --- |
+| CI | [40958689238](https://github.com/auths-dev/auths-proof/actions/runs/40958689238) | Failed: stale gateway closure |
+| PostgreSQL lifecycle | [40958689260](https://github.com/auths-dev/auths-proof/actions/runs/40958689260) | Passed |
+| Python SDK | [40958688998](https://github.com/auths-dev/auths-proof/actions/runs/40958688998) | Passed |
+| TypeScript SDK | [40958689256](https://github.com/auths-dev/auths-proof/actions/runs/40958689256) | Passed |
+| SDK recipes | [40958689237](https://github.com/auths-dev/auths-proof/actions/runs/40958689237) | Passed |
+| Recipe qualification | [40958688981](https://github.com/auths-dev/auths-proof/actions/runs/40958688981) | Passed non-live stages; live qualification stages skipped |
+| Operator package | [40958689233](https://github.com/auths-dev/auths-proof/actions/runs/40958689233) | Passed |
+| Isolation | [40958689244](https://github.com/auths-dev/auths-proof/actions/runs/40958689244) | Passed |
+| Airtable installed demo | [40958689243](https://github.com/auths-dev/auths-proof/actions/runs/40958689243) | Passed |
+| Stripe installed demo | [40958689266](https://github.com/auths-dev/auths-proof/actions/runs/40958689266) | Passed |
+| Independent Google images | [40958689239](https://github.com/auths-dev/auths-proof/actions/runs/40958689239) | Passed application-owned reference |
+| Git signing protocol | [40958688983](https://github.com/auths-dev/auths-proof/actions/runs/40958688983) | Passed |
+| Branch-protection action | [40958689248](https://github.com/auths-dev/auths-proof/actions/runs/40958689248) | Passed |
+| Custody live | [40958689252](https://github.com/auths-dev/auths-proof/actions/runs/40958689252) | Conditional skip |
+
+No workflows were rerun or cancelled during this investigation. No provider
+calls, unattended operator consent, AP070/AP073 source changes, merge or
+deployment were performed. Conditional skips are not passed live evidence.
+The final source head still requires its own applicable CI and acceptance
+checks after the gateway closure and epoch changes.
diff --git a/docs/plans/0074-hosted-evidence-76d008a0.md b/docs/plans/0074-hosted-evidence-76d008a0.md
new file mode 100644
index 000000000..4084344cd
--- /dev/null
+++ b/docs/plans/0074-hosted-evidence-76d008a0.md
@@ -0,0 +1,34 @@
+# AP074 exact-head CI investigation — 76d008a0
+
+Published source: `76d008a020e5b8e6c89e70bcb86ebc34e2c0e29d`, PR213. Repair: `037cb97f5bf2df186f3038ce75e5494d39230e34`, clean isolated checkout, not pushed. Main worktree, AP073 and AP070 source untouched.
+
+## Producer identity
+
+Observed PG and Python source-contract checkouts use PR merge `2507f1387b16fdc6e9da9c51b83130aaa1f32df0`. Its tree `fb2283aa7e1db3550a18ca3cc5cb49a1c532e5fa` equals the published source tree (fetched and diff verified). Source-head labels do not replace this actual producer identity.
+
+## Root matrix
+
+| Root | Hosted evidence | Repair and evidence |
+| --- | --- | --- |
+| Strict clippy rejects missing semicolon in test-only mutation generator | [preflight123733500686](https://github.com/auths-dev/auths-proof/actions/runs/40954513373/job/123733500686), native_differential_tests.rs:384, semicolon_if_nothing_returned | Add semicolon, no lint allowance. `cargo clippy --locked -p auths-gateway --lib --tests --all-features -- -D warnings` passes27.40s. Local log `/tmp/ap074-ci-76d0-gateway-clippy.log`. |
+| Independent information crash reference loses spawned child before readiness | [PG123730998172](https://github.com/auths-dev/auths-proof/actions/runs/40954513640/job/123730998172), child spawn then parent EOFError | Retain parent ownership of Event while spawn child reconstructs its semaphore and until child exits. Exact full runner fails before change and passes22pairs after. Real SIGKILL, abandoned claim, unknown attempt, no reentry/release and conservation assertions unchanged. |
+| Downstream required implementations skipped after preflight failure | [authoritative rollup123736259829](https://github.com/auths-dev/auths-proof/actions/runs/40954513373/job/123736259829), REQUIRED=true/RUN_RESULT=skipped | Fail-closed dependency rollups, not additional independent source defects. No authoritative formal/Kani/translation/compliance/security acceptance claimed. |
+
+## Executed evidence
+
+- PG job actually passed all24 explicit native selectors, including both retention fixtures, before installing the SDK and failing the new information runner. The native log is artifact12933406462, downloaded to `/tmp/ap074-ci-76d0-pg-native.zip` and extracted `/tmp/ap074-ci-76d0-native.log`. Subsequent physical/WAL restore stages were skipped after failure; these are unexecuted on this head.
+- [Python source job123730999956](https://github.com/auths-dev/auths-proof/actions/runs/40954513619/job/123730999956) passed the exact native mutation generator and independent comparator:28requests,26responses,2families. Report artifact12934145521 ZIP SHA256 `a5e66619c969bc7d0067e22bff90996bbdf8fd21e9e80025813a927e832605ee`; downloaded and verified. Report `/tmp/ap074-ci-76d0-mutation-report.json`.
+- [Reference job123730998799](https://github.com/auths-dev/auths-proof/actions/runs/40954513618/job/123730998799) ran104 source-only unit tests successfully, including information parser controls and mutation comparator negatives. Those synthetic parser fixtures establish rejection controls, not native execution.
+- Current-head macOS wheel artifact12933058235 ZIP SHA256 `17c3659e6dbacc1459427daddbf703a3e301c5bed36250dc17b4615ba06d9456` was verified and installed without dependencies/network into `/tmp/ap074-ci-76d0-python`. Exact failed runner replay:
+
+```text
+/tmp/ap074-ci-76d0-python/bin/python qualification/reference/ap074_information_differential.py --native-log /tmp/ap074-ci-76d0-native.log --output /tmp/ap074-ci-76d0-information-repro-current-sdk.json
+```
+
+Result:22pairs,2adapters. This is local repaired reference execution against actual hosted native evidence; it is not a passed hosted information comparator or fresh execution of24native cases. The unchanged AP070 reference emitted Python3.13 SQLite ResourceWarnings during focused tests; no AP070 code changed.
+
+## Terminal matrix
+
+All96checks terminal:65success,12failure,19skipped. Fourteen workflow records:11success,2failure(CI and PG),1conditional skip(custody-live). Independent Python/platform matrix, TypeScript, SDK recipes, installed demos, isolation, operator package and reference simulations pass. [Formal-fast123733500692](https://github.com/auths-dev/auths-proof/actions/runs/40954513373/job/123733500692) passed and explicitly records executed-no-qualification. [Formal-translation rollup123739605415](https://github.com/auths-dev/auths-proof/actions/runs/40954513373/job/123739605415) records REQUIRED=true/RUN_RESULT=skipped; [CI-qualified123739647628](https://github.com/auths-dev/auths-proof/actions/runs/40954513373/job/123739647628) failed closed. These are dependencies of the identified preflight failure, not additional source roots. Authoritative formal/translation/Kani and final qualified evidence did not execute successfully on this revision. No jobs rerun or cancelled. No live provider calls, unattended consent, new PR, merge or deploy.
+
+The two-file repair is one coherent batch. Parent owns combined source closure, independent mutation report source hashes and semantic freeze regeneration. Full AP074 acceptance remains incomplete.
diff --git a/docs/plans/0074-hosted-evidence-977ae61f.md b/docs/plans/0074-hosted-evidence-977ae61f.md
new file mode 100644
index 000000000..88c08862a
--- /dev/null
+++ b/docs/plans/0074-hosted-evidence-977ae61f.md
@@ -0,0 +1,73 @@
+# AP074 hosted evidence and repair batch: 977ae61f
+
+Published source `977ae61f568ea1144d2f8788280bb98f3dbacd55`; PR merge producer `9ae68069aabedbb48cb84e35f79d4e28c33bb435`. Both verified tree `f071d85c7de4f76411db90dec6d7be68f6f5fa2d`; merge parents e2d10a74fa0be5e6bfe2b9f698632c835cd7b11b and source977ae61f. Main published the completed batch; this CI agent does not claim publication of this head.
+
+Read `docs/plans/0074-final-acceptance-audit.md`, `0074-final-coding-checklist.md` and updated AGENTS.md batching policy. Native closure6313af76a8be501ada02f126b5d30713d5baffe49704066b11cd95e59596c159; freeze431/identity172/publicSDK168/surface431. Expected PG exact execution is61 selectors:48ordinary,9final-hardening,2cost,2generated-history. Count the9heredoc selectors separately from shell --exact "$selector". Existing library ignored tests do not establish these passes.
+
+Current14workflow handles: CI41914022680, PG41914023238, Python41914022972, TS41914022951, SDKrecipes41914022783, recipe41914022949, Airtable41914022598, Stripe41914022922, isolation41914023196, operator41914022762, Google reference41914022851, Git41914022750, branch41914022619, custody41914022898conditionally skipped. The cohort is terminal and not qualified: 80 successful, seven failed and nine skipped checks (96 total).
+
+No new local tests/reproductions or workflow reruns during initial triage. Collect all actual roots, complete coherent repairs and only then run the affected batch; no piecemeal verification or weakened assertions/deadlines/platforms. AP070/AP073 untouched, main checkout untouched.
+
+## Confirmed repair roots and execution limits
+
+The PostgreSQL lifecycle job [126868788825](https://github.com/auths-dev/auths-proof/actions/runs/41914023238/job/126868788825) and recipe simulation job [126868788411](https://github.com/auths-dev/auths-proof/actions/runs/41914022949/job/126868788411) expose two root causes. The new review-approval test module imports feature-gated `harness::authoring` without its own matching feature guard, breaking ordinary/loopback-only library test builds. The final-hardening runner additionally requires an unavailable `rg` executable after every successful native selector. Authored repairs guard the module, retain its execution as a new explicit harness-enabled review selector, preserve the exact one-pass/no-ignore assertion through portable `grep -Fq`, and expose rendered Cargo compiler diagnostics before the qualification build fails. No native assertion or coverage is removed. The complete affected repair batch passed the focused verification recorded below.
+
+Retained PostgreSQL artifact 13365141010 has verified ZIP SHA-256 `99fff0461d934d43cd18f76beb8442eb2a04927a48adc0e4044fa23b9e3b13ce`. It contains the native log only: all 48 ordinary required native selectors executed exactly once with one pass and zero failed/ignored; the separate production-process connection test also passed. The final-hardening artifact 13363718099 has verified ZIP SHA-256 `3355e513ea52617b5eb2c38d724bc6d7b2da51c47095159ff332ed4d847e428a`: all nine required native selectors likewise passed once without ignores. Their shell postchecks failed solely because `rg` was absent. The cost, generated-history, independent information comparison and backup/restore stages were skipped downstream; their execution is not established on this head. The repaired inventory will add the review and operation-revision-limit selectors, becoming 48 ordinary + 11 hardening + 2 cost + 2 generated = 63 explicit native selectors.
+
+## Verified retained successes
+
+The independently compared mutation artifact 13364970704 has ZIP SHA-256 `a87ff2a8c25c6241f17c7ea40b3330bbbcd7616e5f0a4d4f26dd744f7707e4bc`; all seven source hashes match this published checkout. Its unchanged contract remains 28 request and 26 response comparisons.
+
+The actual distinct-UID Linux artifact 13364741168 has ZIP SHA-256 `2ae4f17ad88620937c8b1e0a47b7ca4e0c870f606b999a919e8288e5a150628b` and source977ae61f. It records 33 application cases, 12 admin commands, 53 CLI routes and five private CLI refusals; 45 initial/49 replay private files are inaccessible. Source journal bytes are preserved, unauthorized direct-provider requests return401, recipient UID/key guards and trusted positive control hold, public channels exclude plaintext/encoding/raw-digest canaries, and private export is0600 with one verified audit entry and overwrite refusal.
+
+Stripe installed evidence artifact13364616512 has ZIP SHA-256 `ac6288adff74707b8852a07c4fea830231681bbbb7fe6ae604db938597b7ee50`, source977ae61f,23 simulation cases, six verified private audit entries and zero inconsistent/unverified entries. Production qualification is false; this is installed simulation evidence, not live-provider/human-attention qualification.
+
+Airtable installed artifact13364786785 has verified ZIP SHA-256 `69c328abc9a431a34cb6654404a17f4b36f81b71c474bb459e9ea3974b15a495`, source977ae61f. It retains36 simulation cases and13 verified private-audit entries with zero inconsistent/unverified entries. The separate source-free owner browser report has every positive assertion true: unchanged native renderer, approvals withheld until consent, decline/substitution refusal, private owner credentials isolated, one exact write and independent read, installed Python/TypeScript acquisition, separately consented native disclosure with retained recipient ACK, no public HTTP private content, no replay redelivery, same actor read/disclosure/write, unknown response loss without reentry and fresh-process replay without new consent/execution. Production qualification and independent-human-trial remain false.
+
+Formal translation job126874540607 passed using validated same-PR reuse from source5acd62b0, run41903640319 attempt1; it is not a new translation reproduction. Current retained phase artifact13364504382 ZIP SHA-256 `e8a6d15494faebafecbfba0ad7773db50dedf9231c6e3620dee745c2727fc8b7` explicitly records current source977/run41914022680/attempt1 and `execution=reused-pr-run`. Qualification artifact13364957697 ZIP SHA-256 `6c0a539386ccc7d5219ffd45aa9c0bab98ff338203260e33a43df1bbaf0482c2` preserves production closure `5c243bbe3196b6f20772b2d90b397ae43d7545312f1fe07a3b05be795fa471e9` and generated-artifact hash `c09f533647b96a11cf4bc6b5afad4c6dc326913d1b36cb00437e690ea505c367`. The matching translated closure permits reuse; separate authoritative Lean/final evidence remains required.
+
+## Authoritative transitive fixture failure
+
+Authoritative implementation job [126874540692](https://github.com/auths-dev/auths-proof/actions/runs/41914022680/job/126874540692) reached the complete gateway library: **301 passed, zero failed,63 ignored** in62.39s. This clears the earlier broad-run file-store conformance concern for published977 without changing its deadline or assertion. Separately required ignored native cases remain evidenced only by their explicit producer runs.
+
+The workspace then failed `native_application_content_cli_fixture_is_current` at application_content_cli.rs:130. Offline decoding of the actual hosted assertion arrays establishes exactly one differing JSON leaf: `cases[0].request.submission.approval_b64`. Both actual captured CLI request and response equality assertions passed. The real frame corpus had been regenerated for native authority changes, but its downstream CLI golden still carried the previous approval. A search across every tracked file found the old exact approval in only two remaining consumers: `application-content-cli.json` and `application-recovery-cli.json`; both differ from their respective frame corpus only in this approval field. Repair regenerates both through their existing production-CLI/real-Unix-socket constructors, not by copying assertion bytes. No proof, approval, framing, response or fixture assertion is relaxed. The authoritative rollup failure is derivative of this root.
+
+The parent correction909193d6 (amended fromf659217c, identical code/artifacts) enforces the declared64operation-revision limit and corrects bound-table/spec language. The combined repair preserves it and explicitly requires `store::operation_revision_limit_preserves_last_event_without_wrap_or_restart` alongside the18-case review selector. That native repair initially preserved freeze431. The later complete Lean-input and assurance-citation correction requires freeze432, as recorded below.
+
+## Focused repair verification
+
+All three root causes are repaired together, with parent909193d6's revision bound preserved. Local failed-check reproductions and required artifact construction were completed after the corresponding batch was authored:
+
+- `cargo test --locked -p auths-gateway --lib --no-run`: passed20.80s.
+- `cargo test --locked -p auths-gateway --features loopback-provider --lib --no-run --message-format=json`: passed19.96s, preserving the qualification packer's structured build report.
+- Harness-enabled exact review and revision-bound tests: one passed/zero failed/zero ignored each. Review command took24.58s including build; revision command0.54s.
+- Native semantic-closure update and rebuilt unchanged check: one passed/zero failed/zero ignored each. Final closure `61173a0020911d7fb40fe9870a81b9b19a3641b1f5c406c0eeb9b9db3a3b88aa`.
+- `AUTHS_UPDATE_FIXTURES=1 cargo test --locked -p auths-gateway --all-features --test application_content_cli --test application_recovery_cli`: both constructors passed, total16.50s including build. Same command without update passed both in1.01s. Structural comparison confirms each generated golden changes only the stale approval packet.
+- Both changed workflow shell bodies pass `bash -n`. The portable success assertion accepts all nine actual retained hardening logs and rejects absent, failed and zero-pass/ignored evidence.
+- `cargo fmt --all` and `git diff --check` passed. Final `cargo xtask semantic-freeze --update` passed4.50s without changing any release counter or inventory; freeze431 remains.
+
+Logs and machine-readable local results: `/tmp/ap074-ci-977-verification`. These focused results do not replace the required next-head hosted workspace, formal, SDK, installed and PostgreSQL gates. No broad suite was redundantly rerun locally, no workflow restarted, and no time bound, assertion or platform coverage was weakened.
+
+## Terminal cohort and formal evidence
+
+The final source977 cohort contains **96 checks: 80 success, seven failure and nine skipped**. The failures are PostgreSQL126868788825, recipe126868788411, authoritative implementation126874540692 and its rollup126880158416, authoritative Lean126883243174, formal rollup126895630346 and qualified rollup126895668915. The implementation roots are the feature scope, missing runner utility and stale transitive CLI fixtures above. The remaining independent root is stale assurance source citations, described below; rollup failures are derivative. No jobs were manually cancelled or retried.
+
+All installed Python platforms (3.9–3.14 on Linux/macOS/Windows), typing/harness, TypeScript matrix/browser/harness, SDK journeys/scaffold, installed demos and isolation passed. Actual macOS execution resolves the earlier runner-capacity gap for this source. PostgreSQL cost, generated history, the 22 information comparisons and restore stages remain unexecuted on977 because of the earlier failure. Their earlier-head passes do not qualify977 or its repair.
+
+Formal-fast126872323162 passed. Kani126874541107 passed all49 harnesses (5+2+7+7+8+20), zero failures. Its artifact13365354090 has ZIP SHA-256 `3540ea711708d0096f35f8dd76d0c3a509708215d1400bbdbcd0cf0a1e4496c5` and binds source977/run41914022680/attempt1. Translation's validated reuse provenance is recorded above.
+
+[Authoritative Lean126883243174](https://github.com/auths-dev/auths-proof/actions/runs/41914022680/job/126883243174) started at21:39:44UTC and began its clean rebuild at21:43:01. At21:56:27 the compiled assurance audit passed **329 statements with reviewed transitive axioms**, with matching proof coverage. At22:02:49 the shipping Rust refinement tests passed four/zero failed/zero ignored. At22:02:58 the existing bounded-update drift guard correctly failed because59 source-closure citations in the assurance manifest were stale. This is successful compiled proof/audit execution followed by a failed qualification gate, not a qualified Lean phase. The diagnostic artifact13367207757 has ZIP SHA-256 `a7d76cd18696138151f934757f4cfe7b14884439b97f188753c515acc41aab6c` and contains the actual Lean log and compiled329-claim audit. Formal-fast diagnostic artifact13365961292 has ZIP SHA-256 `446923859d85495656235e665252089835b9a24f90ddacbdc1a1b851b1be2cf5`.
+
+The bounded automatic updater, not this agent, pushed `ee7975be5efe8ba93a62570a67b660b816b7ac45` during the manual publication hold. Its only change is59 source-closure SHA citations. Downloaded updater artifact13368050220 has ZIP SHA-256 `501763ff25d4282686d2cc69bd764f45f5ccbd008312ba475abc2022144755e0`, source977/run41914022680/attempt1; its manifest file is byte-identical to that commit. The combined repair preserves this commit in its ancestry.
+
+## Complete Lean reuse and citation repair
+
+The owner additionally requested that native/docs-only pushes stop repeating expensive unchanged Lean builds. The complete contract is [0074-lean-reuse-contract.md](0074-lean-reuse-contract.md). Both Lean jobs now accept only successful clean same-repository/same-PR authoritative evidence with matching complete transitive proof/generated/import/toolchain/configuration/assurance/executor inputs, validated exact producer attempt and artifact digest. Current-head Rust refinement and source/freshness checks still execute. Missing, changed, legacy or untrusted evidence always invokes fresh Lean. Commit IDs and run timing are provenance, never proof-key inputs.
+
+The native metadata constructor now refreshes only existing source-closure citations after source-closure generation. Typed negative controls prove all theorem statements, axioms, claims, reviews, semantic bindings and other evidence remain unchanged. The compiled assurance audit and update drift guard remain required. This prevents a native metadata refresh from producing another automatic citation-only push without treating metadata generation as proof qualification.
+
+All affected verification was run as a complete batch: nine Python provenance tests,35 planner tests,21 native phase/assurance tests and25 qualification-policy tests passed; strict xtask/planner Clippy, phase ownership and source-closure checks passed. The metadata constructor initially identified the required identity advances; after applying only those advances, update and non-update verification passed (62 entries,43 public Rust packages). Logs: `/tmp/ap074-ci-lean-reuse-final`; earlier gateway repair logs: `/tmp/ap074-ci-977-verification`.
+
+The combined formal source closure is `fe66a1c24450b555e7e28d775c5796b68bc0d896445181343c6f8178d3f43751`. Exactly59 source citations were refreshed; normalizing only those hashes makes the complete typed manifest identical to the automatic updater's version. Freeze432 advances formal assurance bytes62, formal source-closure bytes107, evolution85 and public surface432; all other release identities remain unchanged. Native gateway closure remains `61173a0020911d7fb40fe9870a81b9b19a3641b1f5c406c0eeb9b9db3a3b88aa`.
+
+The repaired head must execute all63 explicit native selectors, restored downstream reports and the complete installed/formal/release gates. The first Lean run is deliberately fresh because the old producer lacks the complete-input attestation and the verifier changed. A required final evidence/acceptance-ledger documentation update is planned to demonstrate a real warm hit in both Lean jobs with current-head checks still running. Synthetic negative controls are not a hosted reuse claim. AP074 remains incomplete until final-head acceptance is established.
diff --git a/docs/plans/0074-hosted-evidence-9913d4c7.md b/docs/plans/0074-hosted-evidence-9913d4c7.md
new file mode 100644
index 000000000..96b6d4e12
--- /dev/null
+++ b/docs/plans/0074-hosted-evidence-9913d4c7.md
@@ -0,0 +1,108 @@
+# AP074 hosted evidence: 9913d4c7
+
+Observed terminal state: 29 October 2026, 15:07 UTC.
+[PR #213](https://github.com/auths-dev/auths-proof/pull/213) source revision:
+`9913d4c702451c19de2b601cd634e397db20f951`.
+
+All 96 check records finished: **88 success, eight conditional skips, zero
+failures**. All fourteen workflows finished: thirteen success, one conditional
+skip. This revision is CI-qualified. This is evidence for this published source
+only, not for subsequent disclosure tests or complete AP074 acceptance.
+
+## Workflow results
+
+| Workflow | Run | Result |
+| --- | --- | --- |
+| CI | [40889129833](https://github.com/auths-dev/auths-proof/actions/runs/40889129833) | Success, including CI-qualified |
+| PostgreSQL lifecycle | [40889129739](https://github.com/auths-dev/auths-proof/actions/runs/40889129739) | Success; all 18 explicit native cases executed |
+| Python SDK | [40889129804](https://github.com/auths-dev/auths-proof/actions/runs/40889129804) | Success; wheel producers, installed consumers and gateway harnesses |
+| TypeScript SDK | [40889129810](https://github.com/auths-dev/auths-proof/actions/runs/40889129810) | Success; packed Node/Chromium and gateway harnesses |
+| SDK outcome recipes | [40889129755](https://github.com/auths-dev/auths-proof/actions/runs/40889129755) | Success; installed examples, scaffold and recipes |
+| Gateway operator package | [40889129870](https://github.com/auths-dev/auths-proof/actions/runs/40889129870) | Success; source-free extraction/simulation |
+| Recipe qualification | [40889129736](https://github.com/auths-dev/auths-proof/actions/runs/40889129736) | Success for simulation/candidate/offline paths; conditional live/release jobs skipped |
+| Gateway isolation | [40889129561](https://github.com/auths-dev/auths-proof/actions/runs/40889129561) | Success with distinct UIDs |
+| Airtable Docker demo | [40889129895](https://github.com/auths-dev/auths-proof/actions/runs/40889129895) | 36 cases passed; production qualification false |
+| Stripe Docker demo | [40889130083](https://github.com/auths-dev/auths-proof/actions/runs/40889130083) | 23 cases passed; production qualification false |
+| Google installed images | [40889129830](https://github.com/auths-dev/auths-proof/actions/runs/40889129830) | Success for independent application-owned reference |
+| Git signing protocol | [40889129499](https://github.com/auths-dev/auths-proof/actions/runs/40889129499) | Success |
+| Git signing branch protection | [40889129611](https://github.com/auths-dev/auths-proof/actions/runs/40889129611) | Success |
+| Gateway custody live | [40889129882](https://github.com/auths-dev/auths-proof/actions/runs/40889129882) | Conditional skip; no qualification evidence |
+
+The eight skipped jobs concern the production credential-store journey, Stripe
+test-mode journey, custody store contract, independent first qualification
+signer, independent finite commissioning reviewer, retained production operator,
+complete cleaned-release signing and gateway release verification. No skipped
+job is counted as a pass or live-provider/human-review evidence.
+
+## Formal and implementation results
+
+- [Authoritative implementation](https://github.com/auths-dev/auths-proof/actions/runs/40889129833/job/123506132526)
+ and its rollup passed. Its repeated configurations report 3,008 passed,
+ zero failed and 97 ignored across 609 result suites; these are not unique-test
+ counts. Feature-gated native cases executed separately below.
+- [Compliance](https://github.com/auths-dev/auths-proof/actions/runs/40889129833/job/123506132619),
+ dependency/secrets checks and all three live-reference implementations
+ (OpenTofu, PostgreSQL and records API) and their rollups passed.
+- [Kani](https://github.com/auths-dev/auths-proof/actions/runs/40889129833/job/123506132424)
+ passed with 49 actual successful verification results.
+- [Authoritative Lean](https://github.com/auths-dev/auths-proof/actions/runs/40889129833/job/123512213089)
+ completed a clean 3,270-job build, 329 compiled statements and transitive-axiom
+ audit, four reviewed external links with zero compiled external axioms, four
+ Rust refinement tests and the 49-harness inventory. Fast Lean is separate
+ feedback and is not substituted for this result.
+- [Final formal evidence](https://github.com/auths-dev/auths-proof/actions/runs/40889129833/job/123528105291)
+ passed with translation, clean Lean and Kani (artifact `12907997775`). The
+ formal-translation rollup and
+ [CI-qualified](https://github.com/auths-dev/auths-proof/actions/runs/40889129833/job/123533062348)
+ also passed.
+
+### Translation provenance
+
+[Pinned translation](https://github.com/auths-dev/auths-proof/actions/runs/40889129833/job/123506132456)
+passed by validated same-PR reuse, not a fresh translation. Artifact
+`12905565623` records current head `9913d4c7`, run `40889129833`, attempt 1,
+`execution=reused-pr-run`, reproduction count 2, and reuse from run
+`40867724155`, attempt 1, source head
+`ab36e4b76420a37912d00b2c5b1fac1ad6c556c5`.
+
+The source closure is
+`a002dc42725b1e3e5234f8ba11eea3f8ba8141e2e64f052d950182d4f93a2825`.
+The translation phase closure is
+`a5a98c9c76bde8f03021fd572008961e2c425d11a8eb9ba2080b270b4a65c519`.
+The toolchain closure is
+`baff1cabbf9c83ffca0f4bdb1b8dda685a5ac59638e95d6d12e57f5d5bda9633`.
+The evidence closure is
+`cbc4400d316703bd0a0af4136f65c580047835a03144353343d051214233269f`.
+
+## Native durable execution and producer identity
+
+[PostgreSQL job](https://github.com/auths-dev/auths-proof/actions/runs/40889129739/job/123500802885)
+ran all eighteen explicit cases with one pass and zero ignored each: CLI content
+capacity; file/PostgreSQL process capacity, acquisition crash, mutation crash,
+dependency ordering and fixed-window budgets; file I/O failures; PostgreSQL
+partial-batch rollback and COMMIT reply loss; file/PostgreSQL mutation and
+information retained replay. Mutation crash cases took 1.98 seconds on file and
+10.16 seconds on PostgreSQL. Thirty existing gateway PostgreSQL cases, lifecycle
+and operator-plane tests also passed.
+
+Physical/WAL restore retained before-target and excluded after-target state.
+`replay_continuity_claim=false` remains explicit: a restored database does not
+establish its own freshness.
+
+The actual Linux wheel/PostgreSQL producer checkout was PR merge commit
+`16eb79e57fde2b82fca5550c8e2b4c5596a79046`, with verified tree
+`b8913649cbe69cbe2b4aac8339f5c9fedf708a77`, identical to the published source
+tree. Preserve that producer identity; it is not literally the source commit.
+Demo producers separately checked out the source head.
+
+Both installed Docker demos passed the original substituted-challenge assertion
+unchanged. Neither claims production qualification. Google remains the
+independent AP070 reference and receives no managed-native guarantee here.
+
+## Limits
+
+No CI repair or rerun was needed for this revision. Its final formal run finished
+before the next publication. Later commits require fresh exact-head hosted
+checks. The [implementation ledger](0074-implementation-status.md) remains
+authoritative for the open AP074 acceptance gates; this report closes no
+unexecuted human UI, differential, concurrency, retention or performance claim.
diff --git a/docs/plans/0074-hosted-evidence-ce9d6cf5.md b/docs/plans/0074-hosted-evidence-ce9d6cf5.md
new file mode 100644
index 000000000..872e7b6af
--- /dev/null
+++ b/docs/plans/0074-hosted-evidence-ce9d6cf5.md
@@ -0,0 +1,84 @@
+# AP074 final hosted evidence — ce9d6cf5
+
+Observed terminal state: **2026-10-29 16:13 UTC**. [PR213](https://github.com/auths-dev/auths-proof/pull/213).
+
+Published source head: **`ce9d6cf551af226d1a13d28f2de027ae5b81cbab`**. Main CI is qualified, but the complete PR matrix is **not all green**: two macOS Python consumers never acquired hosted runners, including their one authorized recovery attempt.
+
+All 96 latest check records are terminal: **86 success, 2 cancelled, 8 conditional skips**. All 14 workflows are terminal: **12 success, 1 failure (Python infrastructure), 1 conditional skip (custody live)**. No source/test defect was identified. Isolated checkout `/private/tmp/auths-ap074-ci-c80902ab` is clean at the published head; no source edits, local gates, commits or pushes this round. Parent's unpublished A13/A14, A15 and owner UI work is outside this evidence. Main/AP073/AP070 source untouched.
+
+## Complete workflow matrix
+
+| Workflow | Run | Result |
+|---|---|---|
+| Main CI | [40899439849](https://github.com/auths-dev/auths-proof/actions/runs/40899439849) | Success, including final formal evidence and CI-qualified |
+| PostgreSQL lifecycle | [40899439948](https://github.com/auths-dev/auths-proof/actions/runs/40899439948) | Success; all 20 selected native cases actually ran |
+| Python SDK | [40899439869](https://github.com/auths-dev/auths-proof/actions/runs/40899439869) | Attempt 2 failed solely on two runner-acquisition cancellations; 24 other jobs successful |
+| TypeScript SDK | [40899439801](https://github.com/auths-dev/auths-proof/actions/runs/40899439801) | Success; packed Node/Chromium and both native harnesses |
+| SDK outcome recipes | [40899439802](https://github.com/auths-dev/auths-proof/actions/runs/40899439802) | Success; installed Python/TS examples, scaffold and recipes |
+| Gateway operator package | [40899439793](https://github.com/auths-dev/auths-proof/actions/runs/40899439793) | Success; source-free extraction and simulation |
+| Recipe qualification | [40899439949](https://github.com/auths-dev/auths-proof/actions/runs/40899439949) | Simulation/candidate/offline success; conditional live/release reviewer jobs skipped |
+| Gateway isolation | [40899439954](https://github.com/auths-dev/auths-proof/actions/runs/40899439954) | Distinct-UID success |
+| Airtable Docker demo | [40899439876](https://github.com/auths-dev/auths-proof/actions/runs/40899439876) | 36 cases passed; production_qualification=false |
+| Stripe Docker demo | [40899439762](https://github.com/auths-dev/auths-proof/actions/runs/40899439762) | 23 cases passed; production_qualification=false |
+| Google installed images | [40899440046](https://github.com/auths-dev/auths-proof/actions/runs/40899440046) | Independent application-owned reference passed; no AP070 managed-native claim |
+| Git signing protocol | [40899439950](https://github.com/auths-dev/auths-proof/actions/runs/40899439950) | Success |
+| Git signing branch protection | [40899439930](https://github.com/auths-dev/auths-proof/actions/runs/40899439930) | Success |
+| Gateway custody live | [40899439846](https://github.com/auths-dev/auths-proof/actions/runs/40899439846) | Conditional skipped; no evidence |
+
+The eight skipped jobs are production credential-store journey, Stripe test-mode journey, custody store-contract, independent first qualification signer, independent finite commissioning reviewer, retained production operator, complete cleaned-release signing and gateway release verification. None counted as passed. All required main CI implementations and formal phases passed; none were preflight-skipped.
+
+## Sole remaining root: macOS hosted-runner capacity
+
+Python attempt 1 finished with 24 successful jobs and exactly two cancelled consumers:
+
+- [CPython 3.10 macOS, job123543596333](https://github.com/auths-dev/auths-proof/actions/runs/40899439869/job/123543596333).
+- [CPython 3.11 macOS, job123543596002](https://github.com/auths-dev/auths-proof/actions/runs/40899439869/job/123543596002).
+
+Both had no runner and no steps. GitHub's failure annotation says: **“The job was not acquired by Runner of type hosted even after multiple attempts”**, accompanied by macOS ARM capacity notices. This is not a test or source failure.
+
+After the complete Python matrix settled, the parent explicitly authorized one grouped recovery attempt. The connector returned 403 without mutating anything; the authorized `gh` fallback issued exactly one `POST /repos/auths-dev/auths-proof/actions/runs/40899439869/rerun-failed-jobs`. Attempt 2 started at 15:47:45 UTC on the same source head. Only the unsuccessful consumers were queued for execution:
+
+- [CPython 3.10 macOS retry, job123557778142](https://github.com/auths-dev/auths-proof/actions/runs/40899439869/job/123557778142).
+- [CPython 3.11 macOS retry, job123557777897](https://github.com/auths-dev/auths-proof/actions/runs/40899439869/job/123557777897).
+
+Both retries were cancelled again before execution with the same runner-acquisition/capacity annotations. Attempt 2 finished at 16:02:46 UTC. **No second retry was issued.** Healthy jobs and the separate main/formal workflow were not rerun or cancelled. GitHub copied successful job records to new attempt handles while retaining original execution timestamps: for example the macOS wheel's copied job123557836972 still completed at 15:24:12 UTC, before retry submission. Original successful producer artifacts therefore remain attempt-1 evidence.
+
+Full original and retry job records/annotations: `/tmp/ap074-ci-ce9d-python-retry.json`. No platform, assertion, deadline or approval requirement was removed. The unresolved coverage is specifically the installed CPython 3.10/3.11 macOS consumers on this source head. Other 16 Python consumers, all three wheel producers, typing/API, source tests and both gateway harnesses passed.
+
+## Main CI and formal qualification
+
+- [Preflight123540248975](https://github.com/auths-dev/auths-proof/actions/runs/40899439849/job/123540248975): passed.
+- [Authoritative123543879706](https://github.com/auths-dev/auths-proof/actions/runs/40899439849/job/123543879706) and [rollup123558564984](https://github.com/auths-dev/auths-proof/actions/runs/40899439849/job/123558564984): passed. Log has 609 result suites totaling 3,010 passed, zero failed and 98 ignored across repeated configurations, not 3,010 unique tests. Feature-gated native cases executed separately below.
+- [Compliance123543880074](https://github.com/auths-dev/auths-proof/actions/runs/40899439849/job/123543880074), security, dependencies and all three live-reference implementations (OpenTofu/PostgreSQL/records API) plus rollups: passed.
+- [Kani123543879775](https://github.com/auths-dev/auths-proof/actions/runs/40899439849/job/123543879775) and [rollup123558055121](https://github.com/auths-dev/auths-proof/actions/runs/40899439849/job/123558055121): passed, with **49 actual successful verifications**.
+- [Authoritative Lean123552204257](https://github.com/auths-dev/auths-proof/actions/runs/40899439849/job/123552204257): clean rebuild and audit passed for **329 compiled statements**, reviewed transitive axioms, **four reviewed external links and zero compiled external axioms**, and all four Rust refinement tests. Fast Lean123540249162 also passed; it remains feedback rather than qualification.
+- [Formal evidence123567518646](https://github.com/auths-dev/auths-proof/actions/runs/40899439849/job/123567518646): **PASS at 16:11:45 UTC**, explicitly `translation + clean Lean + Kani`; artifact12912697748.
+- [Formal-translation rollup123572141294](https://github.com/auths-dev/auths-proof/actions/runs/40899439849/job/123572141294): passed.
+- [CI-qualified123572202035](https://github.com/auths-dev/auths-proof/actions/runs/40899439849/job/123572202035): passed.
+
+### Translation reuse provenance
+
+[Translation123543879863](https://github.com/auths-dev/auths-proof/actions/runs/40899439849/job/123543879863) passed by validated same-PR reuse, not fresh translation execution. Phase artifact12909767960 records current head ce9d, run40899439849, attempt1; execution=`reused-pr-run`; reproduction_count=2; reused_from head9913d4c7/run40889129833/attempt1. The durable previous-head report records its underlying fresh ab36/run40867724155/attempt1 two-reproduction provenance. Preserve this chain rather than claiming two new builds.
+
+- Source closure: `a002dc42725b1e3e5234f8ba11eea3f8ba8141e2e64f052d950182d4f93a2825`.
+- Translation phase: `a5a98c9c76bde8f03021fd572008961e2c425d11a8eb9ba2080b270b4a65c519`.
+- Toolchain closure: `baff1cabbf9c83ffca0f4bdb1b8dda685a5ac59638e95d6d12e57f5d5bda9633`.
+- Evidence closure: `cbc4400d316703bd0a0af4136f65c580047835a03144353343d051214233269f`.
+
+Saved phase artifact: `/tmp/ap074-ci-ce9d-formal-translation-phase.zip`.
+
+## Actual native PostgreSQL execution
+
+[PG job123537375377](https://github.com/auths-dev/auths-proof/actions/runs/40899439948/job/123537375377) executed all **20 explicit native selectors**, each one passed and zero ignored: CLI content capacity; file/PG process capacity; file/PG acquisition crash; file/PG mutation crash; **file/PG disclosure crash**; file/PG dependency ordering; file/PG window budgets; file I/O faults; PG partial-batch rollback; PG COMMIT reply loss; file/PG mutation retained replay; file/PG information retained replay.
+
+New disclosure generators passed in 11.15 seconds (file) and 40.86 seconds (TLS PostgreSQL). Existing 30 PG gateway tests passed in 13.74 seconds; lifecycle in 16.89 seconds; operator plane in 0.71 seconds. Physical/WAL restore retained before-target state and excluded after-target state; `replay_continuity_claim=false` remains explicit.
+
+Full log and timestamped execution lines: `/tmp/ap074-ci-ce9d-pg-job.json`.
+
+## Producer identity and evidence files
+
+Linux wheel and PG checked out PR merge **`cc77cdbea1132f2b1e0e48f2a2f96aa8e8d8dfb5`**, not literal source head ce9d. Both have verified Git tree **`11274eea06e9d50ccaf61475bc877e2bda19da89`**. Preserve the actual producer identity and the source-tree equivalence. Demo producers explicitly use PR head.
+
+Final complete check/workflow matrix, producer identity, native execution and retry evidence: `/tmp/ap074-ci-ce9d-evidence.json`. This report: `/tmp/ap074-ci-ce9d-report.md`.
+
+Main/formal work finished and will not be cancelled by the next coherent publication. No source fix or child commit is pending. Remaining runner capacity, conditional live qualification and other AP074 acceptance gaps are not waived. No claim is made for unpublished parent changes. AP070 remains independent; AP073 untouched.
diff --git a/docs/plans/0074-hosted-evidence-d9618a56.md b/docs/plans/0074-hosted-evidence-d9618a56.md
new file mode 100644
index 000000000..ffb16cebc
--- /dev/null
+++ b/docs/plans/0074-hosted-evidence-d9618a56.md
@@ -0,0 +1,25 @@
+# Airtable CI-only fix — published and verified
+
+Observed terminal success on 2026-10-29 at 19:49 UTC. **Main AP074 work remains paused.**
+
+Published exact commit **`d9618a568c4db8fae0b8bc4cb80b77de81c8e7d1`** to `origin/codex/effect-recovery-foundation` for [PR213](https://github.com/auths-dev/auths-proof/pull/213). Before pushing, verified remote and PR head were `cb1ee6d15a8f728b2f6f9ce01021de69b67ac966`, that this was an ancestor of the repair, and that the entire outgoing delta was exactly one reviewed commit changing only `demos/airtable-gateway/tests/owner_browser.py` (11 insertions/two deletions). Non-force push succeeded; remote SHA reverified afterward. No unrelated A19 commits were published. Main worktree, uncommitted retention/comparator work, AP070 and AP073 were untouched.
+
+The [original failure](https://github.com/auths-dev/auths-proof/actions/runs/40911731259/job/123585414288) expected an internal recovery tool identifier in the complete native literal review. The native renderer instead identifies a recovery acquisition using the versioned `auths.gateway-recovery-read/1` contract. The repair checks that rendered contract, acquisition/GET, exact source effect and operation, correct bound read operation, matching native/semantic recovery and a separate reader actor. Existing literal byte equality, authentication, consent/no-entry, substitution, provider-count and replay assertions remain intact. No application/native behavior or security deadline changed.
+
+Existing complete local hosted-package Docker/browser reproduction was rechecked rather than repeated. Publication now has fresh authoritative hosted verification:
+
+- **[Airtable run40942189355](https://github.com/auths-dev/auths-proof/actions/runs/40942189355), attempt1: completed SUCCESS**, source head exactly `d9618a568c4db8fae0b8bc4cb80b77de81c8e7d1`.
+- [Package job123688155935](https://github.com/auths-dev/auths-proof/actions/runs/40942189355/job/123688155935): passed.
+- [Installed-demo job123691155527](https://github.com/auths-dev/auths-proof/actions/runs/40942189355/job/123691155527): passed, including all 36 existing cases and the repaired complete installed owner-browser journey.
+- Evidence artifact **12928066979**, SHA256 **`f61da4e93df77739da78960f20d4baa994583b5126ddb1a2d820328e80c1d147`**, downloaded and checksum verified. Its `owner-journey.json` explicitly identifies exact source `d9618a56` and records unchanged native renderer, withheld pre-consent approvals, decline/no entry, substituted review refusal, private owner credential isolation, one write, independent read and browser/fresh-process replay without extra execution.
+- Package artifact **12927612249**, SHA256 **`b240c4a5e3e0a1b37549fc8f971b6112ae9abea84b246688e7c967a39ba15f95`**, attributed to the same head.
+
+This completes the requested CI-only fix. No additional changes or reruns were needed. Other automatically triggered workflows were preserved; no claim is made here that the entire new-head PR matrix is terminal or green. Owner actions are browser-simulated against a local provider; the artifact explicitly excludes human attention, production custody, live provider qualification, hostile OS and a full content/disclosure journey.
+
+Saved records:
+
+- `/tmp/ap074-ci-d961-evidence.json`: job records, artifact attribution and installed-job log.
+- `/tmp/ap074-ci-d961-owner-journey.json`: verified hosted owner-journey manifest.
+- `/tmp/ap074-ci-d961-airtable-evidence.zip`: checksum-verified hosted evidence archive.
+
+Child checkout is clean. Main's previously cherry-picked duplicate `f6ef54a1` and unpublished work remain preserved; any future ancestry reconciliation belongs to a later authorized resumption, not this paused goal.
diff --git a/docs/plans/0074-hosted-evidence-eb44d8f7.md b/docs/plans/0074-hosted-evidence-eb44d8f7.md
new file mode 100644
index 000000000..462a7293e
--- /dev/null
+++ b/docs/plans/0074-hosted-evidence-eb44d8f7.md
@@ -0,0 +1,128 @@
+# AP074 hosted evidence: eb44d8f7
+
+Status at publication: the eb44 native, installed and clean formal producers below are verified. Its implementation gate retains the obsolete-corpus failure; the repaired consumer head must pass its own complete gates and both real Lean reuse checks. This is producer evidence, not final acceptance.
+
+Source `eb44d8f7082db6ed9ab46c1bf5319e8e736d7b28`; GitHub PR merge producer `8335af6a6f9170a12f4b63feae0f755e87290308`; both tree `195e1ae19c0491af1c7e8fdd6d0c63b87655fe89`. Merge parents are protected base `e2d10a74fa0be5e6bfe2b9f698632c835cd7b11b` and source eb44. Native closure `61173a0020911d7fb40fe9870a81b9b19a3641b1f5c406c0eeb9b9db3a3b88aa`; formal source closure `fe66a1c24450b555e7e28d775c5796b68bc0d896445181343c6f8178d3f43751`; freeze 432.
+
+All twelve applicable workflows outside CI passed. Conditional custody-live41924978034 is skipped and provides no live-provider evidence. Main CI41924977969 is terminal with its known corpus failure. The complete cohort has 96 checks: 85 success, 3 failure and 8 conditional skips. Fast Lean126908887765 passed at22:49:58UTC after the fresh build beginning22:34:06UTC. Kani126913912984 passed at 23:01:31UTC with all49harnesses; translation126913912807 passed at 23:10:35UTC after two byte-identical reproductions. Authoritative clean Lean126929527493 ran23:10:39–23:38:21UTC and passed; its clean rebuild step started23:14:08UTC. No manual reruns/cancellations or new publication occurred after eb44.
+
+## Workflow inventory
+
+| Workflow | Run | Current result |
+| --- | --- | --- |
+| Airtable gateway demo | [41924977955](https://github.com/auths-dev/auths-proof/actions/runs/41924977955) | success |
+| CI | [41924977969](https://github.com/auths-dev/auths-proof/actions/runs/41924977969) | failure: corpus root and dependent rollups |
+| Gateway custody live | [41924978034](https://github.com/auths-dev/auths-proof/actions/runs/41924978034) | skipped |
+| Gateway isolation | [41924977909](https://github.com/auths-dev/auths-proof/actions/runs/41924977909) | success |
+| Gateway operator package | [41924977942](https://github.com/auths-dev/auths-proof/actions/runs/41924977942) | success |
+| Git signing branch-protection action | [41924978062](https://github.com/auths-dev/auths-proof/actions/runs/41924978062) | success |
+| Git signing protocol | [41924978149](https://github.com/auths-dev/auths-proof/actions/runs/41924978149) | success |
+| Google read demo installed images | [41924978033](https://github.com/auths-dev/auths-proof/actions/runs/41924978033) | success |
+| PostgreSQL lifecycle qualification | [41924978051](https://github.com/auths-dev/auths-proof/actions/runs/41924978051) | success |
+| Python SDK package | [41924977978](https://github.com/auths-dev/auths-proof/actions/runs/41924977978) | success |
+| Recipe qualification | [41924978037](https://github.com/auths-dev/auths-proof/actions/runs/41924978037) | success |
+| SDK outcome recipes | [41924978031](https://github.com/auths-dev/auths-proof/actions/runs/41924978031) | success |
+| Stripe gateway demo | [41924978136](https://github.com/auths-dev/auths-proof/actions/runs/41924978136) | success |
+| TypeScript SDK package | [41924978104](https://github.com/auths-dev/auths-proof/actions/runs/41924978104) | success |
+
+## Conditional skips and scope
+
+All eight skipped jobs require `github.event_name == workflow_dispatch` and `github.ref == refs/heads/main` in their checked-in workflows. This cohort is a PR event on `codex/effect-recovery-foundation`, so neither condition is met. These are explicit inapplicable live/operator release paths for this PR event; their skipped status supplies no passing, provider, retained-operator or signed-release evidence. The installed simulations and required native/platform gates are separately executed above.
+
+| Skipped job | Workflow and excluded operation |
+| --- | --- |
+| [verify as a gateway would (126912127149)](https://github.com/auths-dev/auths-proof/actions/runs/41924978037/job/126912127149) | recipe-qualification: verify a manually signed live qualification release |
+| [sign the complete cleaned qualification release (126912126680)](https://github.com/auths-dev/auths-proof/actions/runs/41924978037/job/126912126680) | recipe-qualification: sign the cleaned live release |
+| [independently review finite commissioning (${{ matrix.family }}) (126912125017)](https://github.com/auths-dev/auths-proof/actions/runs/41924978037/job/126912125017) | recipe-qualification: independent finite commissioning review |
+| [independently sign first qualification (${{ matrix.family }}) (126912124758)](https://github.com/auths-dev/auths-proof/actions/runs/41924978037/job/126912124758) | recipe-qualification: independent first-qualification signing |
+| [retained production operator (${{ matrix.family }}) (126912124717)](https://github.com/auths-dev/auths-proof/actions/runs/41924978037/job/126912124717) | recipe-qualification: retained production operator |
+| [gateway journey against Stripe test mode (126906932637)](https://github.com/auths-dev/auths-proof/actions/runs/41924978034/job/126906932637) | gateway-custody-live: actual Stripe test-mode provider journey |
+| [store-contract (126906880394)](https://github.com/auths-dev/auths-proof/actions/runs/41924978034/job/126906880394) | gateway-custody-live: manually dispatched production credential-store contract |
+| [gateway journey on the production credential store (126906879308)](https://github.com/auths-dev/auths-proof/actions/runs/41924978034/job/126906879308) | gateway-custody-live: production credential-store journey |
+
+## Actual native and installed evidence
+
+[PostgreSQL lifecycle job126906875975](https://github.com/auths-dev/auths-proof/actions/runs/41924978051/job/126906875975) passed. Its actual complete retained log matches all 63 required selectors exactly once with one passed/zero failed/zero ignored:48 ordinary,11 final-hardening,2 cost and2 generated-history. Both newly required native review and64-event boundary selectors executed. The additional production connection-process test and 30-store hostile/conformance suite passed separately. Physical-with-WAL restore to a named restore point reports before-target retained and after-target excluded; replay-continuity claim is explicitly false. Connection-generation and exact-record floors protect only their stated state: an unchanged connection floor does not establish safe automatic managed resume after point-in-time rollback of attempt/counter rows. The physical-WAL rehearsal proves before/after storage behavior, not whole-journal replay/accounting continuity.
+
+The independent mutation report compares28 requests and26 responses across maintained Airtable/Stripe recipes. All seven source hashes match this checkout. The information report compares22 pairs after actual execution of the eight named file/PostgreSQL histories; its five source hashes and native-log hash match. Its source_commit is the verified merge producer, whereas installed/cost/history reports explicitly retain source and/or checkout separately. None is relabeled as a different producer revision.
+
+Generated histories retain four seeds and eight concurrent rounds per seed per adapter (32 positive rounds and four negative pairs). Independent expected effect entries agree with actual HTTP and write-transport counts. Distinct OS workers and retained historical-byte preservation hold in every round; each seed has four distinct effects plus the baseline HTTP entry, with fresh-operation/old-approval refusal leaving the source unchanged. These are deterministic generated fixtures, not exhaustive concurrency exploration.
+
+The actual Linux distinct-UID report covers33 application cases,12 admin commands,53 CLI routes and5 private CLI refusals. Initial 45/replay 49 private files remain inaccessible. It reports no new custody/provider entries, direct provider refusals, recipient identity/key checks before content, trusted recipient positive control, no plaintext/encoding/raw-digest sentinel leakage,0600 private export, one verified audit entry and overwrite refusal.
+
+Airtable source-free installed simulation passes 36 cases with 13 verified private audit entries, zero inconsistent/unverified. All owner-browser positive assertions hold: consent withholding/decline/substitution, exact native renderer, private owner credentials, one write and separate read, Python+TypeScript acquisition, separate disclosure consent/native recipient ACK, private content, no replay delivery, same actor across operations and unknown-response/fresh-process controls. Stripe passes 23 cases with 6 verified audit entries, zero inconsistent/unverified. Both explicitly decline production/live-provider/human-attention qualification.
+
+## Verified artifact archives
+
+All ZIP SHA-256 values below match GitHub's authenticated artifact metadata; run metadata binds the exact source head. Only bounded named reports/logs were read. Private nested audit packets were not copied into this report.
+
+| Artifact | GitHub ID | ZIP SHA-256 |
+| --- | --- | --- |
+| Linux distinct-UID | 13369196614 | `b58fb2bf2087540b0cded52e54ac07a3ea1c6d658a0a66bb26e6643ab6426f67` |
+| Mutation differential | 13368214676 | `2f12aad26dcbebd9dd3122159aea9f67869dd536dbd5fd4e53d65f0df81a93cd` |
+| Final hardening | 13370235387 | `1901e57fe4081450b9398a28622fdc5634784d00d5fa4948e8c7607e1b8e520f` |
+| Generated histories | 13369825985 | `4c44f92f269c0988d8cb9e7afe9f32da23d0130ae7be2ea20ccacbbef7061963` |
+| Native cost | 13369072499 | `668af902c3b2ebeecff308eb2933d60f1c679af2bfb2b49e658958c6065affcd` |
+| Information differential | 13368772949 | `18e9449e053625c7c019d1c5b08379774e3435297e98986e7fe87cbe05211e8b` |
+| Airtable installed | 13368319829 | `b8d0d631a952d3647b8ee338db6b5df0e603941b182e925604f9d09ad310acbd` |
+| Stripe installed | 13369173481 | `60aeed41d3b8054e83d2f69111fdfce3beee70f26d55db570258ddc64ab77bfa` |
+| Fast Lean log | 13370357973 | `d20e5620c8de7d5d36513e8869b27e029330749ff670a79942f2771e3dbe2b0f` |
+| Kani phase | 13370384762 | `5f9b7dea6ddf6313592d36da7dff15700fb6ca792f4d3408130ae77c197bb963` |
+| Translation phase | 13371652840 | `68184973fcad190b9fc0c289a8960665f38b7bda87b7ae6ef0317a0f5488b5fa` |
+| Translation qualification | 13370803802 | `bcb31f244dbfa513eedd33ea06dd0be1c10176c4ce64214a2aadb4b54760e537` |
+| Clean authoritative Lean phase | 13373617020 | `f2a256d1be842c86af7cf057b0e46e989e2029f5a4504ca18b2ecf093fd6818c` |
+| Final formal evidence | 13373479017 | `7ac7d07e3b21228650318811c6d54bf09dbc6febbd4e4e793f806b567f52fb8a` |
+
+## Native cost measurements
+
+Eight test-profile samples per path/backend; timings are min/median/max milliseconds, not production SLOs. Adapter and fenced callback totals sum compilation plus execution calls per sample. Snapshot and RSS columns are maxima.
+
+| Backend/path | Verify ms | Compile incl verify ms | Approval ms | Fenced callback ms | Adapter ms | Max snapshot bytes | Max RSS KiB |
+| --- | --- | --- | --- | --- | --- | --- | --- |
+| file/mutation | 0.888/0.946/0.979 | 3.040/3.116/3.142 | 0.238/0.239/0.325 | 11.359/11.660/16.188 | 14.362/14.800/20.507 | 4538 | 34092 |
+| file/recovery | 0.870/0.884/0.912 | 8.627/8.885/8.975 | 0.222/0.224/0.251 | 37.588/38.156/40.827 | 42.149/42.811/54.645 | 12032 | 34136 |
+| file/acquisition | 0.937/0.958/1.518 | 4.627/4.699/7.423 | 0.232/0.245/0.260 | 18.533/18.745/18.960 | 22.289/22.587/22.934 | 7330 | 35516 |
+| file/disclosure | 0.889/0.920/0.945 | 4.114/4.189/4.255 | 0.239/0.242/0.263 | 37.878/38.435/38.691 | 41.773/42.379/52.188 | 16717 | 35580 |
+| postgres/mutation | 0.916/0.955/1.179 | 4.250/4.337/5.506 | 0.237/0.241/0.270 | 11.451/11.750/13.926 | 24.247/25.036/34.465 | 4538 | 38940 |
+| postgres/recovery | 0.874/0.922/0.960 | 10.415/10.585/11.005 | 0.223/0.227/0.239 | 38.249/38.478/40.621 | 63.532/64.258/66.230 | 12032 | 39000 |
+| postgres/acquisition | 0.902/0.940/1.665 | 5.699/5.804/6.515 | 0.231/0.233/0.261 | 18.692/18.941/19.462 | 34.708/40.561/50.609 | 7330 | 41756 |
+| postgres/disclosure | 0.903/0.928/0.959 | 5.222/5.313/5.754 | 0.238/0.240/0.253 | 38.566/39.142/45.296 | 54.699/55.373/64.891 | 16692 | 41816 |
+
+file mutation_recovery: 8 measured inventories; native decode+inventory ms 16.346/51.331/88.234; max attempts 17, limit 4096; max RSS 34136 KiB.
+
+file acquisition_disclosure: 8 measured inventories; native decode+inventory ms 17.950/79.608/142.396; max attempts 16, limit 4096; max RSS 35580 KiB.
+
+postgres mutation_recovery: 8 measured inventories; native decode+inventory ms 17.703/63.310/99.874; max attempts 17, limit 4096; max RSS 39000 KiB.
+
+postgres acquisition_disclosure: 8 measured inventories; native decode+inventory ms 20.013/86.463/155.022; max attempts 16, limit 4096; max RSS 41816 KiB.
+
+
+Raw artifacts bind source eb44, merge8335, native closure61173 and the current test profile. They report actual native verification, compile, approval, adapter/callback timings, buffer lengths and Linux RSS; these eight-sample fixtures are not production SLOs or adversarial throughput guarantees.
+
+## Assurance citation audit
+
+The parent independently compared977→eb44 assurance manifest text. Masking only the 59 existing `kind=source-closure` citations to `formal/qualification/aeneas/source-closure.json` makes the remaining text byte-identical. All 59 current hashes equal raw source-closure file SHA-256 `685fdbfd0d91ae329300d15dbd2d3469b101a358137c8e37dbed35e3559d4601`. No theorem, axiom, claim, review or semantic binding changed. This agrees with the local complete typed-manifest negative controls; metadata refresh does not constitute proof qualification.
+
+## Verified clean Lean producer
+
+[Authoritative Lean job 126929527493](https://github.com/auths-dev/auths-proof/actions/runs/41924977969/job/126929527493) completed successfully at 23:38:21UTC. Its actual log records the compiled audit of329 statements with transitive axioms reviewed, Aeneas qualification with four reviewed external links and zero compiled external axioms, and four passing Rust refinement tests. The clean rebuild executed; no cached phase was accepted.
+
+Artifact 13373617020 has exactly one member, `lean.json`. Authenticated archive SHA-256 is `f2a256d1be842c86af7cf057b0e46e989e2029f5a4504ca18b2ecf093fd6818c`. Its immutable source is eb44, run 41924977969, attempt 1; `execution=executed-clean`, `reproduction_count=0`, and no `reused_from`. Complete proof inputs are `c523f465540e306ac9eaaf49ef460365b467be4bccaafd2160552ce23dc83432`, exactly matching the repaired native computation. Planned Lean closure is `9a9be86efddb2a9db4ee16ed2c25233dce5d56437c7e4395c93f347908472ddc`; toolchain closure is `baff1cabbf9c83ffca0f4bdb1b8dda685a5ac59638e95d6d12e57f5d5bda9633`; evidence closure is `1c446639f484deacf0dc54fbd94da4ec6aceaa69e8e50432ae05696f123c0dbd`. The native reuse validator must independently validate these fields and current source/refinement checks at the consumer; this successful producer is not itself a warm hit.
+
+## Remaining qualification
+
+Main CI authoritative workspace has the corpus failure documented below. Final formal evidence 126942117729 passed at 23:46:11UTC; artifact 13373479017 attests execution=aggregated for the exact source/run/attempt and matching closures. Clean authoritative Lean and its compiled 329-claim audit passed. Qualified rollup126945531900 failed because AUTHORITATIVE=failure from the corpus root; its other listed formal, compliance, dependency, secrets and live-adapter dependencies are success. The three failures are implementation126913912856 and its derivative authoritative126917977445/qualified126945531900 rollups, not three independent roots. Fast Lean, both translation reproductions and all 49 Kani harnesses have passed. The authenticated translation phase reports execution=executed/reproduction_count=2; its qualification record retains zero compiled external axioms and four reviewed external models with all seven case modules. With the successful cold producer preserved, the required corpus repair/final evidence update must demonstrate both Lean phases reuse matching verified evidence while current-head Rust/source checks still execute. No warm hit is claimed yet. AP074 is incomplete pending final acceptance; AP070 stays independent and AP073 untouched.
+
+## Actual authoritative corpus root and repair
+
+Authoritative implementation126913912856/rollup126917977445 failed after the gateway library passed 302 tests, zero failed and 63 ignored in 38.35s. Both failures are in `auths-gateway-fuzz --test corpus`: the committed `observed-by-provider-after-response` seed predates public outcome/3, and `observe-pre-entry` still uses observe/3 instead of/4. Dynamic xtask campaign seeds already use current native fixtures/schemas. The repair binds committed positive/refused outcomes to the existing native canonical signer, keeps old bytes as explicit obsolete refusal controls, and binds current observation frames to the existing native frame constructor. No decoder or assertion is relaxed. No xtask/formal/controller/Cargo input changes are needed; the healthy eb44 proof producer is preserved. Focused complete repair verification passed: both authoritative native seed constructors and unchanged checks, all six fuzz corpus tests, native source-closure update/rebuilt check, formatting and semantic metadata construction. The canonical signed outcome corpus now has 14 current positives,8 current refusals and2 preserved obsolete seeds; current observation frames coexist with the preserved obsolete/3 refusal. No release identity or formal source/assurance file changed. Corrected-head hosted qualification remains pending.
+
+Repair results: `/tmp/ap074-ci-eb44-repair/results.json`. No broad workspace, Lean or Kani suite was rerun locally. The successful hosted clean proof record is eligible for independently validated reuse because the repair changes only native gateway test/corpus/source-inventory files.
+
+The repaired native gateway closure is `4c368d74e5a748ac79752bf9f2da2cbb3447c9ec67755147c80c03a0856a3d81`. The actual native `auths_ci_plan::formal_proof_inputs_digest` function was evaluated read-only against published main eb44 and the repaired isolated tree: both return `c523f465540e306ac9eaaf49ef460365b467be4bccaafd2160552ce23dc83432`. This establishes unchanged complete proof inputs for the seed repair; it is not a synthesized or qualified phase result. The future hosted warm path must independently repeat that validation and all current-head checks.
+
+The two 63 counts name different inventories: 32 library-ignored cases are in the 63 explicit native selectors; another 30 ignored cases passed in the PostgreSQL conformance/hostile suite. The remaining ignored `scenario_tests::conformance_child_process` is deliberately spawned by both file/PostgreSQL conformance parents, twice for submit and twice for batch. Those parents require successful child exits and exact `RACE` witnesses. Their passing hosted results establish the child executions through this harness; there is no separately invented standalone helper run.
+
+## Consumer verification boundary
+
+The corrected consumer head must pass its own required implementation, installed SDK/demo, native adapter, formal aggregation and two validated Lean reuse gates. Those consumer results are pending at publication of this report. The final live verification record is the exact published head shown in [PR213 checks](https://github.com/auths-dev/auths-proof/pull/213/checks), with its immutable run/job/artifact identities recorded in the PR validation notes and retained authenticated final-head report. Passing producer evidence above is not relabeled as consumer execution. Recording that consumer outcome does not require a recursive documentation-only push.
diff --git a/docs/plans/0074-implementation-handoff.md b/docs/plans/0074-implementation-handoff.md
new file mode 100644
index 000000000..c4097b504
--- /dev/null
+++ b/docs/plans/0074-implementation-handoff.md
@@ -0,0 +1,107 @@
+# AP-SPEC-074 implementation handoff
+
+This prompt is for an agent continuing the existing AP074 worktree. It is not
+an instruction to resume paused AP073 work. Implementation is in progress;
+`0074-implementation-status.md` is the authoritative current evidence ledger.
+Initial worktree-creation facts below are retained as historical context.
+
+## Copy/paste prompt
+
+```text
+Make it your goal to implement AP-SPEC-074 completely:
+
+/Users/bordumb/workspace/repositories/auths-proof-base/auths-proof-effect-recovery/docs/specs/0074-bound-effects-and-recovery-state-machine.md
+
+Work only in:
+/Users/bordumb/workspace/repositories/auths-proof-base/auths-proof-effect-recovery
+
+Branch: codex/effect-recovery-foundation
+Baseline: origin/main at e2d10a74fa0be5e6bfe2b9f698632c835cd7b11b
+
+Start by reading AGENTS.md, the required profile/domain abstraction boundary plan,
+and the full AP-SPEC-074. Inspect the worktree and preserve any existing changes.
+Write your own durable goal prompt covering the complete specification, then
+implement it. Keep progress and the acceptance/evidence ledger in
+docs/plans/0074-implementation-status.md.
+
+The goal has exactly two architectural pillars:
+1. One immutable bound effect model shared by native authorization, deterministic
+ request compilation, human approval, admission, execution and reporting.
+2. One authoritative recovery transition model and transaction contract for
+ reservation, provider entry, disclosure, acknowledgement and unknown outcomes.
+
+Finish this foundation before expanding multi-agent delegation. Do not implement
+participant trees, federation, broad identity lifecycle redesign or a new token
+system. Do not merge/cherry-pick the AP073 branch wholesale. The separate
+codex/identity-authority-foundation worktree is paused and contains uncommitted
+work; do not modify, reset, clean or commit anything there. AP073 is not a
+prerequisite or part of this goal.
+
+Main already has the gateway's native verifier integration, closed recipe
+compiler, submission step machine, attempt transitions, durable stores, lifecycle
+reference contracts and deployment clock. Read these first and preserve sound
+mechanisms. Do not create a third lifecycle framework. Complete P0's owner mapping,
+ADR and abstraction case, then perform bounded direct cutovers through P1–P4.
+Centralize shared semantic decisions; do not scatter independent clock, hash,
+approval, retry or accounting fixes through routes, demos and bindings.
+
+Scope clarification: preserve AP070 as an independently implemented,
+application-owned Google broker/reference with its existing narrower claims.
+Do not delete, rewrite or migrate its Python/SQLite admission, execution or
+receipts as part of AP074. Migration, Google metadata and OAuth refresh are
+optional future productization, not AP074 exit gates. Compare retained reference
+behavior without making it consume the implementation it tests. Every managed
+gateway operation must still have one native authority/journal; never give the
+reference managed credentials/permits, use it as an alternate managed route, or
+interpret its receipts as managed guarantees. Preserve all native acquisition,
+disclosure, installed binding/UI, file/PostgreSQL, fault/concurrency, containment,
+formal and A01–A22 requirements. Prioritize those core gates and CI fixes.
+
+The shared model must retain typed differences between acquisition, disclosure
+and mutation. A provider response, receipt signature, state match and recipient
+ACK establish different facts. An unknown outcome must never authorize a blind
+retry, redelivery or refund. Entry permits are issued once, only after the required
+durable acknowledgement; persisted facts cannot recreate lost permits. Fresh
+native authority, approval, configuration and revocation checks must be ordered
+with the final entry decision. Bindings expose Rust semantics rather than
+reimplementing them.
+
+Produce actual behavior, fault-injection, concurrency, differential, installed
+binding and formal/refinement evidence for the specification's invariants and
+A01–A22. The first complete journey is one human and one agent: bounded acquisition,
+controlled disclosure, exact approved write, independent observation and truthful
+recovery. Use real local transport/process boundaries and maintained durable
+adapters. Do not claim live-provider or isolation guarantees from a unit test.
+Record external assumptions and unsupported rollback guarantees precisely.
+
+Commit coherent implementation changes along the way using Conventional Commits.
+Follow the repository's GitHub-first verification policy. Generate required
+artifacts locally as needed to construct the change; do not duplicate broad local
+build/test/lint/formal suites without the policy's justification. The user has
+explicitly authorized early isolated CI triage; fix actual failures without
+weakening contracts. Final hosted verification still requires the completed
+implementation, fixtures and documentation on its exact revision. Push
+explicitly to origin codex/effect-recovery-foundation (never to main), open/update
+a draft PR against main, observe all required hosted workflows on the exact head,
+repair failures and repeat. Do not merge or deploy.
+
+Do not mark the goal complete for a spec, scaffold, wrapper, happy path, passing
+compilation or isolated test. Finish the source cutover and required evidence.
+If genuinely blocked by an external dependency, preserve the work and report the
+specific incomplete criterion, evidence and smallest required intervention.
+Do not weaken checks or silently narrow the scope to manufacture completion.
+
+At completion, report the PR, exact revision, tests/CI results, central owners,
+removed duplicate paths, and remaining deployment assumptions. Do not resume
+multi-agent delegation as an automatic follow-on.
+```
+
+## Initial handoff facts (historical)
+
+- This worktree was created from freshly fetched `origin/main`, independently of
+ draft PR #212 and its uncommitted changes.
+- Only AP074 specification and handoff documentation are authored here.
+- No implementation, test execution, push, CI observation, PR creation or agent
+ dispatch is part of this documentation handoff.
+- The existing AP073 goal was explicitly marked **paused** at the user's request.
+- The specification distinguishes required future evidence from existing code.
diff --git a/docs/plans/0074-implementation-status.md b/docs/plans/0074-implementation-status.md
new file mode 100644
index 000000000..068b4a9ba
--- /dev/null
+++ b/docs/plans/0074-implementation-status.md
@@ -0,0 +1,4182 @@
+# AP-SPEC-074 implementation and evidence ledger
+
+## Resumed — 5 November 2026
+
+The user reactivated the complete AP074 goal. The signed-output/private-audit checkpoint is
+reconciled and published as `7807d04b`. Combined revision `46989b7d` also includes
+native cost producers, strict 209-vector result parity and the holistic CI repair
+batch. Main is adding generated concurrent histories and auditing final closure. The original full objective and
+acceptance requirements remain in force; no completion claim is made.
+
+The CI subagent reports the published `3cb2caad` cohort is terminal: 88 passed,
+zero failed and eight conditional skips. It verified 48 actual native selectors,
+installed owner/UID evidence, 49 Kani harnesses, 329 Lean statements, source
+closures and final evidence rollups. Main did not inspect CI. Its report is
+`/tmp/ap074-ci-3cb2-report.md`; these results do not qualify subsequent changes.
+
+## User-requested pause — 30 October 2026
+
+AP074 remains incomplete. The user requested a committed, pushed pause.
+The completed public mutation-evidence cutover is `1769057a` on PR213.
+All unfinished signed-output/private-audit work is separately committed and pushed
+at `f55aa61a720a5885d68da30daf102689f75446a3` on
+`codex/ap074-public-output-checkpoint`. Its
+[`0074-public-output-checkpoint.md`](https://github.com/auths-dev/auths-proof/blob/f55aa61a720a5885d68da30daf102689f75446a3/docs/plans/0074-public-output-checkpoint.md)
+records implemented changes, construction evidence and remaining integration work.
+That checkpoint is unfinished and must not be mistaken for a qualified PR head.
+
+Resume only on a new user request. Preserve both branches and reconcile the
+checkpoint with subsequent PR CI repairs before continuing. A17, the remaining
+acceptance/evidence items below and final exact-head qualification remain open.
+AP070 remains independent; AP073 remains paused and untouched.
+
+## Durable goal
+
+Implement all of [AP-SPEC-074](../specs/0074-bound-effects-and-recovery-state-machine.md)
+in this checkout, on `codex/effect-recovery-foundation`. Finish the immutable
+native bound effect and the gateway's sole recovery/transaction owner, including
+mutation, bounded acquisition, separate recipient disclosure, exact approval,
+fresh entry authorization, immutable evidence and truthful native projections.
+Cut maintained managed consumers over directly; remove competing decisions for
+the same managed operation. Retain AP070 as an independent application-owned
+reference with narrower claims; its migration is deferred, not an AP074 gate.
+Preserve the native verifier, reviewed recipe compiler, conservative recovery,
+durable adapter guarantees and source-linked formal claims.
+
+Complete P0–P4 and A01–A22 with actual transport, process restart, file and
+PostgreSQL concurrency/fault evidence, differential recipes, installed bindings,
+formal/refinement links and bounded performance measurements. Commit coherent
+changes. The user's subsequent instruction on 28 October overrides the initial
+push timing: commit the current slice, push explicitly to
+`origin codex/effect-recovery-foundation`, open a new draft PR against main, then
+continue with incremental commits. The subsequent explicit CI-triage request
+authorizes investigating failures before implementation is complete. Final
+required hosted workflows must still pass on the completed head. Repair failures; do not
+merge or deploy. Local generation needed to construct artifacts is permitted;
+do not duplicate hosted build/test/lint/formal gates before the first push.
+
+Never touch the paused AP073 checkout or import its unfinished implementation.
+Do not expand participant trees, federation or token systems. Only the explicitly
+requested isolated CI subagent is authorized; this is not delegation product work.
+No stage is complete merely because types, documentation or a happy path exist.
+
+## Baseline and working record
+
+- Implementation started 28 October 2026 from clean HEAD
+ `ec7defb3dbcee1fdcc2839989bdbc4d9110a912d` (specification/handoff commits over
+ the specified main baseline `e2d10a74fa0be5e6bfe2b9f698632c835cd7b11b`).
+- Read repository contract, mandatory abstraction plan and full specification.
+- No AP073 files inspected or changed. No subagents dispatched.
+- Existing source inspected: native sealed action, gateway verifier/compiler,
+ submission driver, transition/recovery leaves, file/PostgreSQL attempt store,
+ lifecycle preparation/acknowledgement contracts, AP070 admission/executor/receipt.
+- P0 committed as `43832adc`.
+- P1 source: canonical private binding, native evidence-selection commitment,
+ effect-free mutation compilation and separate exact Ed25519 human approval.
+ Native signature validation does not replace resource authority or live entry.
+- Source tests authored for identity mutations, canonical/duplicate/limit cases,
+ approval substitution/expiry and a real native signed-proof compilation path.
+- Draft PR: https://github.com/auths-dev/auths-proof/pull/213, initially pushed at
+ `005dfd42`. GitHub connector PR creation lacked permission; authenticated `gh`
+ created the draft. No workflow results have been inspected.
+- The targeted native fixture-generation test completed with one passing test
+ after correcting an invalid empty precondition declaration in the test recipe.
+ Command: `AUTHS_UPDATE_FIXTURES=1 cargo test -p auths-gateway --lib -- --exact observed_tests::effect_native_tests::native_compilation_approval_and_reverification_keep_exact_request`.
+ Artifact: `bindings/fixtures/gateway/effect-binding.json`. This run used
+ `005dfd42` plus the committed null-precondition fixture correction; its scope
+ is compilation, canonical round-trip and exact approval, not A01 completion.
+ This local run constructs the required canonical artifact; it is not a broad
+ pre-push verification run. No hosted workflows observed or formal gates run.
+- Initially the new effect module was not connected to submission. The installed
+ approval cutover below now replaces that proof/action-only interface. P1 is
+ still open for review completeness, full binding persistence and integration
+ evidence; the existing reservation/entry lifecycle has not been replaced yet.
+- Added entry validation over the retained original proof/action. It re-runs the
+ full verifier, compares the complete newly compiled binding, and checks the
+ original human approval at the supplied transaction time. The installed exact
+ approver policy is itself in the effect identity. The same native fixture
+ generation test passes after this change, including independently expiring
+ resource authority while human approval remains current. Transactional wiring
+ and fencing are still pending; this API alone makes no linearization claim.
+- Native private review now renders the sealed command through the same
+ projection as offline recipe review. It carries the exact effect identity and
+ canonical binding; no binding/UI request reconstruction is needed. Review
+ data explicitly contains private content and has a separate 128 KiB limit.
+- Spec hardening committed and pushed as `c75e16ba`: deployment containment,
+ review completeness, concrete fencing requirements, compositional properties,
+ operational recovery and the future delegation boundary.
+- Required operator `EffectPolicy` now pins partition, epoch and approver. The
+ engine constructor validates it; installation `/6` and operator attestation
+ `/2` require and authenticate it. Native `effect-policy` emits canonical
+ configuration; `install` and `operator-request` require `--effect-policy`.
+ `GatewayEngine::compile_effect` reads the installed connection and deployment
+ clock without provider I/O, custody or reservation. Entry fencing and replacing
+ the baseline submission path remain pending.
+- Native fixture generation passed again while producing `effect-policy.json`,
+ including the private-review assertions previously authored but unexecuted.
+ Policy refusal and operator-attestation substitution cases are authored, not
+ yet executed. No hosted workflows have been observed.
+- Maintained install callers now pass explicit policy input. Synthetic operator
+ rehearsal uses the native synthetic fixture; live rehearsal and example
+ journeys require operator-selected policy files. The production reference
+ journey requires `AUTHS_QUALIFICATION_EFFECT_POLICY` before infrastructure or
+ provider work. Live approval-channel integration remains pending; no synthetic
+ approver is installed implicitly into a live journey.
+- The installed mutation and commissioning APIs now require native exact effect
+ approval. The shared driver owns `ApprovedEffect`, accesses its single sealed
+ command, and revalidates original proof/approval against prepared configuration
+ before claim and before send. Its clock now uses the existing deployment clock.
+ These checks are not a transaction fence. Full binding is not yet persisted in
+ the attempt journal; ancillary reads and replay observation still need their
+ separately authorized acquisition/recovery cutover.
+- Application socket `/2` requires bounded canonical approval bytes; `/1` and
+ missing approvals are rejected. Python/TypeScript wrappers carry those native
+ bytes without hashing or signing. Private `effect-review` and `effect-approval`
+ CLI commands expose native review/signing data and canonical envelope assembly.
+ Source tests cover obsolete/missing/oversized frames and installed wrong-signer
+ or expired approval with zero custody/transport calls; these tests are authored,
+ not executed. Test-only post-verification driver oracles use explicitly synthetic
+ approvals and establish no human or installed-authority claim.
+- The targeted native generator passed while constructing the `/2` frame and
+ `effect-approval-request.json`. Running the native CLI to construct
+ `effect-approval.json` succeeded and its bytes equal the native signed corpus.
+ This artifact construction also compiled the gateway binary. No broad local
+ suite or hosted workflow was run/observed. Installed binding journeys, example
+ proof preparation/review, qualification consumers and semantic closure/formal
+ regeneration still require updates for the new approval round trip.
+
+## Immediate implementation continuation
+
+The checkpoint sections below are historical records; the phase and acceptance
+ledgers at the end describe current closure. Do not treat earlier "pending"
+notes as a reason to reimplement a later completed source cutover.
+
+1. Resolve hosted CI failures using the real native approval/recovery contracts.
+ Preserve tests and scope claims; no fake approvals or weakened gates.
+2. Close file/PostgreSQL independent-process capacity, dependency-race, crash and
+ ambiguous-ack evidence. Prove the final fencing order and managed deployment's
+ non-bypassability with real transport, counters and restricted processes.
+3. Finish native recovery actions/charge/evidence projections and the installed
+ authenticated human review/approval plus managed binding/UI journey.
+4. Complete two mutation recipe and independent read/disclosure differentials,
+ generated histories, source-linked formal/refinement and bounded performance.
+5. Pass all final required hosted checks on the completed revision. Keep AP070's
+ independently application-owned broker/reference intact. Google metadata,
+ OAuth refresh and migration are future integration, not completion blockers.
+
+## Current owner decision: AP070 remains independent
+
+The user's explicit scope decision supersedes historical checkpoint requirements
+below that called for AP070 migration. AP074 owns canonical authorization,
+admission, execution and recovery for managed gateway operations. AP070 retains
+its own Python/SQLite broker, token custody, routes and narrower claims as an
+independent reference. No migration is required for AP074 closure; no native
+acquisition/disclosure, adapter, crash/concurrency, binding/UI, containment,
+differential, formal/performance or hosted gate is waived or marked complete.
+
+Prior AP070 work is preserved in `fb5c5a0a`: `google_drive.py` now delegates pure
+metadata checks to new `google_metadata.py`, receives a typed MIME representation,
+and returns typed denials for malformed MIME/decimal values. A 36-case pure
+reference corpus/generator was added. No admission.py, read_executor.py,
+receipts.py, token vault, broker routes or SQLite ownership was deleted or
+migrated. Only the pure generator ran; no full broker or live Google claim follows.
+No further AP070 source changes are authorized by this scope update.
+
+Source/configuration observations supporting safe deferral: AP070's compose runs
+its own broker/relays and SQLite state; its agent transport calls execute_read
+and TokenVault. It does not call GatewayClient or a managed application socket;
+the gateway does not load the Google adapter. Compliance inventories separate
+its application-owned state/protocols from managed gateway state. This is not a
+runtime containment proof. Required managed evidence must still establish that
+references cannot obtain managed credentials/permits, authorize the same managed
+operation, or have their receipts accepted as native entry/recovery facts.
+Independent application-held provider credentials/external writers are outside
+managed containment and must stay explicit in deployment claims.
+
+Future managed Google productization requires its own reviewed capability,
+metadata/custody evidence and one native journal. AP070 remains useful for
+independent comparisons; document divergences rather than rewriting the oracle
+to consume the implementation it is meant to test.
+
+## Full-binding journal cutover (in progress)
+
+- Mutation admission now receives the sealed `ApprovedEffect`; the unchecked
+ request-only claim helper exists only under test/testkit compilation.
+- Attempt schema `/4` persists canonical full binding and original exact signed
+ approval in the immutable fixed fields. Historical signature validation and
+ redundant namespace/operation/action/recipe checks reject corrupt state.
+- The sole replay key includes installed namespace, partition and epoch plus
+ operation ID, and deliberately excludes the effect digest. Full binding
+ collisions produce `gateway.effect.operation-conflict`; equal bindings remain
+ replay even when approval is renewed. Neither replaces the initial bytes.
+- Every installed CLI store open receives operator policy; the engine rejects
+ a different store policy. Existing native budget counter identities remain
+ unchanged so new operation epochs do not reset grant capacity.
+- File reads now limit bytes before allocating/decoding records and rollback
+ files. This is a read bound, not yet the required bounded lock/transaction.
+- Shared file/PostgreSQL conformance source adds restart, changed selected
+ authority/recipe/action, renewed approval, partition/epoch isolation and
+ original-byte preservation checks. These new behavioral cases are authored,
+ not executed. Existing reference expectations distinguish conflict from replay.
+- Required local artifact generation passed for `cd3a3448` plus this journal
+ source: `pending_vectors::attempts::attempt_scenarios_v3_are_current` (one
+ test) and `observed_tests::effect_native_tests::native_compilation_approval_and_reverification_keep_exact_request`
+ (one test), each via `AUTHS_UPDATE_FIXTURES=1 cargo test -p auths-gateway --lib -- --exact ...`.
+ The native corpus now includes the canonical scoped operation key. Native CLI
+ envelope generation also passed after correcting its policy type import;
+ output remains byte-identical to the native approval corpus. No hosted CI, broad local suite, formal gate or actual adapter conformance run is claimed.
+- P1 still needs semantic review completeness and integration evidence. P2
+ still needs atomic fencing and a consumed permit. Identical replay's existing
+ observation path must be replaced by independently authorized bounded recovery;
+ the updated reference cases do not establish that future contract.
+
+## P2 fence preparation
+
+The concrete fence design is recorded in ADR 0017 before entry behavior changes.
+File and PostgreSQL adapters now expose one bounded snapshot-and-replace
+operation. All PostgreSQL gateway mutations acquire the same table-scoped
+advisory transaction lock, including connection updates and allocation sweeps.
+File locking uses the existing shared writer lock with bounded acquisition.
+The native decision runs once under that lock, before durable replacement; a
+callback failure, deadline or ambiguous commit returns no acknowledgement.
+The API cannot itself issue an entry permit.
+
+Read sets are bounded to 64 distinct records with a non-expiring first target.
+File reads use lock-held helpers to avoid recursive lock acquisition.
+PostgreSQL uses READ COMMITTED snapshots after acquiring the lock, retains
+sorted batch insertion and explicitly commits replacements and sweeps. Two-second
+lock/decision bounds preserve tighter installed SQL timeouts; sub-millisecond
+pool values are rejected instead of rounding to disabled timeouts.
+
+The shared conformance source now races each of replacement, insertion and
+sweeping against a held fence, checks the next snapshot sees the predecessor,
+and checks a refused decision preserves the target and is never retried.
+These tests are authored but not executed. Required semantic-closure artifact
+generation passed after fixing an error-type inference ambiguity introduced by
+the store error conversion; it compiles the source but does not exercise these
+adapter races. Final regeneration after the millisecond-bound correction also
+passed: `AUTHS_UPDATE_FIXTURES=1 cargo test -p auths-gateway --lib -- --exact semantic_closure::tests::semantic_closure_is_current` (one artifact-generation
+test), on `96e47805` plus this fence source. This is compilation/artifact evidence,
+not execution of file/PostgreSQL conformance or final-entry behavior.
+
+Remaining P2 work: shared configuration/revocation record and every writer's
+participation, Reserved/EntryGranted reducer and immutable audit event, checking
+all reservation membership and final native authority under the fence, and
+issuing/consuming a permit only from a new acknowledged transition. The old
+Attempting send path is not represented as compliant final entry.
+
+## Shared configuration checkpoint
+
+- Added the non-expiring shared configuration record and PostgreSQL schema `/7`.
+ Its installed pin covers recipe, trust, effect policy, connection/workload,
+ profile, qualification contract and compiled semantic closure. Bound-effect
+ `/2` and compiler `/2` bind that pin; obsolete binding formats are refused.
+- Trusted installation initializes the record. Explicit native operator review
+ and activation require both candidate identity and the exact prior revision.
+ Activation and qualification publication share the adapter writer fence.
+- Shared qualification memory monotonically retains sequence/time floors and
+ revoked signers/qualifications across restarts and configuration changes.
+ Bounded-set or revision saturation persists denial, never authorization with
+ forgotten entries. Hosts with a different bundle or installation fail closed.
+- Compilation and entry preparation use early shared checks. These are not the
+ final grant transaction. Reserved/EntryGranted, audit facts, reservation
+ membership checks and consumed transport permits remain outstanding.
+- File/PostgreSQL source conformance covers missing state, stale hosts, competing
+ revision tokens, restart, permanent revocation and saturation. These behavioral
+ tests are authored, not executed. Qualification test callers now publish
+ operator reloads explicitly instead of mutating only their process cache.
+- Required artifact generation passed on `4e0f435d` plus this configuration
+ source: `configuration::tests::native_configuration_fixture_is_current`,
+ `vectors::schemas::schema_vectors_are_current` and
+ `vectors::verification::verification_vectors_are_current` (one test each).
+ The native CLI regenerated `effect-approval.json` byte-identically to the
+ binding corpus and compiled the updated binary. Final source-closure generation
+ `semantic_closure::tests::semantic_closure_is_current` passed (one test).
+ These are artifact construction, not adapter race, formal or hosted CI evidence.
+
+The configuration checkpoint is committed and pushed as `8dbd9939`. A follow-up
+source audit found the PostgreSQL health projection still reported literal schema
+6; it now derives the version from the active schema constant. Required semantic
+closure regeneration passed after that correction (one artifact-generation test).
+
+## Mutation entry grant checkpoint
+
+- Attempt `/5` now creates `Reserved`; the existing transition owner alone
+ permits `Reserved -> EntryGranted` and freezes its private grant audit.
+ Post-grant state cannot become `NotEntered`. A restart projects a grant to
+ unknown while retaining reservations as reservations in the journal.
+- The existing step machine requires an acknowledged `GrantEntry` before `Send`.
+ The runtime checks native original authority, exact approval, connection,
+ configuration, qualification, time and all count/sum slots under one fence.
+ Qualification evidence is frozen once to exclude process-local reload races.
+ Commissioning checks the current permanent budget against its acknowledged
+ pre-custody floor and merges retained revocation state in the same read set.
+- Every managed write port, including HTTPS, consumes a sealed one-use permit
+ and derives the request from its retained command. An ambiguous commit yields
+ no permit. Transport failures after the grant retain unknown, even before send.
+- Native `effect-entry.json` generation passed on `3cbe3ad8` plus this source:
+ one file-backed grant, stale second-grant refusal, immutable audit and
+ crash-before-send projection. The generator also emits valid native stage and
+ transition fuzz seeds; previous-schema seeds are retained as negative inputs.
+ This synthetic post-verification corpus does not prove installed authority.
+- Updated attempt-scenario and final semantic-closure generators passed. Native
+ CLI approval-envelope generation also passed and matched the unchanged corpus. New
+ file/PostgreSQL concurrent-grant, reducer and property-test source is authored;
+ broad suites, installed races, formal gates and hosted CI remain unexecuted.
+- P2 remains open: resumable reservations with fresh authority/approval, explicit
+ revisions, full private audit/projection treatment, installed clock/revocation
+ races and all persistence/transport failpoints. Formal translated output must
+ be regenerated against the changed shipping predicates. P3/P4 and all complete
+ acceptance gates remain open. No AP073 or delegation work was performed.
+
+## Reservation restart and revision checkpoint
+
+- Attempt `/6` adds a required revision bounded to exact JSON integers. The
+ shipping transition predicate requires exactly one increment for every update,
+ rejects stale/skipped revisions and fails closed at saturation. CAS still
+ compares the complete previous bytes; the revision is not a rollback anchor.
+- An equal binding in `Reserved` can reopen without replacing its initial
+ approval, claim, first checkpoint or count/sum slots. Existing grants and
+ terminal records remain replay; changed full bindings remain conflicts.
+ Reopening produces no transport capability.
+- Grant audit `/2` freezes the approval actually used, its fresh authorization
+ start time, grant revision and fresh pre-entry evidence. The grant rechecks
+ current native authority under the fence using the resumed submission's
+ deadline. Historical admission approval remains immutable, and post-grant
+ snapshots expose the reads used for that grant.
+- Required native `effect-entry.json` generation passed on `fd8803c0` plus this
+ source: `AUTHS_UPDATE_FIXTURES=1 cargo test -p auths-gateway --lib -- --exact
+ store::entry_tests::native_entry_fixture_is_current` (one generator test).
+ The expanded synthetic file-backed sequence exhausts its count/sum capacity,
+ checkpoints, reopens after the initial approval and submission deadline expire,
+ refuses stale deadline/approval, grants once with fresh evidence, retains the
+ original slots and audit, and refuses another grant. Canonical stage/transition
+ fuzz seeds were regenerated. This is not installed proof or provider evidence.
+- Attempt-scenario and semantic-closure artifact generation also passed (one
+ generator test each). Remote artifact commit `8201545d` was preserved before
+ final closure generation. No broad suite or hosted workflow was observed.
+- Equivalent PostgreSQL restart source is authored, not executed. Public pending
+ projections, independent recovery authorization, installed fault/race evidence,
+ acquisition/disclosure/ACK and translated formal artifacts remain open.
+
+## Native review-contract checkpoint
+
+- Recipe source `/3` requires `auths.effect-review-contract/1`; source `/2`
+ is refused. Every non-binding profile argument must have one compatible
+ presentation, with a bounded trusted label and explicit unit for integers.
+ Native binding fields, exact destination/request, installation, reservations,
+ limits, recovery declarations and uncertainty are always present in private
+ review. Consequences are mandatory operator-owned recipe data. The gateway
+ performs no provider-name inference or unit conversion.
+- Recipe review `/3`, private effect review `/2`, private signing output `/2`
+ and effect compiler `/3` make the cutover explicit. Review semantics participate
+ in the recipe digest and effect identity. There is one native full-value
+ renderer (`auths.literal-complete-review/1`), bounded to 1 MiB. It emits literal
+ ASCII JSON, escapes control/bidi/Unicode content, never interprets markup,
+ and refuses unknown renderer versions or oversized output. The CLI publishes
+ signing-request files only after complete local review output succeeds.
+- Native required artifact generation passed on `6d38bb1c` plus this source:
+ `recipe::tests::native_review_contract_fixture_is_current` and
+ `observed_tests::effect_native_tests::native_compilation_approval_and_reverification_keep_exact_request`
+ (one generator each). The former emits valid recipe fuzz seeds and exact
+ negative review inputs for missing fields/consequences, unsupported version,
+ wrong type, control labels and duplicate fields. The latter includes a hostile
+ bidi/content sentinel, lossless literal-render round trip and renderer refusal.
+ These establish native compilation/presentation only, not installed human
+ approval or human understanding. Entry and semantic-closure generators passed.
+ Native CLI approval-envelope construction passed and matched the regenerated
+ signed effect corpus exactly; it also compiled the updated binary.
+- OpenAPI derivation emits a source `/3` draft with incomplete consequences and
+ integer units. It cannot compile until an operator supplies those facts.
+ Derivation tests use explicitly synthetic declarations for transport comparison;
+ they do not turn descriptions from an untrusted document into review authority.
+- OpenAPI corpus artifact generation completed for local documents and pinned
+ GitHub/OpenAI inputs. Todoist's current URL does not match its pinned bytes;
+ the two full-vendor cases were explicitly skipped during generation. Its
+ derived recipe/review/output commitments were updated from the native-generated
+ excerpt after verifying the previously committed vendor and excerpt recipes
+ and profiles were byte-identical (the corpus requires this equivalence).
+ Existing vendor provenance remains unchanged. Full pinned-vendor re-execution
+ is pending; no digest pin was changed and ordinary configured verification
+ still fails for missing/mismatched files. Inputs retained for later generation:
+ `/tmp/auths-ap074-openapi-inputs` (GitHub/OpenAI only).
+- Remaining review work is the installed authenticated approval-channel journey,
+ supported frontend rendering and failure boundaries, privacy checks, and
+ cross-language/installed evidence. P2–P4 remain open and no complete A01–A22
+ acceptance claim follows from this checkpoint.
+
+## Replay projection checkpoint
+
+- Application replay of a granted or terminal effect now reads only its full
+ matching native binding and returns the retained journal outcome. It never
+ enters the credential/observation path. A reserved effect remains resumable
+ only through fresh approval/authority and the single-winner grant already
+ implemented. Changed bindings remain explicit conflicts.
+- The public native result adds `reserved` and its bounded journal revision.
+ An unacknowledged grant reloads the journal: pre-grant state is reserved,
+ committed grant state is unknown, and unavailable/corrupt state is
+ indeterminate (`gateway.attempt.state-unavailable`). None returns a permit.
+ Python and TypeScript decode the native variant; they do not decide recovery.
+- Native `replay-projection.json` artifact construction passed on `ed907973`
+ plus this source using `scenario_tests::native_replay_projection_fixture_is_current`.
+ It drives the real file store and submission driver with a synthetic command
+ oracle, concurrent reopened-store replay, pre-commit failure, post-commit lost
+ acknowledgement and unavailable reread. Counts show zero replay custody/read/
+ write calls and no journal change. It also constructs native valid/invalid
+ reservation decoding cases consumed by authored binding tests.
+- Qualification execution forbids replay leases and new confirmation credit.
+ Maintained family corpora, live harness expectations and examples now distinguish
+ historical outcomes from binding conflicts. Retained operator observation tests
+ explicitly invoke that path rather than using replay. That existing operator
+ path, automatic post-write reads and other ancillary reads still need independent
+ bounded native acquisition authority/accounting; this checkpoint does not close
+ A11/A15, installed bindings, full adapter faults or the authenticated journey.
+- Attempt-scenario, stable-code, custody and final semantic-closure corpus
+ construction also passed via their exact native artifact generators. Binding tests, broad conformance,
+ translated formal artifacts and hosted CI are not yet executed. No AP073 work
+ or delegation occurred.
+
+## Independent recovery compilation checkpoint
+
+- Bound-effect `/3` and compiler `/4` use one sealed mutation/acquisition sum,
+ one native verifier, one exact approval and one complete private renderer.
+ Fixed-window mutation reservations remain fixed; source-lifetime read
+ reservations declare separate attempt-count and response-byte obligations.
+- A protected observable journal record supplies the source identity, exact GET,
+ response-derived locator when available, projection and echo. The new native
+ read call binds that contract and a distinct read operation. Fresh reader
+ authority is independent of the source writer. Mutation recipes cannot claim
+ the reserved native recovery-read tool, and mutation admission rejects reads.
+- Installed `effect-recovery-call` authors the private native call;
+ `effect-review --recovery-of` compiles its independently signed proof using
+ current installed configuration, connection and deployment time. Both are
+ effect-free: no credential lease, provider read, reservation or journal update.
+- Required `native_recovery_acquisition_fixture_is_current` generation passed:
+ real signed native proofs and file-derived source target; expired writer,
+ distinct reader, exact read review/approval, approval expiry/substitution,
+ changed source/contract/limit/namespace/operation refusal, reserved mutation
+ tool refusal, mutation-admission refusal and unchanged source journal. The
+ source grant uses a synthetic oracle and performs no provider send. This is
+ compilation evidence, not installed recovery execution or A15 closure.
+- Mutation binding, entry, attempt-scenario and stable-code native artifact
+ generators passed. Native CLI approval-envelope construction passed and
+ matched the regenerated signed corpus exactly. These runs used `eb3c31fb`
+ plus this compiler source; remote artifact commit `31a825d0` was inspected and
+ preserved by fast-forward. After the final boxed protocol representation,
+ recovery compilation and `semantic_closure::tests::semantic_closure_is_current`
+ generation both passed on `31a825d0` plus this source (one generator each).
+- Acquisition admission, conserved charging, fenced entry, evidence freshness
+ and transport remain unwired. The compiler currently refuses changed source
+ connection/generation and managed bounded-policy extensions; safe credential
+ replacement and enforcement of read-policy bounds are still required work.
+ Existing direct operator re-observation and all ancillary reads still need
+ replacement. No full acceptance gate, broad local suite or hosted CI is claimed.
+
+## Recovery acquisition journal checkpoint
+
+- Attempt journal `/7` has one common immutable header, approval history, entry
+ audit and codec with closed mutation/acquisition state alternatives. Both
+ protocols use the existing kernel transition predicate and adapter fence.
+ Acquisition cannot acquire mutation observation facts or change protocol.
+ Support bundle `/2` projects protocol and stage separately.
+- Recovery admission atomically inserts the read claim and one permanent
+ `auths.effect-allocation/1` slot charging one invocation and the bound maximum
+ response bytes. All read IDs share the source's 16-slot capacity. Equal replay
+ retains the original allocation; exhaustion records an uncharged refusal;
+ sweeping never refunds source-lifetime allocation. PostgreSQL source schema
+ `/8` admits this non-expiring kind; obsolete disposable schemas are rejected.
+- A read grant fences the read record, exact observable source, allocation and
+ installed authority/configuration/connection dependencies. Only acknowledged
+ persistence returns a one-use read permit. Lost permits project unknown and
+ cannot be reconstructed. Bounded response bytes remain private; public facts
+ distinguish acquisition response from mutation observation and disclosure ACK.
+- `GatewayEngine::acquire_recovery` now connects native read approval, admission,
+ custody, final grant, permit-consuming transport and response recording.
+ Recipes needing additional credential reads fail closed until those reads
+ have their own acquisition accounting. No application/operator execution route
+ or source-evidence refinement has been cut over by this checkpoint.
+- The native recovery artifact generator passed on `964764f2` plus this source:
+ real signed reader proofs, file journal, repeated reservation without charging,
+ single grant and stale contender refusal, reopen after permit loss, concurrent
+ competition for the last 15 of 16 slots, no sweep refund, bounded private
+ response/replay and mixed-protocol support projection. Entry validation and
+ provider response are synthetic. This is not installed authority, actual
+ provider transport, independent-process or PostgreSQL execution evidence.
+- Mutation entry/replay, attempt scenarios, stable/production codes, custody and
+ qualification schema/verification artifact generators also passed on this
+ source. Native mutation binding and final semantic-closure construction passed;
+ CLI approval-envelope construction matched the signed corpus exactly. The new
+ kernel protocol negative tests and revised Kani oracle are
+ authored, not executed. Translated formal artifacts and hosted CI remain open.
+- Still required: admission/current-configuration ordering, safe replacement
+ credentials and managed read-policy bounds, freshness-bound refinement of the
+ source, all ancillary read cutovers, typed disclosure/ACK, authenticated UI and
+ bindings, installed fault/race evidence and final formal/CI gates. No A01–A22
+ acceptance criterion is closed by this checkpoint.
+
+## Recovery evidence and operator cutover checkpoint
+
+- Journal `/8` links newly recorded source evidence to the independently
+ authorized acquisition effect, operation, revision, original receipt time and
+ exact private response digest. It preserves source binding, actor attribution,
+ grant, response history and charges. Source/read/allocation and the deployment
+ clock are checked under the existing adapter fence; no new transaction owner
+ or provider capability is introduced by evidence recording.
+- Evidence is fresh through age 30 seconds; older or future-dated bytes cannot
+ newly refine a source. HTTP error, wrong version, malformed/missing/multiple
+ records and duplicate JSON members are unusable. Value equality without the
+ expected echo cannot resolve an unknown write. Recipe observation parsing now
+ rejects duplicate members at every depth while retaining its byte/depth bounds.
+- Exact supported echo/value evidence can refine an unknown mutation. Its private
+ provenance is immutable; repeated recording after expiry returns history with
+ the original time and charges. The source's retained contract remains available
+ for exact read replay after refinement; new read entry still requires an
+ eligible source under the grant fence.
+- Removed `GatewayEngine::reobserve` and the operator `reobserve` command. Admin
+ request/response `/2` exposes `recover` with a second bounded native
+ proof/action/approval frame and `recovery-record` for previously acquired
+ evidence without provider/custody I/O. Both submission channels reuse one
+ bounded input decoder. The former operator driver is a test-only reference.
+ Automatic post-write and other ancillary/application read paths still require
+ cutover; they are not made compliant by removing the operator endpoint.
+- Native recovery report `/1` exposes separate source/acquisition stages,
+ revisions, retained count/byte capacity and evidence disposition. It excludes
+ provider bytes, locators, echo, action commitments and private response digests.
+ This report is not a content-disclosure capability.
+- `recovery-refinement.json` extends the signed-proof/file-journal corpus with
+ unusable/inconclusive cases, age 30/31 and future-receipt boundaries, simultaneous
+ refinement (one update and one exact history result), expired-writer history,
+ unchanged receipt time/charges, recipe substitution refusal, redacted reports
+ and closed admin frame grammar. Provider bodies, entry validation and the
+ controllable clock remain synthetic. Installed sockets/provider transport,
+ lock-wait expiry, PostgreSQL, crash/ACK-loss and full binding evidence are open.
+- Required native recovery/entry/replay, attempt-scenario, stable/production-code,
+ custody and qualification schema/verification artifact construction passed.
+ Final CLI approval-envelope construction matched the signed corpus; final
+ semantic closure generation passed on `1322d182` plus this source. Remote bot
+ commit `1322d182` changed only formal qualification source-closure hashes and
+ was preserved by fast-forward. No broad local suite or hosted CI was run.
+- This source does not close A01–A22. Safe credential replacement, managed read
+ bounds, ordered admission, remaining read cutovers, disclosure/ACK, complete
+ UI/bindings, generated histories, final formal/release artifacts and hosted CI
+ remain required.
+
+## Submission observation cutover checkpoint
+
+- Removed automatic post-write observation from the shared submission order and
+ driver. A recorded response stops with `ResponseRecorded`; unknown transport
+ stays conservative. Removed the obsolete mode enum, resume/read-back phases
+ and actions, and test-only operator entry from the shipping driver.
+- Direct `ObservableGatewayAttempt` observation-writing methods are now
+ crate-private and test-only. The shipping evidence refinement path is the
+ source/read/allocation recovery fence. Retained record/store conformance
+ mechanics request their observation explicitly from a test oracle and make no
+ AP074 read-authority claim.
+- Updated authored scenario stages, credential/read/lease counts and exact
+ request lists, plus measured qualification replay expectations. The generated
+ replay corpus now measures a successful response with one lease, one write,
+ zero provider reads and no evidence; its replay performs no additional I/O
+ and preserves the complete snapshot. Python/TypeScript parity consumers also
+ decode this native response case.
+- Required native replay, attempt-scenario and final semantic-closure artifact
+ construction passed (one exact generator each), on `9f717d39` plus this source.
+ The replay generator exercises the new response/no-read assertions; the
+ scenario generator constructs expected corpus data, not adapter conformance.
+ No broad local suite or hosted CI has been run. Existing installed/harness
+ journeys with automatic-observation expectations still require explicit
+ independent acquisition/approval cutover, so full-suite readiness is not
+ claimed. Generated/translated formal links remain pending. No acceptance
+ gate closes.
+- Next: bound/account prerequisite and application reads, ordered admission,
+ replacement credentials/read-policy limits, disclosure/ACK, maintained
+ consumer/UI cutover and complete installed/fault/formal/CI evidence.
+
+## Prerequisite declaration checkpoint
+
+- Compiler `/5` now commits the full ordered prerequisite acquisition plan in
+ the mutation transport identity and retains the original closed read values
+ beside the verified command. Native private review renders each typed purpose,
+ exact request, response limit, one-call allowance, transport/freshness bounds
+ and no-disclosure declaration. Old compiler bindings are refused directly.
+- Account comparisons pin the installed account commitment/domain; denied reads
+ pin their ordinal and refused statuses; ceiling reads resolve verified values;
+ pre-entry reads pin pointers and selected native authority. The seven-read
+ maximum excludes onboarding probes and post-write observations.
+- A new signed native fixture covers Stripe's four prerequisites and Airtable's
+ pre-entry plan. Initial generation caught fixture-only trust namespace and
+ delegation-depth errors, now corrected. The shared test trust helper now uses
+ the requested MCP service instead of the default harness service. It does not
+ broaden installed trust.
+- Required native prerequisite, effect/review/approval, recovery acquisition,
+ mutation-entry and replay artifact generators passed on `119d1f78` plus this
+ source (one exact test each). Entry generation refreshed the signed journal
+ and transition fuzz seeds. Native CLI approval-envelope construction matched
+ the signed corpus, and final semantic-closure generation passed. These are
+ artifact-construction results, not full-suite or hosted CI evidence.
+- Preserved remote bot commit `7c7ff71b` by fast-forward after inspecting its
+ isolated qualification source-closure hash update. No hosted workflow results
+ have been inspected.
+- This is compilation/approval work, not durable read accounting. Existing raw
+ prerequisite calls remain an explicit open cutover. They must next consume the
+ retained compiler requests through the shared journal, capacity accounting and
+ one-use permits. No A01–A22 criterion closes; installed UI/bindings, disclosure,
+ replacement credentials, ordered admission, faults and formal/CI remain open.
+
+## Durable prerequisite checkpoint
+
+- Journal `/9` now has a closed recovery/prerequisite acquisition-plan variant.
+ Both use the existing shared codec, reducer, permanent count/byte allocation,
+ fenced entry and response persistence. PostgreSQL storage schema stays `/8`;
+ its current record kinds suffice. Obsolete attempt records are rejected.
+- A prerequisite's component key binds the parent logical operation and compiler
+ ordinal, so a changed binding cannot evade replay by changing its digest.
+ Its bounded canonical transport witness proves approved plan membership but
+ cannot construct a transport request. Only the retained sealed read reaches
+ the permit and transport.
+- Read grants fence the child, still-reserved parent, allocation and current
+ native dependencies. Submission's account, denied, pre-entry and ceiling
+ reads now consume these permits. Responses are evaluated only after durable
+ recording. Lost permits/responses remain charged and cannot be reminted.
+- Final mutation entry fences every required child response and allocation and
+ refuses missing, uncertain, future or stale evidence. Receipt timestamps are
+ immutable; signed pre-entry observations and ceiling bases use the earlier
+ acquisition grant time conservatively. Seven reads add 14 dependencies, for
+ at most 52 of the existing 64-record fence bound.
+- `prerequisite-entry.json` adds eight file-journal histories across native
+ Stripe/Airtable plans: recorded response, missing read, lost permit and stale
+ response. Equal reservations preserve capacity, competing handles cannot
+ both grant, reopening returns history, substituted membership fails, and
+ future/stale reads are unusable. This uses explicit-time oracle grants and
+ synthetic bodies with zero provider calls. In particular, the Airtable case
+ does not establish installed pre-entry requirement eligibility.
+- Required artifact generators passed (one exact test each): prerequisite
+ compilation/entry, independent recovery, mutation entry/fuzz seeds, native
+ replay projection and semantic closure. Earlier compilation attempts caught
+ stale test-only raw-read calls and zeroizing-buffer type mismatches, corrected
+ before generation passed. No broad local suite or hosted CI was observed.
+- Still open: ordered initial admission, tighter selected observer evidence
+ validation at the final fence, full prerequisite fault/transport/PostgreSQL
+ evidence, application/onboarding reads, safe credential replacement and read
+ policy bounds, disclosure/ACK, Python-owner removal, installed UI/bindings,
+ formal/performance and final CI. No A01–A22 criterion closes at this checkpoint.
+
+## Final prerequisite evidence checkpoint
+
+- Installed mutation entry now evaluates the required response purposes under
+ the existing writer fence after native revalidation and acquisition membership
+ checks. It reuses the existing account/denied, recipe ceiling/ratio and native
+ selected-observation predicates; no second evaluator or transport owner.
+- Signed pre-entry observations must verify canonically against the pinned
+ observer and match the exact acquired subject, facts and original grant time.
+ Selected native ages and anchor validity are evaluated at the fenced time;
+ the acquisition's 30-second ceiling cannot relax them. The installed
+ configuration commitment now includes the gateway's selected observer.
+- Relative-ceiling basis, response digest and time must match the retained read,
+ and original argument, ratio and binding checks still apply. Missing or extra
+ checkpoint evidence is refused. Account, ceiling and pre-entry JSON share
+ recovery's bounded duplicate-rejecting decoder.
+- New native predicate artifacts cover 15 Airtable evidence cases and 15 Stripe
+ guard/checkpoint cases. They include canonical invalid signatures, changed
+ signer/subject/facts/time, selected-age limit/limit-plus-one, account/currency
+ mismatch, excess capability, altered basis/digest/time, missing/extra evidence
+ and duplicate JSON members. Stripe compilation now uses a known synthetic
+ account commitment for a genuine matching comparison in this test corpus.
+- The exact prerequisite compilation/entry/guard generator, signed-observation
+ predicate generator and semantic-closure generator passed. The new signature
+ case decodes and re-encodes a canonical observation with altered signature
+ bytes, so it specifically exercises signature refusal. Construction caught
+ test-import/parser-visibility issues, corrected before generation passed.
+ No broad local suite or hosted CI was inspected.
+- These are native signed effects and synthetic response/predicate results, not
+ installed lock/process/provider or PostgreSQL evidence. The test-only store
+ grant oracle intentionally remains separate from installed native entry.
+ No A01–A22 acceptance case closes. Ordered admission, remaining read routes,
+ replacement credentials/read bounds, disclosure/ACK and Python removal,
+ authenticated UI/bindings, full faults/formal/performance and CI remain open.
+
+## Ordered initial admission checkpoint
+
+- Mutation, independent recovery and prerequisite claims now validate native
+ authority/approval and current connection/configuration/qualification under
+ the same writer fence as their atomic claim and capacity insert. The adapter
+ decision can stamp only the first record's bytes; other batch fields remain
+ fixed. Initial admission time is sampled inside the fence.
+- Count/sum frontier discovery remains bounded candidate selection. Insertion
+ checks target vacancy and validates immutable sum predecessors under the
+ fence. Acquisition admission checks the eligible source in the same snapshot.
+ Mutation admission uses at most 55 of the existing 64 records. The existing
+ two-second lock/decision bounds and file rollback manifest remain in use.
+- Runtime callers now supply installed validation before claim publication;
+ recovery preparation precedes admission and is reused for entry. Transport
+ grant timestamps are sampled after admission, while the original monotonic
+ turn deadline remains enforced. No admission result itself grants transport.
+- The new fixture caught expiry being checked while constructing a replacement
+ candidate before reading existing history. Replay now loads retained history
+ first, preserving the original reservation and charges after approval expiry;
+ a later grant still requires eligible current approval and native authority.
+- `effect-admission.json` covers real file batch atomicity, missing/changed
+ dependencies, a dependency writer ordered before admission, collision/replay
+ suppression of the decision, admission time, expired-approval refusal without
+ writes, retained count/sum charges and expired-entry refusal. Native authority
+ uses an explicit-time approval oracle; dependency bytes are synthetic. The
+ same conformance source is authored for PostgreSQL, not executed locally.
+- Required artifact construction passed: new native admission, prerequisite
+ compilation/entry/guard, independent recovery, mutation entry/fuzz seeds and
+ replay projection (one exact generator each). Construction also caught and
+ corrected test Arc/zeroizing-buffer type mismatches. Final semantic closure
+ is regenerated with this checkpoint. No broad local suite or hosted CI was
+ observed, and no A01–A22 gate closes.
+- Remaining implementation: application/onboarding reads, safe credential
+ replacement and read-policy bounds, controlled disclosure/authenticated ACK,
+ Python-owner removal and authenticated UI/bindings. Complete installed
+ authority/adapter/process faults, differential/formal/performance and final
+ hosted evidence remain required.
+
+## Application recovery route checkpoint
+
+- Removed the unapproved provider `read-back` request from the production
+ engine, application protocol, CLI and Python/TypeScript client methods.
+ Observation `/3` exposes local outcome/pre-entry history only; `/2` is refused.
+ Application `auths.gateway-recover/1` carries a source operation and the same
+ bounded native proof/action/exact-approval envelope as operator recovery.
+- The application adapter calls `GatewayEngine::recover` directly. It returns
+ only the native redacted recovery report or native refusal, never provider
+ response bytes or a newly signed provider field. Existing source eligibility,
+ native verification, exact approval, capacity, fencing and refinement still
+ apply. This is not general content acquisition or recipient disclosure.
+- CLI `app-recover` and SDK `recover` methods use this frame. Thin clients return
+ bounded native JSON bytes unchanged; authored parity tests use the native
+ frame corpus. The old synthetic observation reference now lives only in
+ direct tests and the private test harness control socket, never the shared
+ application route. Maintained reference consumers and current docs reflect
+ that restricted scope. The synthetic submission harness refuses recovery
+ rather than pretending to implement installed authority.
+- Native artifact construction passed: application socket framing/dispatch and
+ production CLI/socket round trip, including the harness feature build. The
+ counting/refusal server establishes decoder, routing and exact transport
+ bytes; it does not establish installed authorization or provider execution.
+ Retired submit/admin/read-back fuzz seeds are explicit rejection cases, with
+ current submit/recovery seeds generated from the native envelope.
+- Test-only raw claim/approval helpers no longer compile into the standalone
+ harness when unused. Final source closure is regenerated with this slice.
+ No broad local suite, language suite or hosted CI was run/observed, and no
+ A01–A22 acceptance gate closes.
+- Remaining: onboarding/rotation reads and recovery ancillary reads, safe
+ credential replacement, managed read-policy limits, general acquisition and
+ controlled disclosure/ACK, Python-owner removal, authenticated UI/bindings,
+ and complete installed/process/PostgreSQL/fault/formal/performance/CI evidence.
+
+## Managed recovery read budget checkpoint
+
+- Independent recovery compilation now consumes selected native count/sum
+ policies instead of refusing all managed bounds. The bounded argument is
+ exactly `maximum_response_bytes`; every sum reserves the full approved
+ response ceiling. Mutation recipe spend requirements remain mutation-owned.
+ Read count/byte identities are separate from mutation count/amount identities.
+- Mutation and recovery use the same fixed-window candidate arithmetic, slot
+ encoding, sum-predecessor checks and final membership validation. Recovery
+ admission inserts managed slots and the permanent source allocation in one
+ fenced batch. Replays add no charge; failed contention inserts nothing.
+ Prerequisites do not reserve the parent policy again.
+- Compiler `/6` binds the managed reservations. Journal `/10` retains immutable
+ acquisition counter membership and rejects the old layout. Decode checks
+ exact counter keys against the signed binding; final entry checks actual
+ slot identity, ownership, cumulative amount and window with current native
+ authorization under the fence. Public snapshots report policy counter counts
+ alongside the source allocation without exposing private keys or content.
+- `recovery-policy-budgets.json` contains three real file histories: final count,
+ final byte sum and sum admission after an already charged unknown read. Each
+ races two independently approved operations for the last capacity, retains
+ exactly one winner, refuses substituted membership and old-window entry,
+ preserves lost-permit charges through reopening, and requires a newly
+ approved binding for a fresh window. Wrong argument, too-small ceiling and
+ too-small sum policies are refused. Entry uses an explicit-time oracle, not
+ an installed provider or PostgreSQL deployment.
+- Required recovery, mutation binding/approval, prerequisite compilation/entry
+ and predicates, admission, mutation entry/fuzz, attempt scenario and replay
+ artifact generators passed. Application frame/CLI and final semantic closure
+ regeneration also passed. The native binding generator now writes and checks
+ the exact approval envelope alongside its signing request. No broad local
+ gate or hosted CI has been run. No A01–A22 acceptance criterion closes at this checkpoint.
+- Remaining read implementation: onboarding/rotation and recovery ancillary
+ reads, plus safe replacement credentials. General acquisition, controlled
+ disclosure/authenticated ACK, Python-owner removal, authenticated UI/bindings,
+ full installed/process/PostgreSQL/fault/formal/performance and hosted evidence
+ remain required.
+
+## Recovery credential prerequisite checkpoint
+
+- Recovery compilation now includes the installed recipe's account and
+ denied-capability reads in its immutable request commitment and native
+ review. Compiler `/7` uses the complete acquisition transport witness;
+ journal `/11` rejects superseded recovery plans. The installation-only
+ credential probe remains an onboarding obligation.
+- Mutation and recovery share the credential-plan compiler, prerequisite
+ record/allowance, consumed-permit acquisition helper and account/capability
+ predicates. A child can enter only while its mutation or independent recovery
+ parent is reserved. Another prerequisite cannot be its parent. Recovery final
+ entry fences all required children/allocations and evaluates their purpose
+ and freshness alongside native authority, original source and managed slots.
+- The production recovery runtime no longer rejects recipes wholesale for
+ declared credential checks. It acquires and records them in order, stops on
+ a failed account/capability predicate, and validates them again at final
+ entry. Child replay loads original evidence; unknown children cannot remint
+ permits. Their explicit allowances do not charge the parent's managed budget
+ again. At most five recovery children give a maximum 48-record entry fence.
+- `recovery-prerequisites.json` covers recorded, missing, lost-permit, stale,
+ wrong-account and excess-capability histories using native Stripe recovery
+ proofs and real file state. Native actual-purpose evaluation is reported
+ separately from oracle entry. The corpus checks full human review, exact
+ recompile, missing-parent refusal, corrupted membership witnesses, retained
+ child charges and restart projections. It does not claim installed provider
+ transport or PostgreSQL evidence.
+- The prerequisite/recovery artifact generators passed after correcting a
+ missing type import and making the Stripe source fixture retain its required
+ response-derived refund locator. The fixture does not invent an observable
+ target for a source without that locator. Binding/approval, mutation entry,
+ admission, attempt/replay, code inventory, application framing and production
+ CLI/socket generators also passed. Final source closure is regenerated with
+ this checkpoint. No broad local gate or hosted CI has been observed.
+- Remaining: first-install and candidate-rotation read authorization/journaling,
+ safe replacement credentials, general acquisition, disclosure/ACK, Python
+ decision removal, authenticated UI/bindings and full installed/process/adapter/
+ formal/performance/CI acceptance. No A01–A22 criterion is fully closed.
+
+## Candidate credential compilation checkpoint
+
+- Compiler `/8` adds the reserved native candidate-check tool and sealed target.
+ The full proposed connection and exact predecessor/absence are pinned; a
+ rotation preserves descriptor, account, authorizations and administrative
+ state. Native verification, selected authority, managed read bounds, binding
+ construction and exact approval share the existing recovery compilation path.
+- Candidate plans retain recipe order (probe/account/denied), with the final
+ declared read as the primary acquisition and prior reads as declared children.
+ No-read recipes cannot invent a network operation. Private review includes
+ the actual GET/HEAD and full descriptor/previous state. Source-lifetime keys
+ use connection ID, credential generation and reference commitment; changing
+ the operation or timestamp cannot reset the 16-attempt allowance.
+- The native candidate corpus covers first install, active and disabled
+ rotations, HEAD final reads and probe-only checks; it rejects substituted
+ candidate/predecessor/recipe/contract/limits/extra arguments, changed connection,
+ expired or reused approval, revoked predecessors and changed rotation accounts.
+ Existing mutation/recovery admission and orphan-child routes refuse candidates.
+ Probe predicates accept the exact value and reject wrong/duplicate JSON values.
+- This is compilation and explicit refusal evidence, not candidate execution.
+ Staged custody, current-or-absence fencing, candidate admission, ordered reads,
+ safe publication and onboarding/rotation CLI cutover remain required. The
+ runtime still uses the previous onboarding checks until that direct cutover.
+ Journal `/11` is unchanged; no candidate journal decoder is introduced here.
+- Required native artifact construction passed for candidate compilation and
+ the changed compiler identity's binding, approval, recovery, prerequisites,
+ admission/entry, replay, attempt and code fixtures. Application framing passed.
+ The CLI generator exposed macOS inheritance of the listener's nonblocking flag;
+ the accepted test stream now explicitly uses blocking I/O with its existing
+ five-second deadline. CLI regeneration and final source closure then passed
+ on `15b3cad1` plus this source. No broad local gate or hosted CI was run; no
+ A01–A22 acceptance criterion closes.
+
+## Candidate acquisition journal/runtime checkpoint
+
+- Journal `/12` adds candidate source plans to the shared acquisition owner.
+ Admission pins the actual expected predecessor or first-install absence;
+ ordinary and candidate entry share native verification/configuration/approval
+ checks. Candidate checks preserve disabled state and still require the exact
+ workload/profile. Revocation or any predecessor replacement prevents later
+ child and primary grants. The source/native dependency on the same connection
+ row is read once, with both predicates evaluated against that snapshot.
+- Entry audit `/3` records absence separately from a present record commitment.
+ Native candidate histories exposed the previous all-present audit assumption;
+ the explicit layout change makes first-install grants representable without
+ weakening mandatory dependency checks. Old layouts are refused.
+- The engine candidate acquisition API leases the exact staged credential after
+ admission and uses the shared ordered-read helper. It validates earlier probe,
+ account and capability responses before primary entry; final entry repeats
+ their actual predicates/freshness. Response/unknown facts retain allocations
+ and never grant publication. Cross-proposal allocation collision checks prevent
+ a credential identity from changing its already established byte ceiling.
+- The native candidate artifact generator covers first install, active/disabled
+ rotation, HEAD-final and probe-only variants using actual native fence checking
+ and real file state. Histories cover missing/bad/stale children, predecessor
+ changes before admission/child/main entry, lost main permits, replay and reopen.
+ Configuration/clock and response bodies are synthetic; no installed provider,
+ publication or PostgreSQL claim is made. A capacity fixture initially reused
+ equal limits; its probe-only recipe now changes the limit to exercise refusal.
+- Artifact construction passed on `f0a364fd` plus this source: 36 candidate
+ journal histories and three cross-proposal capacity cases; recovery/prerequisite
+ compilation, binding/approval, actual prerequisite predicates, admission/entry,
+ replay, attempt, code/custody, application frame/CLI and final semantic closure.
+ One extraction build required correcting a moved helper's module qualifier;
+ the new capacity case required actually differing byte ceilings. These were
+ corrected before successful regeneration. No broad local suite or hosted CI
+ was run or observed.
+- Remaining onboarding work: durable staging/authoring CLI, retained final-purpose
+ validation and safely fenced publication, then removal of the old raw check
+ paths. General acquisition/disclosure/ACK, safe recovery credential replacement,
+ Python-owner removal, authenticated UI/bindings and full acceptance evidence
+ remain open. No A01–A22 criterion is fully closed.
+
+## Candidate publication checkpoint
+
+- The operator API publishes exact staged candidate records through the existing
+ store writer fence. It validates the root/child records and allocations, fresh
+ native authority/approval/configuration/qualification, actual final and child
+ credential predicates, and predecessor/absence before connection insertion or
+ replacement. No acquisition state, charge or evidence timestamp changes.
+- Exact already-current state has a separate result with no causal attribution,
+ capability or evidence refresh. The runtime checks exact custody beforehand
+ and reloads through the generation-floor owner before scheduling retirement.
+ Independent custody availability is not claimed atomic with shared state.
+ Publication errors do not claim the connection write definitely did not occur.
+- Candidate artifact construction passed on `57332c58` plus this source: 80
+ native file-journal histories, including publication success/replay and
+ refusals for bad final purpose/version, unknown primary, stale primary/child,
+ expired approval, changed configuration and changed predecessor. Response
+ bodies, clock and installed configuration remain synthetic. This is not
+ installed provider/custody, process-fault or PostgreSQL evidence.
+- Code inventory and final semantic-closure artifact construction also passed.
+ No broad local gate was run. The publication API is an operator method only;
+ no application framing or existing provider request identity changes.
+- Durable staging/proposal CLI and removal of raw onboarding/rotation paths
+ remain next. General acquisition/disclosure/ACK, safe recovery replacement,
+ Python-owner removal, authenticated UI/bindings and full acceptance evidence
+ remain open. No A01–A22 criterion is closed and no hosted CI was observed.
+
+## Private candidate proposal checkpoint
+
+- Added `auths.gateway-credential-candidate/1`, bounded canonical private JSON
+ containing exact candidate/predecessor CBOR and recipe digest. It preserves
+ the target for future CLI approval/restart steps without serializing authority,
+ custody, a check result or a transport capability. The native decoder derives
+ the plan from the pinned recipe and revalidates the connection relationship.
+- The five-target native generator now passes the decoded proposal into every
+ signed compilation and journal/publication history. It authors exact-roundtrip
+ and wrong-schema/recipe/record/predecessor, unknown/duplicate/noncanonical and
+ oversize refusals. Initial artifact compilation exposed a test assertion's
+ demand for private-plan Debug; the assertion now compares equality without
+ adding a private-data Debug implementation.
+- Native candidate artifact generation passed after the assertion correction:
+ five proposal roundtrips/negative sets and all 80 journal/publication histories.
+ Final semantic closure is regenerated with this source. No broad local gate
+ or hosted CI was run; no acceptance criterion closes at this checkpoint.
+- Staging and CLI cutover remain pending; this format alone does not implement
+ them. No new schema migration, legacy reader or alternate execution path was
+ added. Existing candidate compiler/binding/journal identities are unchanged.
+
+## Candidate rotation staging and CLI checkpoint
+
+- `stage_candidate_rotation` checks installed configuration, current exact
+ connection, non-revocation, prefix, read plan, current custody and generations
+ before storing a successor. Durable cleanup notes precede the custody write.
+ It never leases, reads a provider, publishes, or discards another staged secret.
+ The fixed deployment-clock proposal can be persisted through approval/restart.
+- Added production `candidate-stage-rotation`, `effect-candidate-call`,
+ `effect-review --candidate`, `candidate-acquire` and `candidate-publish` CLI
+ commands using native owners. Stage/publish use operator custody identity;
+ acquisition uses runtime read identity. Proposals/calls are new mode-0600
+ files. Existing output and permissive proposal files are refused.
+- Read-bearing recipes now refuse the old `rotate`/`rotate-prepare`/
+ commitment-only `rotate-commit` routes before I/O. Recipes without reads keep
+ the offline operator path. Joining an already-published connection confirms
+ its exact credential reference without any provider read. First-install raw
+ checks and maintained deployment/qualification journey cutovers remain open.
+- Independent operator processes exposed persistent file-custody snapshot
+ overwrites. The existing custody owner now reloads bounded canonical state
+ under a stable cross-process sidecar lock. File-lock contention fails
+ immediately; stale handles cannot overwrite committed candidates or hide a
+ completed deletion. Reads allocate once into zeroizing buffers, with length
+ bounds and exact EOF checks. No new dependency, credential identity or
+ lifecycle owner is introduced; shared connection/custody atomicity is not claimed.
+- Native artifact construction passed on `861271a3` plus this source: five
+ durable staging cases (active/disabled, revoked, missing cleanup journal,
+ bad prefix), same-secret replay/different-secret refusal, custody reopen,
+ unchanged connection/witness and refusal of unapproved guarded rotation.
+ The 80 native candidate journal/publication histories and code/custody
+ inventories passed as well. Initial generator errors were a missing semicolon
+ and a fixture directory lacking mode 0700; both were corrected.
+- The production CLI artifact passed with actual competing processes, one
+ staging winner, held-custody-lock refusal, exact native call bytes, output
+ preservation and owner-private permissions. Its initial compile required a
+ borrowed operation string; its synthetic installation then needed the same
+ pinned qualification root and empty loaded bundle as the production loader.
+ Those fixture inputs were corrected instead of weakening configuration checks.
+- After the file-custody change, native staging regeneration passed again,
+ including stale-reader observation of a completed deletion; the independent
+ CLI process generator also passed again. Final source closure was regenerated
+ after all source edits.
+- These are scoped file/CLI facts, not a fully approved installed provider
+ rotation, AWS/PostgreSQL conformance or process-crash qualification. First
+ installation, full approved provider journeys, safe recovery credentials,
+ general acquisition/disclosure/ACK, Python-owner removal, authenticated UI and
+ complete fault/formal/performance/hosted evidence remain required. No A01–A22
+ acceptance criterion closes. No broad local gate or hosted CI was run.
+
+## First-install staging and onboarding cutover checkpoint
+
+- Read-bearing `install` now persists the existing canonical connection format
+ as a private draft, stages exact generation-one custody and saves a private
+ proposal. The shared connection stays absent until approved acquisition and
+ fenced publication. Draft identity/account/descriptor/timestamps survive
+ process restart. `candidate-stage-installation` explicitly resumes from it.
+- The stage initializes durable cleanup notes before custody and pins the host
+ floor to the exact unpublished candidate. An existing matching secret supports
+ retry after ambiguous custody acknowledgement; conflicting secrets or floors
+ refuse. Candidate compilation/acquisition/publication require the existing
+ floor. Runtime cannot repair it; explicit operator restaging may initialize
+ an unpublished floor. Independent custody/floor/output failures retain cleanup
+ obligations and do not imply that custody was never written.
+- Removed production raw onboarding exports and the install checker. The old
+ closed evaluator remains test-only; the pure probe predicate remains shipping.
+ The offline install helper now requires the recipe and refuses declared reads
+ before custody. Join requires cleanup notes before confirmation. Development
+ loopback selection moves from install to approved candidate acquisition.
+- Native candidate artifact construction passed on `1c23ca31` plus this source,
+ retaining all 80 journal/publication histories and adding five first-install
+ staging histories: exact reopen/restage, bad prefix, existing connection,
+ substituted floor and corrupt cleanup journal. The actual CLI generator passed
+ with `testkit-harness,loopback-provider`, covering installation, an independent
+ resume process, identical proposal bytes, private files and conflicts.
+- Initial generator failures were missing test-oracle imports, an incorrect
+ result-field assertion, comparing process-specific witness scopes across two
+ engines, and a test variable shadowing the recipe lock. Corrections retain
+ each process's own zero-I/O witness and the intended refusal assertions.
+- Required generators: `observed_tests::effect_native_tests::candidate_tests::native_candidate_compilation_fixture_is_current`
+ (library), `native_candidate_cli_fixture_is_current` (candidate_cli integration
+ target), code inventories and final semantic closure. These construct scoped
+ artifacts; they are not broad local gates, installed approved provider
+ qualification or hosted evidence. No hosted CI has been observed.
+- Final candidate/code/custody generation passed (four exact tests); the
+ production CLI artifact passed (one exact test), and final semantic closure
+ passed (one exact test). Fetch found upstream bot commit `67506ada`, updating
+ only the prior custody source's formal closure. It was inspected and retained
+ by fast-forward without observing hosted workflow results.
+- Remaining: maintained deployment/example scripts and complete approved
+ installation/rotation journeys; safe recovery credential replacement; general
+ acquisition/disclosure/ACK; Python-owner removal; authenticated UI/bindings;
+ full process/adapter/differential/formal/performance/hosted evidence. No A01–A22
+ acceptance criterion closes at this checkpoint.
+
+## Approved candidate process journey checkpoint
+
+- Added a real production-CLI journey over private installation files, file
+ custody, generation floor, cleanup journal and actual local HTTP sockets.
+ Test-only fixed-key native authoring supplies real canonical proof/action and
+ trust; `effect-candidate-call`, `effect-review`, `effect-approval`, acquisition
+ and publication still use the shipping compiler, verifier and transaction
+ owners. No oracle entry predicate or hand-built approved binding is used.
+- `candidate-process-journey.json` records six histories: successful first
+ installation with active rotation; rotation of a disabled predecessor; bad
+ probe; wrong account; excess final capability; and lost primary response.
+ The provider counts and checks exact request lines, version/credential headers
+ and absence of account-scope headers. Successful first install and rotation
+ each make exactly four declared GETs. Earlier failed predicates stop later
+ reads. Failed/unknown evidence cannot publish a connection.
+- Independent acquisition processes race the same operation: one set of reads
+ reaches the provider. A new process replays retained history without re-entry;
+ every non-connection journal record and allocation remains byte-identical
+ across replay and publication. Lost final response retains unknown and its
+ invocation charge. Exact publication replay reports `already-current` without
+ provider I/O. Disabled rotation preserves state while advancing generation.
+- Invalid approval and old exact approval for a new operation cause no provider
+ calls. Approval alone cannot publish. Native literal review includes every
+ credential request; public acquisition output is scanned for credential and
+ account sentinels. The test signer simulates an external reviewer and makes no
+ human-presence, live-provider, process-isolation or PostgreSQL claim.
+- Exact artifact generation passed on `ed557dab` plus this source:
+ `AUTHS_UPDATE_FIXTURES=1 cargo test -p auths-gateway --features testkit-harness,loopback-provider --test candidate_journey -- --exact approved_candidate_process_journey_fixture_is_current`.
+ The final six-case run passed in 7.27 seconds. Final semantic-closure artifact
+ generation also passed (one exact test). Initial test failures required
+ clearing macOS's inherited nonblocking socket flag and using the existing
+ `disable --store-only` command. Test cleanup avoids a second panic during
+ unwinding. No production check or predicate was weakened to pass the journey.
+- Maintained caller inventory still includes Stripe/Airtable demos, qualification
+ production/live/custody-fault scripts and the refund-approval example. Their
+ mutation-only trust/grants need explicit candidate read authority and exact
+ approval; production first install also needs the qualification bootstrap
+ order preserved. Merely replacing install/rotate command names is insufficient.
+ The no-read operator simulation retains its offline path.
+- This closes the previously missing local approved candidate transport evidence,
+ not P3 or A01–A22. General acquisition/disclosure/ACK, safe recovery credentials,
+ maintained consumers, authenticated UI/bindings and full adapter/fault/formal/
+ performance/hosted evidence remain required. Hosted CI has not been observed.
+
+## Native disclosure acknowledgement and reducer checkpoint
+
+- Extended the existing kernel transition owner with typed disclosure and a
+ distinct terminal `Acknowledged` stage. ACK bytes are separate from provider
+ response/observation fields. All protocols retain immutable identity,
+ allocation, grant and monotonic revision checks. Disclosure cannot acquire
+ provider facts; mutation/acquisition cannot carry ACKs. The provider outcome
+ presence predicate refuses the disclosure terminal stage.
+- Native `auths.gateway-disclosure-ack/1` binds effect, source, protected content,
+ raw-key recipient, channel, delivery instance, projection, ceiling and actual
+ byte count. It uses existing canonical/domain commitment and strict Ed25519
+ verification, derives the recipient through the existing raw-key descriptor,
+ rejects duplicate/unknown/noncanonical/obsolete/oversized input before use,
+ and returns a sealed historical assertion. Debug is redacted and encoded
+ buffers are zeroized. Verification supplies no execution authority.
+- Late evidence requires no current-authority or time argument. It must still
+ match the retained delivery and protected content length and pass the shared
+ reducer. Exact repeated ACKs are read-only comparisons of consistent terminal
+ history; they do not advance revision, change evidence or issue another grant.
+- Exact artifact construction passed on `cac8c75b` plus this source:
+ `AUTHS_UPDATE_FIXTURES=1 cargo test -p auths-gateway --lib -- --exact effect::disclosure_tests::native_disclosure_acknowledgement_fixture_is_current`.
+ The final run passed one test in 0.03 seconds. Its fixture covers 576 protocol/
+ stage pairs, four disclosure histories, immutable state-field substitutions,
+ revision saturation, changed binding dimensions, independently re-signed ACK
+ assertions, foreign signatures, exact byte-count limits and malformed codecs.
+ The initial compiler error combined raw-key construction and principal
+ derivation errors with incompatible Result types; each is now mapped at its
+ own boundary. Final semantic-closure generation passed one exact test.
+- Kani source includes disclosure, the new ACK field and an exact terminal-replay
+ obligation. Kani and translated Lean/refinement evidence have not been run
+ or regenerated; they remain P4 work. No broad local gate or hosted CI was run.
+- This is a native evidence/reducer slice, not an installed release path. Content
+ acquisition/custody, disclosure compilation/approval, transactional admission
+ and entry, authenticated recipient transport, persistent ACK recording and
+ AP070 Python-owner removal remain required. Synthetic fixed allocations in
+ this corpus do not establish durable accounting or actual byte delivery. No
+ A01–A22 acceptance criterion closes.
+
+## Explicit content acquisition compiler checkpoint
+
+- Added the digest-bound opt-in `content` capability with its own
+ `auths.gateway-content-read/1` schema. It selects one string/enum resource,
+ existing fixed/verified path and account headers, bounded sorted literal query
+ pairs, and UTF-8 or JSON-string projection. The operator review includes the
+ complete declaration. An absent declaration cannot compile content reads;
+ the reserved native read tool cannot be installed as a mutation tool.
+- Added `ContentReadTarget` and `compile_content_acquisition` to the existing
+ acquisition compiler. Fresh native read proof, full private target/connection,
+ source allowance, managed policy counters, prerequisites, exact approval and
+ complete native review share the existing effect owner. Full revalidation
+ recompiles content effects. Mutation/recovery/candidate admission cannot be
+ borrowed by a content effect. No content transport or permit was added here.
+- Source allowance keys bind recipe, connection ID, credential generation/
+ reference and exact resource/account arguments, excluding logical operation
+ IDs and administrative metadata. The full connection still binds the approval.
+ The native corpus checks that disable/enable timestamp changes preserve keys
+ while old bound approval refuses the changed installation. Durable allocation
+ enforcement remains part of the next journal work, not a compiler claim.
+- Pure projection returns exact private zeroizing bytes, requires status 200,
+ declared media type and optional UTF-8 charset, and rejects malformed UTF-8,
+ duplicate JSON, missing/non-string projections and both raw/projected limits.
+ Six typed response codes distinguish these facts from native authority denial.
+- Exact artifact construction passed on `72fe870e` plus this source:
+ `AUTHS_UPDATE_FIXTURES=1 cargo test -p auths-gateway --lib -- --exact observed_tests::effect_native_tests::content_tests::native_content_compilation_fixture_is_current pending_vectors::codes::code_inventory_is_current`.
+ The final two-test run passed in 0.06 seconds. The corpus uses real native
+ proof/approval signatures for both projections, fresh/expired recompile,
+ different resource/connection/limit, extra arguments, approval reuse and stable
+ capacity keys; it checks the declared projection limits and all six error
+ codes. Final semantic-closure generation passed one exact test.
+- Initial fixture construction used a nonexistent review method and accidentally
+ supplied content source as profile-schema fields. The fixture now invokes
+ `render_literal` and edits the recipe source while preserving its profile lock.
+ No production check was weakened. No broad local gate, Kani, Lean or hosted CI
+ ran. This is compilation/pure projection evidence only.
+- Fetch found bot commit `5b9d9ecd`, changing only formal source-closure hashes
+ for the prior ACK/kernel source. It was inspected and retained by fast-forward;
+ no hosted result was inspected.
+- Next requirements are the content plan in the shared journal, response metadata
+ and protected retention, separately authorized disclosure entry/transport/ACK
+ persistence, standalone read-only installation and AP070 metadata predicates/
+ consumer cutover. The existing recipe still includes a mutation declaration;
+ this opt-in content capability is not a claim that standalone read-only
+ installation or the Google domain cutover is complete. A01–A22 remain open.
+
+## Content acquisition journal and transport checkpoint
+
+- Content now has a typed plan in the existing acquisition owner. It binds the
+ exact current connection, resource capacity, request/projection and credential
+ prerequisites. Admission and final grant share the existing native fence and
+ permanent count/byte allocations. Content cannot borrow candidate/recovery
+ admission or ordinary response evidence. Current connection changes refuse.
+- Added engine compilation/acquisition APIs using the existing preparation,
+ custody and acquisition sequence. Terminal replay precedes transport/custody
+ preparation. Public results remain acquisition facts without raw bytes,
+ content commitments or a disclosure permit.
+- The sealed content permit sends the compiler-owned GET and fixed query over
+ the pinned transport. It requests identity encoding and disables automatic
+ decompression; other request paths retain their query prohibition. A distinct
+ response phase privately retains bounded raw bytes, status, version agreement,
+ receipt time and missing/single/invalid media/encoding facts (128 bytes each).
+ It maps to the existing kernel response stage, not a new reducer. Oversized
+ responses and lost permits retain unknown and full charges. Partial read
+ accumulation is zeroized on error.
+- Exact artifact construction passed on `eafd6a2b` plus this source:
+ `AUTHS_UPDATE_FIXTURES=1 cargo test -p auths-gateway --lib -- --exact observed_tests::effect_native_tests::content_tests::native_content_journal_fixture_is_current observed_tests::effect_native_tests::content_tests::native_content_compilation_fixture_is_current`.
+ The final two-test run passed in 2.04 seconds. The new fixture uses installed
+ native entry validation, the file adapter and eight real loopback HTTP calls,
+ checks 16 permanent allocations and seventeenth refusal, one-use grants,
+ reopened replay/unchanged bytes, exact queries and request headers, retained
+ response metadata, record downgrade rejection and public sentinel privacy.
+ Source/connection equality is checked separately. An initial compile failure
+ used a nonexistent policy key method; the test now uses the existing native
+ `GatewayAttemptKey::for_operation`. No production check was weakened.
+- This is not a full engine/CLI process journey, PostgreSQL/fault evidence,
+ content projection provenance, protected content commitment or disclosure.
+ Protected projected-content construction must consume the recorded metadata
+ and reject bad encoding/media/version/projection without refetch. Read-only
+ installation, AP070 metadata and Python-owner removal, maintained consumers,
+ authenticated UI/bindings and P4 evidence remain. No A01–A22 case closes.
+- After the bounded-buffer zeroization change, final generation of
+ `native_content_journal_fixture_is_current` and
+ `semantic_closure::tests::semantic_closure_is_current` passed together (two
+ exact tests, 2.15 seconds). No broad local gate or hosted CI has been run or
+ observed. Bounded decoded response buffers and partial transport accumulation
+ are zeroized on drop. Fetch found no new upstream commits.
+
+## Protected content projection checkpoint
+
+- The previous goal turn made progress (`71fd4f7f`, committed/pushed); this
+ continuation adds protected content to that same response transaction.
+ Attempt `/13` rejects all obsolete `/12` state, without a compatibility reader.
+ A complete content response now retains either protected readiness or the
+ exact typed projection refusal. Neither result grants disclosure authority.
+- The compiler and retained-response paths share one pure projection leaf.
+ Retained version agreement, identity encoding and exact media/projection are
+ enforced; empty content is valid. Encoded/wrong-media/malformed/over-limit
+ responses stay recorded and charged. Public snapshots expose only byte count
+ or stable refusal; response-version/encoding have distinct native codes.
+- Protected content uses the pinned workspace HMAC-SHA256 primitive plus native
+ domain/canonical framing, with a fresh private 32-byte key and exact source,
+ projection, byte count and private content digest. Loads reproject and verify
+ the keyed commitment (or recompute the refusal); replay preserves the key and
+ original record bytes. No plaintext digest, key, commitment or body is public.
+ Existing journal access protection remains required; no encryption or defense
+ against a malicious journal/key administrator is claimed.
+- Exact artifact generation passed on `71fd4f7f` plus this source: content
+ compilation/journal, code inventory, attempt scenarios, prerequisite entry
+ and native entry/fuzz seeds (six exact generators, 9.31 seconds). The initial
+ compile failure used a nonexistent pointer helper; it now invokes the existing
+ response-pointer and escape validators. A schema-edit script initially read
+ ripgrep stdin rather than the tree; it was corrected to an explicit `.` root.
+- The strengthened `native_content_journal_fixture_is_current` generator then
+ passed (one test, 3.30 seconds). Version-2 evidence contains fifteen real HTTP
+ cases: repeated/empty/exact/over-limit/invalid UTF-8 content, identity/gzip/
+ duplicate encodings, wrong/missing media, missing required version, HTTP denial
+ and oversized response, plus lost permit and seventeenth source refusal.
+ It rejects key/tag/length/schema/body/version/media mutations and swapping
+ protected material between independently valid sources containing the same
+ bytes. Reopened replay leaves the private key and complete record unchanged.
+- `cargo xtask arch --update` completed required dependency-snapshot generation;
+ the reviewed delta is only `auths-gateway -> hmac` (no package or other edge
+ changes). Building that generator surfaced existing unused observation/raw-read
+ helpers, now gated to their test/testkit consumers instead of shipping unused
+ production read interfaces. Final semantic-closure generation passed (one exact
+ test, 0.28 seconds). These are artifact construction, not broad local gates or
+ hosted CI evidence. Fetch found no upstream commits; no hosted results were
+ inspected.
+- Next: sealed retained source and native disclosure compilation, then shared
+ disclosure admission/entry, authenticated transport and persistent ACKs. The
+ standalone read-only installation, AP070 metadata/owner cutover, full
+ engine/CLI/UI/binding journey and P4 adapter/fault/differential/formal/performance
+ evidence remain required. No A01–A22 case closes at this checkpoint.
+
+## Sealed source and disclosure compiler checkpoint
+
+- Based on `10a548ee`, added journal-only `RetainedContent` provenance with no
+ public plaintext accessor/deserializer. It requires validated ready content;
+ rejected/unknown responses remain ineligible and no source read is repeated.
+- Native `auths_gateway_disclosure_v1` compilation binds exact source/commitment,
+ projection/length, raw-key recipient, channel and delivery instance. It verifies
+ a separate native proof, uses existing complete private review/exact approval
+ and full fresh recompile, and has no provider credential prerequisites.
+- Added disjoint disclosure count/byte declarations, sixteen lifetime deliveries
+ per source, and existing managed policy arithmetic on the release byte argument.
+ Changing operation/recipient/channel does not create a new lifetime allowance.
+ These remain obligations until disclosure journal admission is implemented.
+- Required artifact construction executed two exact generators: content journal
+ and code inventory. Both passed (3.03 seconds execution after compilation).
+ The journal fixture still performs fifteen real local HTTP reads and adds
+ separate native release/approval/ACK cases for empty and nonempty content.
+ Cases cover managed budgets, wrong budget arguments, old read authority,
+ changed recipient/channel/instance/partition, reused approval, full recompile
+ and expired approval. No content or private correlations enter the artifact.
+- Final content-journal and semantic-closure artifact generators passed together
+ (two tests, 3.06 seconds). Review includes the source resource arguments and
+ projection contract, not just opaque commitments. Preserved upstream generated
+ formal closure commit `622a58fe` by fast-forward before this commit.
+- Disclosure entry, authenticated delivery, persistent ACKs and the full P3/P4
+ cutover remain incomplete. No acceptance case is closed by compilation.
+ No broad local gate, formal execution or hosted CI observation was run.
+
+## Shared source-allocation checkpoint
+
+- Based on pushed disclosure compiler commit `93f4915c`.
+- Extracted the existing source count/byte declaration leaf and canonical
+ append-only allocation row/key/occupied-slot validation into the existing
+ effect/capacity and store/capacity owners. Acquisition source eligibility,
+ lifecycle, schema and charges are unchanged; disclosure declarations now use
+ the same arithmetic with their own units and source keys.
+- Required content-journal and recovery-acquisition artifact generators both
+ passed (two tests, 3.64 seconds); their fixtures remained byte-for-byte current.
+ No new provider calls, admission authority, retry or refund rule is introduced.
+- Candidate, prerequisite and semantic-closure artifact generators also passed
+ (three tests, 37.73 seconds). Behavioral fixture bytes stayed unchanged; only
+ the native semantic closure changed with the source extraction.
+- This reduces duplication needed by disclosure admission; it does not yet
+ implement that admission or a delivery grant. P3/P4 and A01–A22 remain open.
+
+## Disclosure journal checkpoint
+
+- Based on `15d0bb8d`, added a closed disclosure protocol to the shared journal,
+ reducer projection and support listing. Journal `/14` rejects obsolete `/13`;
+ related source, qualification claims and generated schema/fuzz fixtures move
+ together with no migration reader.
+- Native engine proposal/compile/admit APIs now use sealed retained provenance,
+ exact approval and the existing installed authority/configuration fence. Atomic
+ admission reserves the shared permanent source allocation and selected native
+ count/byte slots. It emits no content, credential lease or delivery permit.
+- Durable ACK recording validates exact retained binding/length and the original
+ grant via the common reducer. Exact replay is revision-preserving; concurrent
+ CAS conflicts have one bounded reload. Late history after connection disable
+ remains recordable. The production authenticated grant is still unwired.
+- Required artifact construction passed five exact generators in 9.69 seconds:
+ content journal (including new disclosure journal), code inventory, attempt
+ scenarios, prerequisite compilation and native entry/fuzz corpus. The fixture
+ executes installed file admission, duplicate concurrency, source/connection/
+ approval refusal, sixteen source allocations and independent policy exhaustion,
+ conflicting binding, reopening and no disclosure/public private correlations.
+- ACK persistence tests use explicitly synthetic retained grants, never real
+ recipient authentication or sends. They cover pre-grant/foreign ACK refusal,
+ concurrent identical ACKs, late unknown refinement, immutable revision/charges
+ and reopened replay. No process/PostgreSQL/isolation/complete A08 claim.
+- Final content/disclosure journal and semantic-closure generators passed
+ together (two tests, 4.91 seconds), including ten record mutations and the
+ immediate obsolete schema. The byte-policy case exhausts its sum while count
+ capacity remains; public refusals distinguish it from source exhaustion.
+- Production recipient authentication and final grant/transport, AP070 owner and
+ binding/UI cutovers, complete P4 evidence and hosted CI remain outstanding.
+ No A01–A22 case is closed by this checkpoint; no broad local or hosted gate ran.
+
+## Authenticated disclosure entry and transport checkpoint
+
+- Based on pushed `3f23e5f3`. The preceding status-only turn changed no source;
+ this continuation finishes the pending native grant and Unix transport slice.
+- `UnixDisclosurePolicy` commits exact gateway/recipient UIDs and a private channel
+ identifier. Both peers check OS credentials; the recipient signs a fresh
+ challenge over the full bound delivery statement under an authentication domain
+ distinct from ACK. The owned socket remains sealed through the final grant.
+- Final disclosure entry uses the existing writer fence and sole reducer. It
+ revalidates retained source/MAC, native proof/approval/configuration, permanent
+ source allocation and managed policy membership before persisting the immutable
+ authentication transcript plus entry audit. Only acknowledged persistence
+ returns the one-use send permit. Attempt `/15` rejects obsolete `/14` and
+ granted records lacking valid authentication. Historical decode creates no
+ socket or permit. Public projections remain redacted.
+- `GatewayEngine::disclose` composes admission, fresh recipient authentication,
+ final native entry and exact retained-byte delivery. Recipient code explicitly
+ ACKs after complete receipt or retains native evidence for late recording.
+ Authentication/ACK frames are bounded to 4,096 canonical bytes, content to
+ 65,536 and exact projected length. Five-second handshake/fifteen-second session
+ deadlines supplement, never replace, native fenced authority time.
+- Required artifact construction passed five exact generators in 8.94 seconds:
+ content journal (now also producing `disclosure-delivery.json`), code inventory,
+ attempt scenarios, prerequisite compilation and native entry/fuzz corpus.
+ Extended content/delivery generation then passed in 5.31 seconds. New source
+ evidence uses actual Unix socket pairs and the installed native file fence over
+ a real retained HTTP response. It covers actual receipt/ACK, lost ACK, reopened
+ no-redelivery, late ACK, dropped permit, competing authenticated channels with
+ one winner, and connection/configuration/approval/source changes after
+ authentication. It also refuses wrong UID/key/nonce, malformed/duplicate/
+ unknown/noncanonical/oversized authentication and corrupted retained proofs.
+- The configuration-change fixture initially named a nonexistent store variant
+ and passed a raw key. It now uses `EffectConfiguration` and the native key
+ wrapper; the repaired artifact passed without changing production checks.
+- Final content/disclosure and semantic-closure artifact construction passed
+ together (two exact tests, 5.31 seconds). Fetch found no upstream commits.
+- The earlier synthetic-grant ACK corpus remains explicitly scoped historical
+ unit evidence. New socket/file evidence is not an independent-process crash,
+ PostgreSQL, hostile-agent isolation, full engine/CLI/UI or A07/A08/A22 closure.
+ These, AP070 source-owner removal, standalone read-only installation, maintained
+ consumers and complete P4 evidence remain required. No broad local gate,
+ formal execution or hosted CI observation was run. AP073 remains untouched.
+
+## Native content application, CLI and control-binding checkpoint
+
+- Based on pushed `d5cfeb99`; the preceding goal turn made concrete progress and
+ left a clean worktree. This continuation adds the maintained application path
+ over that native disclosure runtime, not another executor or state owner.
+- Content acquisition now projects resource arguments from the native-verified
+ action. The existing acquisition compiler shares its body below that sealed
+ result. Native fixture parity confirms the same binding as explicit-target
+ compilation; candidate and recovery artifacts remain unchanged.
+- Four closed application envelopes expose content acquisition, disclosure,
+ redacted history and late ACK. Native admission returns history/refusal or Ready;
+ Ready only transfers the owned socket to native authentication and fresh final
+ entry. The transport inherits the application deadline. Control loss cannot
+ recreate a permit. Historical queries check recipe/namespace/partition/epoch;
+ acquisition/disclosure facts are typed, private content stays in its owner.
+- The Rust recipient client checks gateway UID/channel/key before sending private
+ proof/approval. CLI `content-acquire`, `content-history`,
+ `disclosure-acknowledge` and `disclosure-receive` call native operations. Receipt
+ saves new 0600 content and ACK files before sending ACK and reports only
+ `recipient-ack-sent`, never durable server persistence. Review supports content
+ and disclosure targets and can save the exact private recipient binding.
+- Python/TypeScript control methods carry acquisition/history/ACK envelopes and
+ return native bytes without effect hashing, projection, signing, accounting or
+ retry logic. Their fixture and bounds tests are authored, not executed. The
+ recipient protocol remains Rust-owned; full installed binding/UI integration
+ and native call-authoring CLI work remain open.
+- Required native application frame/content compilation/content journal artifact
+ generation passed three tests in 5.89 seconds. The first build caught two missed
+ `DisclosureConnection` deadline destructures; both peers now honor the parent
+ deadline. No production check was weakened to repair compilation.
+- `AUTHS_UPDATE_FIXTURES=1 cargo test -p auths-gateway --test application_content_cli -- --exact native_application_content_cli_fixture_is_current`
+ passed (one generator, 0.89 seconds). It executes the production CLI for three
+ control envelopes against real refusal sockets; it does not exercise installed
+ engine authority, recipient CLI delivery or actual provider I/O.
+- The native socket/file delivery generator now exercises the production recipient
+ client with an explicit control-frame shim, plus real approved native file grants.
+ It is not an installed application-dispatcher claim. Native history is tested
+ against retained snapshots and wrong-recipe refusal. The frame generator covers
+ closed routing, duplicate/obsolete/unknown/override/oversize refusal and writes
+ four maintained fuzz seeds; the fuzz dispatcher is updated for those schemas.
+- Final required artifact generation passed six exact tests in 34.94 seconds:
+ application content frame, content compilation/journal, candidate compilation,
+ recovery acquisition and semantic closure. Candidate/recovery artifacts stayed
+ byte-for-byte unchanged. No Python/TypeScript suite, broad local gate, formal
+ execution or hosted CI observation ran. No A01–A22 case is closed. AP073 remains
+ untouched; full installed process/adapters/UI and AP070 owner cutover remain.
+
+## Native call authoring and complete CLI process composition checkpoint
+
+- Based on `189bb1ea`. The preceding user turn was a status-only response, not
+ implementation progress; this continuation revalidated the clean worktree and
+ performed the next implementation/evidence step. No AP073 files or agents used.
+- Native `content_authorization_call` uses the existing bounded duplicate-free
+ JSON parser and `ContentReadTarget` against the installed connection. Native
+ `disclosure_authorization_call` uses `DisclosureTarget` from retained source
+ provenance. CLI `effect-content-call` and `effect-disclosure-call` save native
+ canonical calls to new absolute 0600 files. No duplicate contract/hash, proof,
+ approval, reservation, custody lease or transport authority is introduced.
+- `content_journey.rs` generates `content-process-journey.json` through actual
+ production CLI/daemon processes, native signed proofs/exact approvals, local
+ file custody/journal and a counting loopback HTTP double. Test-only fixed-key
+ trust authoring explicitly admits a bounded same-service tool set; each signed
+ action/grant retains a single permission and undergoes native verification.
+- The journey acquires content, authenticates a recipient CLI, saves exact bytes
+ and canonical ACK in 0600 files, confirms durable ACK, kills/restarts the daemon
+ and proves history replay produces neither new output files nor provider reads.
+ A second native recipient process receives bytes and retains ACK without sending
+ it; both peers are killed. Reopen reports unknown with one delivery/32 bytes
+ retained. Replay returns history without delivery; exact late ACK refines the
+ same charged journal, and repeated ACK is idempotent.
+- The same daemon then performs an exactly approved write and a separately
+ authorized recovery GET. The provider echo produces native `observed-by-provider`
+ evidence, distinct from state-match evidence and the original HTTP response.
+ Three total provider requests are asserted. Write/recovery replay adds none.
+ A separate operator process disables the connection before another approved
+ disclosure; it refuses with no content/ACK output. Historical ACK remains visible.
+- Public control results exclude content/credential sentinels. Resource authoring
+ rejects duplicate, extra, nonobject and 65,537-byte inputs, accepts the 65,536-byte
+ boundary and preserves existing output files. The native recipient trust channel
+ remains a test fixture, not an authenticated human UI or isolated hostile agent.
+- Artifact generation initially caught fixture API/import mistakes, the query's
+ expected percent encoding, and an incorrect test expectation that echo-linked
+ observation also sets state-match. The final assertions retain the native
+ distinction; production verification/reducer predicates were not weakened.
+- Executed artifact command:
+ `AUTHS_UPDATE_FIXTURES=1 cargo test -p auths-gateway --features testkit-harness,loopback-provider --test content_journey -- --exact native_content_process_journey_fixture_is_current`.
+ Final complete process/crash/resource-boundary generation passed one test in
+ 3.08 seconds (one subprocess helper filtered from the parent run and invoked
+ explicitly as a separate recipient process inside the journey).
+- Required exact native content compilation, candidate compilation, recovery
+ acquisition and semantic-closure generation passed four tests in 32.62 seconds.
+ Content/candidate/recovery fixtures stayed byte-for-byte unchanged; the source
+ closure and embedded digest were refreshed. The final process fixture was
+ generated after the resource limit and output-preservation assertions landed.
+ No local broad build/test/lint/formal suite was run.
+- This strengthens A07/A08/A22 with real file/process composition; no entire
+ acceptance row closes. Authenticated UI, installed Python/TypeScript packages,
+ hostile UID/egress containment, PostgreSQL, remaining crash/race/fault points,
+ full recovery operations, AP070 ownership removal, differential/formal/performance
+ and required hosted evidence remain outstanding. No broad local gate or CI
+ observation was performed.
+
+## Standalone content-only installation checkpoint
+
+- Based on pushed `c80902ab`, the previous process-composition increment. This
+ continuation implements the missing read-only installation needed before AP070
+ can move its managed reads to the native owner.
+- Source `/3` may omit `write` only with the reserved native content tool and a
+ valid required content capability. The compiler rejects mutation observation,
+ echo, pre-entry/preconditions, relative ceiling and spend bounds in that case.
+ Its lock admits binding/resource/account fields only; existing credential and
+ registered-header validation remain authoritative.
+- Lowering now has a closed `Mutation(Box) | ContentOnly` alternative.
+ ContentOnly has no write plan, and mutation construction refuses it before
+ request/custody/transport. Native acquisition, disclosure, candidate credential
+ checks and the journal/reducer are reused. No dummy write URL/body, GET write
+ method, second read executor or provider dispatch was added.
+- Rust review method and mutation recovery APIs now return Option. Content-only
+ review/audit render absent mutation method/recovery rather than an imaginary
+ recorded-write capability, with empty write path and zero body allowance. The
+ content GET declaration remains explicit. All affected Rust/fuzz callers were
+ cut over directly; existing mutation JSON/source identities are unchanged.
+- `native_read_only_recipe_fixture_is_current` writes source/lock/review under
+ `bindings/fixtures/gateway/content-only`, a negative read-only corpus and native
+ fuzz seed. It covers missing content, custom tool, mutation adjuncts and write
+ injection under the reserved read tool, plus absence of a native write plan.
+- `native_read_only_process_journey_fixture_is_current` installs the real daemon
+ with read/release-only native trust, refuses mutation review/submission and
+ mutation recovery from acquired content, then performs exactly one counted
+ provider GET and separate authenticated recipient disclosure/ACK.
+- Final required artifact generation passed six exact native tests in 32.17
+ seconds: read-only source, existing review contract, content/candidate/recovery
+ compilation and semantic closure. Existing mutation review, content compilation,
+ candidate and recovery artifacts remained byte-for-byte unchanged. The final
+ read-only and mixed read/write process generators passed two tests in 3.05
+ seconds; the previous mixed process artifact also remained unchanged. No broad
+ local test/build/lint/formal suite was run.
+- The user's subsequent explicit request to investigate/fix CI authorizes earlier
+ CI triage. A CI subagent works only in an isolated checkout and reports commits
+ for integration after this increment; it does not modify the main worktree.
+ Final required hosted verification on the completed specification remains open.
+- No complete A01–A22 case closes. Google metadata/reference parity, refresh
+ custody, AP070 Python state-owner removal, installed language/UI, hostile
+ containment and complete adapter/fault/formal/performance evidence still remain.
+
+## User-requested CI triage and inventory correction checkpoint
+
+- After the explicit user request to delegate CI failures, an isolated CI agent
+ inspected hosted failures at `c80902ab`. Main implementation continued without
+ shared-file edits. The first isolated fix `991c223d` was reviewed and integrated
+ as `9c939036` after the read-only increment `cfaa8596` was committed.
+- That fix updates the Python gateway export inventory from 17 to 18 (total 190),
+ the TypeScript declaration digest, Python 3.9 runner matrices/artifact names to
+ Ubuntu 24.04, and both qualification family corpus/contract source hashes.
+ Qualification assumptions and declarations were preserved. The agent reports
+ the API snapshot check and six focused qualification tests passing; exact
+ commands/hosted job links are to be recorded with its completed report.
+- Hosted preflight also found the semantic freeze stale. Required artifact
+ construction `cargo xtask semantic-freeze --update` identified nine changed
+ subjects. Freeze 391 and only those subject versions were advanced; regeneration
+ succeeded. The current formal source closure did not change. This refresh is
+ not a formal proof execution or a completed hosted gate.
+- Remaining reported CI roots include stale native/installed tests expecting
+ implicit observation, obsolete raw re-observation simulation calls, and demos
+ lacking exact native effect approval/installation policy. CI fixes must use
+ actual native authorization/recovery or truthful response-only expectations;
+ tests/approvals may not be bypassed. Investigation continues in the isolated
+ checkout, with deeper maintained demo integration still part of AP074.
+
+## AP070 concrete metadata reference checkpoint
+
+- Based on pushed `1ccfc09a`. The native content-only installation is available;
+ the next provider-specific gap is the metadata gate preceding a Drive content
+ GET. Current AP070 Python admission/release still remains to be removed.
+- The Drive reader now calls one pure concrete metadata validator returning a
+ closed document/plain-text representation. Custody, transport, status/media,
+ task checks and state decisions remain outside that validator. Existing request
+ selection uses the typed result. The existing 16 KiB bound is checked again;
+ malformed MIME and excessive decimal conversions return typed refusal.
+- A deterministic domain corpus freezes identity, download restriction, MIME,
+ optional size/trashed semantics, malformed/nonobject inputs and byte boundaries.
+ It explicitly labels historical duplicate/alternate-encoding, nonboolean trash
+ and Unicode decimal acceptance as cases requiring stricter native refusal.
+ Matching a weak historical parser is not a native acceptance requirement.
+- Required artifact construction:
+ `PYTHONPATH=demos/google-employee-delegation/broker/src AUTHS_UPDATE_FIXTURES=1 python3 demos/google-employee-delegation/broker/tests/test_google_metadata_reference.py`
+ passed one generator in 0.002 seconds. It imports only the pure domain contract;
+ no native package, Google credential, provider call or cloud account is used.
+- This is concrete reference evidence ahead of extraction, not a completed native
+ metadata gate or AP070 cutover. Full native/compiler/journal/freshness parity,
+ both supported content paths, token refresh custody, bindings/UI and removal of
+ the competing Python shared state owner remain open. No A01–A22 gate closes.
+
+## Integrated native CI fixes and preserved acceptance scope
+
+- The isolated CI agent's `e705e05c` and `890b42cd` were reviewed and integrated as
+ `4bbff153` and `40662227`. The test harness now compiles real native source
+ effects with a connection descriptor and revalidates them. Its explicitly
+ private fixed-key control plane supplies test approvals to installed SDK
+ consumers; it is not a human approval or production endpoint.
+- Chained-step recovery installs write/recovery trust before entry, keeps replay
+ Unknown without a read, and uses an independent reader proof/exact approval,
+ real reservation/one-use permit and journal refinement. The final-entry fence
+ remains explicitly a unit-test oracle; installed fencing is separate evidence.
+ Ordinary write tests assert ResponseRecorded and no implicit observation.
+- Agent verification at the isolated source commits: 14 file observed tests,
+ 15 TLS PostgreSQL observed/bounds tests (isolated project/port 55449, removed
+ afterward), 19 bounds tests plus 1 expected PostgreSQL ignore, 4 audit tests,
+ packed Node and Python observation consumers, 4 Python public API tests, API
+ snapshot check and 6 qualification tests passed. These are focused hosted-failure
+ reproductions, not a claim that the integrated branch passed final hosted CI.
+- Original hosted roots at `c80902ab`: [preflight freeze](https://github.com/auths-dev/auths-proof/actions/runs/40783379364/job/123151257660),
+ [Python API](https://github.com/auths-dev/auths-proof/actions/runs/40783379255/job/123150389757),
+ [Python runner](https://github.com/auths-dev/auths-proof/actions/runs/40783379255/job/123150389685),
+ [TS API](https://github.com/auths-dev/auths-proof/actions/runs/40783379522/job/123151940534),
+ [installed Node](https://github.com/auths-dev/auths-proof/actions/runs/40783379522/job/123151940374),
+ [PostgreSQL](https://github.com/auths-dev/auths-proof/actions/runs/40783379447/job/123150390095),
+ [qualification](https://github.com/auths-dev/auths-proof/actions/runs/40783379390/job/123152045258),
+ [operator simulation](https://github.com/auths-dev/auths-proof/actions/runs/40783379435/job/123152776374).
+- Maintained Stripe/Airtable demos, north-star and scaffold workflows still need
+ proper effect-policy/exact approval integration. These are managed consumer
+ requirements and remain required. They cannot be waived by retaining AP070.
+- Integrated required artifact generation passed six exact native generators in
+ 35.57 seconds before the small chain-helper extraction; only source closure
+ changed. Final closure-only regeneration after that extraction passed one
+ exact generator in 0.08 seconds. No AP070 migration, test ignore, authorization
+ bypass or completed A01–A22 claim is made.
+- Final semantic-freeze construction after integrating the SDK/harness changes
+ succeeded at freeze 392. Only the three changed identity/public-SDK/public-surface
+ subjects reported by the generator advanced, alongside the freeze version.
+ This refresh does not claim a hosted pass. The CI agent is authorized to push
+ the exact integrated revision non-forcibly to the existing PR branch.
+
+## Independent production-process source-capacity checkpoint
+
+- Based on pushed `6009de25`. The previous user-facing status was not new
+ implementation progress; this continuation revalidated the clean tree and
+ implemented a real cross-process capacity case rather than another in-process
+ store-instance test. AP070 remains independent and unchanged.
+- `native_content_process_capacity_fixture_is_current` starts two production
+ daemons with separate private custody, joined to one native connection and
+ shared file attempt store. Native CLI authoring, real proofs, exact review and
+ approvals prepare 17 operations for the same resource. Fifteen reads consume
+ all but the final source-lifetime allocation; two independently running gateways
+ then contend for it. Exactly one enters the counting HTTP provider.
+- The case asserts 16 actual provider calls, 16 permanent source allocations and
+ 16 durable entry audits; the loser retains a NotEntered native record with
+ `gateway.acquisition.exhausted`, no entry, no allocation and no policy slots.
+ Aggregate retained capacity is 16 calls and 16,384 full-ceiling response bytes.
+ Cross-process history agrees. Killing both daemons and reopening preserves
+ byte-for-byte journal/grant/allocation data; all 17 replays add no provider I/O.
+- Initial construction mistakenly shared one local credential file between both
+ daemons, so transient custody contention could refuse a contender before the
+ admission race. The final fixture uses the supported two-host install/join
+ layout, retaining independent local custody and one shared journal. No custody
+ error was ignored or production authorization check relaxed.
+- Public acquisition snapshots currently expose NotEntered and charges but not
+ its specific refusal code. The test checks that reason in the protected native
+ journal; the missing public reason remains a projection limitation for A10.
+- Required artifact construction:
+ `AUTHS_UPDATE_FIXTURES=1 cargo test -p auths-gateway --features testkit-harness,loopback-provider --test content_journey -- --exact native_content_process_capacity_fixture_is_current`
+ passed one generator in 6.52 seconds after the retained native-record assertions
+ were added. The artifact is `bindings/fixtures/gateway/content-process-capacity.json`.
+- Scope is actual production daemon/CLI/native authority and final file fencing
+ for source-lifetime acquisition count/byte ceilings. It is not managed fixed
+ windows/count/sum, PostgreSQL, hostile UID containment, distributed filesystem
+ fencing or power-loss evidence. A04 remains partial and no A01–A22 gate closes.
+ The PostgreSQL/process/fault cases and CI batch corrections are the next work.
+
+## Native file/PostgreSQL independent-process capacity checkpoint
+
+- The next common case drives the full native GatewayEngine in two separate OS
+ processes over each maintained adapter, rather than substituting the test
+ EntryValidation oracle. Parent and peers use explicit native trust, real signed
+ actions, exact approvals and current installed dependency/configuration checks.
+ The deployment clock and signing keys are explicitly fixed test inputs.
+- `native_process_tests.rs` attaches peers to the same file location or isolated
+ TLS PostgreSQL schema. Each process has separate private persistent custody.
+ After fifteen native acquisitions, ready peers are released through a bounded
+ test-only barrier and contend for the final source-lifetime allocation while a
+ real local HTTP double counts exact requests and credential insertion.
+- Both adapters record exactly sixteen grants and real GETs, sixteen calls and
+ 16,384 full-ceiling response bytes, plus one exhausted not-entered record.
+ Reopening the store/engine and replaying all seventeen operations grants no
+ further provider access. There is no unchecked effect/approval or final-entry
+ authorization oracle in this path.
+- Required artifacts `native-process-capacity-file.json` and
+ `native-process-capacity-postgres.json` were generated by exact native tests
+ with features loopback-provider,testkit-harness. The file generator passed in
+ 2.20 seconds; the ignored PostgreSQL generator passed in 8.54 seconds against an
+ independently started TLS fixture on port 55451. Its schema is disposable and
+ removed on exit. PostgreSQL used certificate verification and synchronous
+ commits. The dedicated containers/volume were stopped and removed after
+ construction. Gateway source-closure generation passed one exact generator in
+ 0.08 seconds after the new module was finalized.
+- This is a privileged test installation of the native engine, not production
+ commissioning/image qualification; the production-CLI file case is separate.
+- PostgreSQL CI now explicitly runs both feature-gated process cases; merely
+ discovering the normal default-feature suite would not execute these tests.
+ This does not claim a hosted result until that workflow passes on the new head.
+- Scope is full native entry checking and real process/adapter/provider boundaries
+ for source-lifetime acquisition count/bytes. Managed fixed-window count/sum,
+ all crash/fault and dependency-race schedules, hostile agent containment and
+ human UI remain open. A04 remains partial; no acceptance row is closed here.
+
+## Holistic CI contract and prerequisite-ownership checkpoint
+
+- Integrated isolated batch `ce308315` as `08f0064c`, directly after the two
+ native process-capacity commits. The batch preserves a shared parent's
+ reservation when a different process owns an unresolved prerequisite read.
+ Missing/in-flight evidence cannot be recorded as a terminal negative provider
+ predicate and suppress that owner's valid write. No retry permit is minted.
+- Deterministic paused-read regressions on file and TLS PostgreSQL retain
+ Reserved for the contender, complete ResponseRecorded for the owner and count
+ exactly one write, three credential reads and one ceiling read, with no
+ recovery read. These use the post-verification reference driver; the separate
+ native process cases still establish full native acquisition checking.
+- The batch repairs all currently inspected preflight/API/scenario/simulation
+ roots together: canonical compliance roles with preserved evidence anchors,
+ Python GatewayReserved snapshot, stale quorum proof fixture, complete review
+ declarations, native application frames, code/source inventories and strict
+ workspace lint. No lint suppression, security assertion or store deadline was
+ weakened. Synthetic qualification reports remain explicitly response-only
+ references and do not establish independently authorized native recovery.
+- In the isolated checkout, actual `xtask ci preflight` passed, including strict
+ all-target/all-feature workspace clippy and real compliance. The gateway run
+ passed 277 behavioral tests with ignored TLS PostgreSQL cases enabled and four
+ test threads; after source-closure regeneration its separate exact check passed
+ 1/1. Five changed CLI targets passed 8/8 tests. Python API inventory passed.
+ An earlier unrestricted local run hit the existing file-lock deadline under
+ I/O saturation; the deadline/assertions were retained, the exact isolated case
+ and bounded-parallel full behavior run passed. This is not a hosted green claim.
+- Remote remains `6009de25` until the next integrated push. Required hosted
+ formal/authoritative/secret/dependency checks still need the final combined
+ source. Early green workflow subsets or skipped aggregate wrappers do not
+ close the goal.
+
+## Maintained demo native approval and explicit recovery checkpoint
+
+- Both managed demo installers now create native exact-effect policies, stage
+ credential-bearing candidates, separately authorize/acquire their native
+ credential checks and publish retained evidence. Additional hosts join the
+ same connection/alias, trust, policy and shared store with separate custody.
+ Installing the software observer explicitly reviews and activates the changed
+ configuration using native revision/candidate CAS tokens.
+- Mutation packets now carry native exact-effect approvals unchanged through
+ installed Python and TypeScript GatewayClient surfaces. The scripted reviewer
+ signs the full native literal review/signing request using an operator-only
+ ephemeral key. Invalid actions that cannot be reviewed are deliberately sent
+ with a different valid effect's approval as a binding-substitution negative
+ case; no placeholder or bypass approval is accepted.
+- Reserved candidate/recovery reads have their own private actor grant and
+ native quorum proofs. A bounded development reviewer socket can authorize only
+ listed fixture sources, at most four explicit read requests per source. The
+ native owner still determines whether the exact read can be compiled,
+ admitted, charged, entered or recorded. Python supplies no outcome reducer.
+- Simulation labels its automatic reviewer honestly. Live test-provider mode
+ requires explicit consent on a separate owner-only FIFO for the entire saved
+ native literal review; the application cannot read/write that channel. A
+ missing/different token or ten-minute expiry fails closed before signing.
+ This is development Docker-operator custody, not an independent-human trial,
+ remote approval inbox or production signing-custody claim. Actual live-provider
+ execution of this channel has not been run.
+- Writes expose ResponseRecorded without implicit read-back. A separate approved
+ native recovery operation returns the unchanged tagged report with original
+ mutation facts, independent read charges and evidence disposition. Exact replay
+ retains current facts without provider I/O. A different challenge under the
+ unchanged installed context is denied before custody; no second installation
+ silently activates different trust to make that action eligible. A Stripe
+ unknown without a retained response locator cannot manufacture an observation.
+- Updated evidence keeps independent exact HTTP mapping/read-back, negative
+ pre-custody checks, budgets, two-process single-write races and process kills.
+ Airtable crash replay now explicitly proves zero reads/writes before a
+ separately authorized recovery read establishes the linked observation. The
+ frontend displays the complete native literal review using textContent and
+ distinguishes response facts from the separate native recovery report.
+- Targeted diagnosis of the known hosted demo failures rebuilt the native
+ loopback binary at `08f0064c`. Same-UID reproductions passed 23 Stripe cases,
+ 34 Airtable cases and two Airtable SIGKILL/reopen cases. Both TypeScript clients
+ compiled and executed their gateway transport. The diagnostic Python native
+ extension and npm build were the CI checkout's c809-era artifacts, with current
+ wrappers; these are not fresh exact-head installed-release evidence. Temporary
+ kit/harness files were outside the repository; no authorization check or native
+ engine was mocked, while OS ownership wiring was adapted to the current UID.
+- A concurrent diagnosis found a truthful interim Unknown projection for a race
+ contender inspecting the winning process after durable entry but before its
+ response persistence. The external counter and protected final journal showed
+ one write and ResponseRecorded. The race tests retain exactly-one-write and
+ a successful response, accept that interim native projection, and additionally
+ require final exact replay to converge without any read/write/custody entry.
+ This does not permit a terminal refusal to poison the parent or an unknown
+ outcome to retry. The prior prerequisite ownership defect still fails it.
+- Owner-only consent mechanics (two exact/substituted-token cases) and malformed/
+ unlisted reviewer requests (three cases returning no proof or approval) passed
+ the same-UID diagnostic with unchanged native counters. They are now authored
+ as explicit demo selftests. Docker UID containment, complete native
+ log/output sentinels, browser execution and newly built installed package
+ evidence remain required in hosted CI. No A01–A22 row is fully closed. The CI
+ agent continues the separate example/scaffold cutover while core fault,
+ fixed-window capacity, dependency-race, differential/formal/performance gates
+ remain open. AP070/AP073 were not changed.
+
+- Semantic freeze 394 records the updated managed demo/compliance contract;
+ identity protocol 140, public SDK contract 137 and public surface 394 were
+ explicitly advanced. Generated source/formal identities were not reclassified
+ as proved by these inventory updates.
+
+## SDK example and generated-scaffold cutover checkpoint
+
+- Integrated hardened CI batch `a13e960f` as `8ab0136d`. Python, packed npm and
+ generated scaffold consumers now use native staged candidate acquisition and
+ publication, exact effect approvals, explicit configuration activation and
+ separately authorized recovery. Mutation HTTP response and the unchanged
+ native recovery report remain separate; positive recovery asserts the native
+ original source's provider link and observed stage, not a synthesized success.
+- A separately installed Python operator SDK authors reserved read proofs for
+ the TypeScript application. This is declared deployment tooling; the actual
+ application still uses installed npm GatewayClient, without claiming native
+ TypeScript read-proof authoring. One authored operator transport helper is
+ copied byte-for-byte into the scaffold, with a direct equality gate. It does
+ not interpret provider fields or decide native admission/recovery outcomes.
+- The operator channel accepts only closed bounded requests, canonical proof/
+ action base64url and source tokens, rejects duplicate/malformed JSON, applies
+ a five-second whole-frame deadline and sixty-second native command deadline,
+ and returns redacted typed refusals. Malformed or disconnected peers cannot
+ terminate the listener or consume its valid-review allowance. Native authority,
+ capacity and one-use entry remain enforced by the gateway, independently of
+ this finite development review channel.
+- Live exact signing requires separate owner-only FIFO consent after reading the
+ entire native literal-review.txt and signing-request.json. Simulation alone
+ uses the explicitly labeled finite scripted reviewer. Incomplete/unsupported
+ native reviews and declined live consent cannot reach the signer.
+- Actual isolated reproductions passed the Python journey (19.55 seconds),
+ packed npm journey (24.25 seconds), generated project's full native selftest,
+ and thirty scaffold/operator tests. The latter include real malformed Unix
+ peers followed by a healthy request, finite deadlines and refusal before any
+ signature. A real positive recovery packet was also authored/reviewed through
+ the hardened transport. These used current wrappers and repaired native gateway
+ source but c809-era local Python/WASM artifacts; fresh exact-head installed
+ artifacts, live-human behavior and Docker isolation remain hosted obligations.
+- Existing hostile manager/quorum/budget/provider cases, offline audit/tamper,
+ source-owned echo checks, idempotency state-loss demonstration, private key modes
+ and second-install zero-provider-I/O assertions were preserved. No AP070 source
+ or paused AP073 state was changed. These CI repairs do not close A01–A22.
+
+## Native acquisition process-crash checkpoint
+
+- `native_crash_tests.rs` executes the full native content acquisition engine in
+ separately spawned OS processes over each maintained adapter. It reuses native
+ call authoring, proof verification, exact approval, installed dependency fencing
+ and separate local custody. Fixed public keys, fixed deployment time and a
+ privileged test installation remain explicit assumptions; no EntryValidation
+ oracle replaces the gateway's checks.
+- A test-only transparent store proxy pauses unchanged native decisions under
+ the actual adapter fence, or pauses before forwarding actual durability
+ acknowledgement. The proxy never selects a transition, rewrites a record or
+ constructs a permit. Production builds contain no selectable failpoint.
+- Both file and TLS PostgreSQL exercise seven SIGKILL schedules: before
+ reservation commit; after reservation commit before acknowledgement; after
+ acknowledged reservation but before entry commit; after entry commit before
+ its acknowledgement/transport; while a real HTTP response is partial; after
+ complete response receipt before persistence; and after response commit before
+ return to the caller. The receiving process is confirmed killed by signal 9;
+ the separate HTTP double records actual requests. No caller result was written.
+- Reopening checks the protected native phase, durable grant and original
+ allocation, plus native public history/charge projections. Absence and reserved
+ state can resume only through current native proof/approval/entry checks, with
+ one original allocation and one GET. A lost grant remains charged Unknown
+ even when no HTTP request was sent, and cannot lease or fetch again. Lost
+ response persistence remains Unknown without reconstructible content. Durable
+ response persistence retains ready content despite caller-response loss.
+- Every terminal replay preserves private journal bytes, native boundary counters
+ and public snapshot. Each allocated invocation retains one call and the full
+ 1,024-byte ceiling, including unknown; it is not refunded to actual response
+ size. There is no automatic refetch, redelivery, observation or write.
+- Required artifact generators `native_file_process_crash_fixture_is_current`
+ and `postgres_native_process_crash_fixture_is_current` generated reproducible
+ seven-case files under bindings/fixtures/gateway. Initial exact runs passed in
+ 1.45 and 2.66 seconds respectively; final assertions additionally check the
+ native recovered stage, no post-grant custody lease and retained-content
+ projection. PostgreSQL used an independent canonical TLS fixture on port55451,
+ certificate/server-name verification and synchronous commits; schema isolation
+ and cleanup are provided by TestAttempts. The dedicated Docker containers,
+ network and volume were stopped/removed after final generation. Both explicit feature-gated commands
+ were added to PostgreSQL CI so normal default-feature discovery cannot omit them.
+- Final regeneration after the complete native assertions passed both adapter
+ generators in one invocation (two adapter tests and the no-input peer helper,
+ all three passed in 3.96 seconds). The source-closure generator passed in 0.06 seconds before rebuilding
+ that final test executable; its embedded-closure verification then passed
+ 1/1 in 0.05 seconds. The known strict gateway lint reproduction passed
+ without suppressions after adding the instrumentation. Source closure and release metadata must match
+ the committed final source before publication. These are native acquisition
+ transaction-boundary/process-crash results, not injected sync/rename failures,
+ partial-batch fault schedules, ambiguous database commit delivery, power-loss,
+ hostile-process isolation or all mutation/disclosure crash boundaries. A05
+ remains partial and A06 remains open; no acceptance invariant was weakened.
+
+- Integrated release freeze 395 advances identity protocol 141, portable ABI
+ bindings 125, public SDK contract 138 and public surface 395 for the combined
+ source/inventory changes. It does not certify new formal or provider claims.
+
+## Combined publication observation
+
+- Pushed non-force to origin/codex/effect-recovery-foundation at exact head
+ `af01d3769a2d6a1b90820a506a2f797487ebbe4a`; GitHub confirms draft PR213
+ remains open at that head: https://github.com/auths-dev/auths-proof/pull/213.
+ The push includes native capacity/crash evidence and both substantial CI repair
+ batches, demos, examples/scaffold, inventories and regenerated source identities.
+- Exact-head hosted runs were confirmed live: main run 40799196205,
+ Python run 40799196247, TypeScript run 40799196258, TLS PostgreSQL run 40799196262,
+ SDK recipes run 40799196408, qualification run 40799196229,
+ operator package run 40799196193, Airtable run 40799196227, Stripe run 40799196249 and
+ isolation run 40799196259. Pending jobs are not passes; skipped custody/live-provider
+ workflows do not establish provider qualification. The CI agent continues
+ holistic triage with backoff and must collect all actual roots before a batch.
+- The PR description reports this work as an incomplete foundation and retains
+ every remaining acceptance/formal/installed/operational gate. No merge or
+ deployment was requested or performed. Subsequent implementation/checkpoint
+ commits need their own final exact-head verification before goal closure.
+
+## Native dependency ordering and expiry checkpoint
+
+- `native_dependency_tests.rs` adds independent native worker/operator OS
+ processes over the maintained file and TLS PostgreSQL adapters. Both handles
+ finish installation/review before contention begins. A transparent test-only
+ store wrapper schedules the unchanged native entry decision against actual
+ adapter fences; it cannot construct a permit or select a native transition.
+- Connection revocation and exact configuration activation each run in both
+ orders: a dependency change committed first prevents a grant/provider call;
+ an entry decision ordered first may exercise its one grant. The grant fence
+ explicitly contains the connection and effect-configuration records, each
+ physically distinct from the attempted effect. The competing writer is a
+ separately spawned operator using the maintained connection CAS or reviewed
+ configuration activation, not a fabricated record snapshot.
+- Four additional schedules hold the real connection writer fence while the
+ worker waits to grant entry. The maintained test clock advances without
+ modifying signed proof/action/approval bytes. Exact human approval and native
+ resource-authority expiry are tested independently, each with a still-valid
+ positive control. Expiry blocks entry; the positive controls record the new
+ grant time after waiting. Admission time remains the original signed/event
+ time. Existing allocations and the full call/byte charge remain unchanged.
+- Native public history is inspected through a reopened handle and must agree
+ with the actual result while preserving private journal bytes. Real HTTP
+ request counts distinguish admitted entry from blocked attempts. These cases
+ do not establish observer-evidence aging, threshold-selection changes,
+ mutation/disclosure dependency races, fixed-window budget contention,
+ power-loss or hostile-process isolation. A12/A13 remain partial.
+
+- Required fixture generation passed the two adapter cases plus the no-input
+ peer helper (three tests, 3.58 seconds). After regenerating the integrated
+ gateway source closure, a rebuilt executable verified both dependency files,
+ both existing process-capacity files, both seven-case crash files, all three
+ no-input peer helpers and the closure without update mode: ten tests passed
+ in 7.62 seconds. The shared-helper default authoring remains byte-identical
+ to the previous capacity/crash fixtures. Strict gateway library/test clippy
+ passed without new suppressions. The dedicated canonical TLS PostgreSQL
+ fixture used port55451, verified CA/server name and synchronous commits.
+ The dedicated containers, network and volume were removed after verification.
+ These are local artifact-construction/regression results; fresh hosted
+ verification remains required.
+
+## Additional holistic CI repairs
+
+- Integrated isolated repair `ae3154f3` as `050fa219`: patched dependency
+ requirements/lockfile and corresponding vetted crypto API callers, preserved
+ high-S rejection and all canonical signature fixtures, regenerated the SDK
+ baseline for the existing GatewayReserved export, corrected the synthetic
+ credential marker and documented fingerprint-specific historical public-data
+ exceptions. No broad secret-scan allowlist or dependency advisory exception
+ was added.
+- Hosted macOS setup initially failed DNS; its one exact-head job retry exposed
+ a separate artifact-handoff bug. Consumers now download immutable IDs emitted
+ by their actual producers, including unique per-platform matrix outputs.
+ Missing producer output fails closed. No consumer guesses its producer's
+ attempt number, and no head, platform, source-free or live approval requirement
+ is weakened. The repair spans Python, TypeScript, qualification and release
+ workflows as one batch.
+- Agent-local reproductions passed strict workspace check/clippy, dependency
+ policy, 1,597 unchanged wire vectors, conformance, 68 crypto/custody tests,
+ staged/full-history secret scans and workflow lint. Those results diagnose
+ the prior hosted failures; they are not successful hosted evidence for the
+ combined new head. The previous head's formal translation subsequently failed
+ and is under separate investigation before the next combined publication.
+ Its inventory and natural model still describe the superseded Attempting
+ state. The repair must model current protocol/revision/entry/ACK facts,
+ distinguish late source-bound response/recipient evidence from linked provider
+ observation, and retain strict provenance and no-blind-retry invariants.
+- The existing AP070 reference and paused AP073 worktree remain untouched.
+
+- Release freeze 396 advances only the generator-reported changed subjects:
+ formal source closure104, identity protocol142, external custody6, public SDK
+ contract139 and public surface396. The gateway semantic closure includes the
+ new tests and integrated crypto/custody/lockfile sources. These identities
+ record reviewed source changes; they do not establish completed formal proofs
+ or close an acceptance gate.
+
+## Native fixed-window capacity checkpoint
+
+- `native_budget_tests.rs` adds three full native acquisition histories per
+ adapter: count-only exhaustion, response-byte sum exhaustion with count
+ headroom, and simultaneous count/byte limits. A real bounded extension is
+ signed into each resource grant and passes the ordinary native compiler and
+ entry checks; exact human approval remains a separate signed fact. This
+ test-only authoring extension constructs no admission fact or entry permit.
+- After one prefix request reserves a count slot and the entire 1,024-byte
+ response ceiling, two independent worker OS processes with separate local
+ custody contend for the remaining allowance in the same fixed 3,600-second
+ window. Exactly one receives a grant and sends HTTP. The other retains a
+ not-entered record with the precise window/sum refusal, no allocation, no
+ counter slot and zero call/byte charge. The successful pair retains two
+ call charges and 2,048 reserved bytes despite only 21 response bytes each.
+- Assertions load the native count/sum slots through their recorded counter
+ witnesses and verify the original window index/end, operation ownership and
+ cumulative byte ceiling. No third slot exists. A reopened native engine
+ returns all original snapshots and preserves every inspected attempt/counter
+ byte, execution witness and provider count. These cases use actual adapter
+ transactions and local HTTP, with fixed public test keys/clock; they do not
+ establish mutation/disclosure fixed-window contention, cross-window expiry,
+ clock rollback, partial-batch/durability faults, human review or isolation.
+- Initial artifact generation passed the file matrix (1.28 seconds) and TLS
+ PostgreSQL matrix (2.60 seconds), each containing three histories. The strict
+ gateway library/test lint reproduction passed after the complete test helper
+ factoring, without suppressions. Exact feature-gated CI commands and consumer
+ evidence anchors include both adapters. A04 remains partial.
+- Final source-closure generation and rebuild verified the two new budget
+ artifacts together with both native capacity, crash and dependency matrices,
+ the three no-input peer helpers and the closure without update mode: all
+ twelve tests passed in 4.69 seconds. Existing unbounded authoring/capacity/
+ crash/dependency artifacts stayed byte-identical after the bounded-grant helper
+ factoring. The dedicated TLS fixture's containers, network and volume were
+ stopped/removed afterward. Fresh exact-head hosted verification remains open.
+
+- Release freeze397 advances the reported identity protocol143, public SDK
+ contract140 and public surface397 subjects for this bounded-authoring/test
+ inventory update. No portable wire change, new formal proof or completed
+ acceptance claim follows from these source identity updates.
+
+## Native file I/O failure checkpoint
+
+- `FileGatewayAttemptStore` has private `cfg(test)` failure state and hooks at
+ its actual data-sync, rollback-journal installation, batch-target installation,
+ replacement and directory-sync call sites. Production builds contain no
+ selector/state, environment switch or selectable failure behavior. Native
+ record decisions and every preceding successful filesystem operation are
+ unchanged; the selected failing I/O return is simulated rather than produced
+ by an actual faulty kernel/device.
+- `native_file_fault_tests.rs` authors fifteen full native acquisition schedules
+ in independent worker OS processes, using signed bounded resource grants,
+ separate exact approvals/custody and real local HTTP. A transparent scheduling
+ wrapper arms the adapter only after the unmodified native decision under the
+ real writer fence. It cannot manufacture a reservation or entry permit.
+- The admission matrix covers sync/install of the rollback journal, each of
+ four target-install prefixes (attempt, lifetime allocation, count, byte sum),
+ and each of three directory acknowledgements. Entry and response each exercise
+ data sync, replacement and directory acknowledgement. Authoritative reopen,
+ native history, all-or-none charges, immutable signed inputs, actual provider
+ requests and execution witnesses distinguish rollback, resumable reservation,
+ charged lost-grant unknown and retained response evidence.
+- Authored assertions require a partial batch to leave no attempt/allocation/
+ count/sum targets; a committed reservation may progress only through native
+ fresh checks; a visible grant after failed acknowledgement cannot produce a
+ new permit. Unknown outcomes remain charged even when no HTTP request was
+ made. Terminal replay preserves every journal JSON byte and performs no new
+ lease or HTTP request. Required artifact generation passed all fifteen
+ schedules in 3.37 seconds. The first run encountered the local HTTP double's
+ existing two-second header deadline on a heavily contended host and correctly
+ projected Unknown; its positive-resume assertion failed. The test executor
+ pool is now bounded to two threads per process instead of multiplying the
+ host's CPU count. No native deadline, assertion or security invariant changed.
+ Strict gateway library/test lint passed without suppressions. After source
+ closure regeneration and rebuild, fourteen selected native artifact/helper/
+ closure tests passed without update mode in 6.37 seconds. The two adapter
+ lifetime/fixed-window capacity, crash and dependency matrices stayed byte-
+ identical with the bounded test executor. The dedicated TLS fixture's
+ containers/network/volume were stopped and removed after verification.
+- These schedules do not establish actual device EIO/power loss, PostgreSQL
+ partial-batch/ambiguous-commit delivery or the complete mutation/disclosure
+ fault matrices. No acceptance condition is waived, and A06 remains incomplete.
+
+- Release freeze398 advances identity protocol144, public SDK contract141 and
+ public surface398 for the generator-reported source/inventory changes. The
+ regenerated gateway closure includes the fault instrumentation and matrix.
+ Hosted exact-head verification and the complete A06 gate remain required.
+
+## Native PostgreSQL partial-admission checkpoint
+
+- `native_postgres_fault_tests.rs` exercises four real server-side failure
+ histories through independent native gateway worker processes and the
+ maintained TLS PostgreSQL adapter. A trigger in each locally owned, isolated
+ test schema raises an exception on insert1,2,3 or4 of the actual four-record
+ admission batch. No shipping adapter fault switch, forged decision or alternate
+ authorization path is introduced. All source/proof/approval/custody and local
+ HTTP composition is the same native path as the capacity cases.
+- A PostgreSQL sequence survives transaction rollback and witnesses exactly the
+ rejected insert number. Thus each schedule proves the reached server row
+ prefix, including earlier successful inserts, rather than merely returning a
+ synthetic adapter error. After the exception, native history is unavailable
+ and direct server counts show no attempt, lifetime allocation, count slot or
+ byte-sum slot. The child execution witness and independent HTTP counter both
+ show no custody lease or external entry.
+- The trigger is removed only after the failed worker exits. A reopened native
+ engine revalidates the unchanged signed inputs under the same logical operation
+ and produces one response, one lifetime allocation, one count slot and one
+ byte-sum slot. Exact terminal replay preserves the attempt bytes and execution
+ witness, with no second HTTP request or charge. Temporary trigger/sequence
+ objects remain confined to the owned schema that the fixture drops.
+- Required artifact generation passed four histories in 4.72 seconds. Exact
+ feature-gated PostgreSQL CI invocation and consumer evidence anchors include
+ the new matrix. Strict gateway library/test lint passed without suppressions.
+ After closure regeneration and rebuild, all fifteen selected native artifact,
+ helper and closure cases passed without update mode in 6.15 seconds. Existing
+ capacity, crash, dependency, fixed-window and file-fault artifacts remained
+ byte-identical after adding the private peer execution witness. The dedicated
+ TLS containers/network/volume were stopped and removed after verification.
+ This is transaction-abort/partial-batch evidence, not ambiguous commit delivery,
+ server/power failure or the complete mutation/disclosure fault matrix. A06
+ remains partial and no acceptance requirement is weakened.
+
+- Release freeze399 advances the reported identity protocol145, public SDK
+ contract142 and public surface399 subjects. Fresh hosted exact-head evidence
+ remains required; source-identity counters are not formal acceptance evidence.
+
+## Formal translation and recovery correspondence repair
+
+- Integrated isolated formal repair `ffa982fa` as `884b5bd9`. The hosted
+ translation failure on `af01d376` exposed an obsolete 123-function inventory
+ and a model still using Attempting/implicit read-back. The shipping kernel
+ already has the direct Reserved/EntryGranted cutover; the model, generated
+ translation, exact refinement and assurance inventory now describe that
+ implementation together. This repair changes no shipping runtime behavior.
+- Reviewed inventory126 adds six closed protocol/grant/ACK checks and removes
+ three obsolete mode/implicit-read helpers. External and opaque local functions
+ remain zero. The reviewed qualification boundary digest was explicitly renewed
+ for that inventory change; tool pins, warnings, four existing external models,
+ axiom policy and CFG parity requirements remain intact.
+- The natural attempt model includes three protocols, JSON-safe bounded revision,
+ immutable grant and ACK evidence. Exact translated refinement covers every pair
+ of views including revision overflow. Separate late-response and late-recipient-
+ ACK theorems retain source/grant lineage without asserting provider observation;
+ the complete unknown-exit theorem retains strict linked-evidence requirements
+ for ObservedByProvider. Submission ordering proves durable grant before send,
+ one send at most, no post-send NotEntered/refund, no implicit response read,
+ and stored replay without a new send. Runtime signature/challenge/provenance,
+ filesystem/database atomicity and physical I/O remain separate obligations.
+- Isolated source reproduction passed both authored models, exact refinement,
+ full pinned Lean build (3,305 jobs), all eight qualification case modules,
+ two clean byte-identical translations, and a compiled transitive axiom audit
+ of329 statements: seven added, seven updated,315 unchanged, none removed.
+ Only the existing propext/Quot.sound/Classical.choice foundational axioms occur.
+ The boundary-pin unit test passed1/1 and the staged v8.28 secret scan passed.
+- Final non-update `cargo xtask formal --skip-kani` passed existing qualification,
+ 329-claim audit, byte-stable semantic vectors,4/4 Rust refinement tests and the
+ 49/49 Kani harness inventory. Bounded Kani was deliberately not duplicated
+ locally; the prior published head's hosted Kani job passed, and the next exact
+ head still requires hosted verification. Mutation campaign snapshots remain
+ explicitly stale pre-release evidence rather than newly passed campaigns.
+- The integration's sole generated source-closure conflict was resolved using
+ the source generator. Its output exactly equals the reviewed formal batch;
+ no digest was hand-edited. Main's independently generated gateway closure and
+ native fault/concurrency fixtures remain separate evidence. A20 and the full
+ AP074 acceptance/performance gates remain incomplete.
+
+- On the combined checkout, the required proof-coverage generator reproduced
+ study006 byte-for-byte and emitted unchanged current coverage artifacts
+ (110/1,031 decision-scope functions and1,289/11,353 code lines). The added native
+ test/fault scaffolding is not advertised as proved decision code. Release
+ freeze400 advances only the changed formal assurance-manifest61, generated
+ translations20, qualification17, formal source-closure105 and public-surface400
+ subjects. Identity protocol145 and public SDK142 remain unchanged by this
+ formal-only integration. No full acceptance gate is declared closed.
+
+## Native PostgreSQL COMMIT-reply loss checkpoint
+
+- `native_postgres_commit_tests.rs` and its private opaque TCP proxy author six
+ full native acquisition histories: actual commit success versus actual backend
+ abort during admission, entry-grant persistence and response persistence.
+ A schema-local deferred constraint trigger pauses the native COMMIT at a
+ separate advisory lock. An independent direct TLS client observes the waiting
+ backend through PostgreSQL locks and asserts that its pg_stat_ssl session is
+ encrypted. The test proxy never terminates TLS, parses SQL or fabricates a
+ store result; trust roots, server-name checks and native store deadlines remain
+ the maintained implementation's checks.
+- For committed transactions, the independent observer releases the trigger,
+ reads the newly committed exact operation phase and confirms that server reply
+ bytes are withheld before disconnecting the worker's transport. For aborted
+ transactions, the observer terminates only the fixture backend identified by
+ that owned-schema barrier, verifies backend exit/rollback, then disconnects.
+ The real adapter returns an error at the selected native decision exactly once.
+ Proxy forwarding resumes for fresh pool connections; unavailable immediate
+ caller projection remains a typed indeterminate result rather than authority.
+- Native error/custody/HTTP witnesses, unchanged signed inputs and independent
+ server counts verify all-or-none charges. An aborted admission leaves no claim
+ or allocation/counter rows. A committed admission or aborted grant retains a
+ charged reservation and can continue only through native fresh checks. A
+ committed grant with a lost acknowledgement stays charged Unknown even with
+ zero actual HTTP reads; it can never lease/fetch again. An aborted response
+ persistence retains charged Unknown after one read; a committed response
+ remains recorded despite reply loss. Native reopening/replay preserves every
+ inspected gateway-row byte and execution witness, with one permanent lifetime
+ allocation and one count/sum slot, no refund, no extra HTTP and no new permit.
+- Initial required artifact generation passed all six real transaction/transport
+ histories in2.25 seconds. Strengthened independent TLS and pre-reopen row-count
+ assertions passed in2.36 seconds; strict gateway library/test lint passed with
+ no suppressions. After closure regeneration and rebuild, all seventeen
+ selected native artifact/helper/closure tests passed without update mode in
+ 6.92 seconds, including both adapters' previous capacity/crash/dependency
+ matrices. Their files stayed byte-identical. Dedicated fixture containers,
+ network and volume were stopped/removed after verification.
+ The explicit feature-gated PostgreSQL CI command and consumer anchor name this
+ matrix. No shipping adapter/kernel behavior or fault switch is introduced.
+- These cases cover acquisition COMMIT reply delivery over an actual TLS
+ connection, with committed and aborted database outcomes. They do not claim
+ PostgreSQL power-loss/whole-store rollback protection, hostile OS containment,
+ live qualification or the full mutation/disclosure fault matrix. A06 remains
+ partial until those required protocol-specific schedules are complete.
+
+## Published ab36 hosted observation
+
+- The combined security/artifact, native fault/concurrency, formal correspondence
+ and release-identity changes were pushed non-force as
+ `ab36e4b76420a37912d00b2c5b1fac1ad6c556c5` to the open draft PR213.
+ At the first observed checkpoint all12 non-main/non-live workflows passed;
+ the main CI workflow is still running with76 successful checks,8 conditional
+ skips and no failures. Pending checks are not passes. Custody-live is skipped
+ and provides no qualification evidence.
+- The [exact-head PostgreSQL job](https://github.com/auths-dev/auths-proof/actions/runs/40867724293/job/123426622975) passed with actual execution of all
+ eleven explicit feature-gated native cases (each1passed/0ignored), the thirty
+ ordinary ignored PostgreSQL cases and physical/WAL restore assertions. Fresh
+ installed Docker Stripe/Airtable demos passed23/36 cases respectively, each
+ explicitly production_qualification:false. Installed Python/TypeScript platform
+ harnesses, SDK recipes/scaffold and independent Google images passed. Distinct
+ UID isolation executed and passed2/2 cases. These are evidence on ab36, not
+ evidence for the newer unpublished COMMIT-loss source.
+- Subsequently, [authoritative implementation](https://github.com/auths-dev/auths-proof/actions/runs/40867724155/job/123431724494)
+ failed a CLI frame fixture that still omitted required exact approval. The CI
+ agent found two stale fixtures, including vacuous commissioning-field negative
+ cases. Its isolated repair uses the current canonical native submission and
+ requires approval to decode before forbidden-field checks; binary tests passed
+ 9/9. The full failing workspace command is being reproduced with no-fail-fast
+ to collect downstream roots before one batch. Later compliance and Kani jobs
+ passed; translation and its downstream proof/evidence remain pending. These
+ local CLI repairs are not yet published-head passes.
+- The existing CI agent continues holistic exact-head triage. No new agent,
+ AP070 migration, AP073 modification, live action, merge or deployment occurred.
+
+- Release freeze401 advances the reported identity protocol146, public SDK
+ contract143 and public surface401 subjects for the new fixture/consumer
+ inventory. The shipping store and translated kernel are unchanged. Fresh
+ exact-head hosted execution is required before acceptance claims.
+
+## Integrated canonical application controls and CI attribution
+
+- Reviewed isolated CI batch31969a61 and integrated it as d7eea272. Two binary
+ negative-frame tests now begin with the canonical native submission fixture
+ and assert that exact approval actually decodes. The commissioning-field
+ negatives are no longer vacuous, and every forbidden provider/authority field
+ remains rejected. The socket-path tests supply real fixture proof/action/
+ approval so clap parsing cannot mask the intended socket failures.
+- Qualification's disposable synthetic source commit disables inherited Git
+ signing only on that single command. No user/repository setting, real commit
+ signing, qualification signer or signing acceptance gate is changed. Focused
+ isolated checks passed binary9/9, socket13/13 and pipeline2/2.
+- The actual failed workspace command was reproduced with no-fail-fast to expose
+ downstream issues before publishing another batch. Its305 result suites
+ reported1,503 passed,4 failed and47 ignored. Failures were the now-repaired
+ socket caller and synthetic signing environment, expected gateway closure drift
+ after binary source edits, and one typed Unavailable under broad-run file-store
+ contention. The original binary root had already been repaired before that run.
+ The identical compiled conformance test subsequently passed1/1 in8.82 seconds
+ with unchanged two-second lock/fence bounds and unchanged exact-once assertions.
+ This is not advertised as a successful full-workspace run or a store fix.
+- Published ab36's pinned translation job123431724418 passed two clean identical
+ translations. Authoritative Lean job123448869608 is still running and final
+ formal evidence follows it; no completed qualification is claimed yet. Hosted
+ compliance and all49 Kani harnesses passed, with zero Kani failures. Fresh
+ Python tests passed428 with11 skips, and native/WASM live-demo parity passed.
+- Producer attribution is retained precisely: standalone Python/TypeScript and
+ default-checkout CI/isolation jobs built the recorded PR merge revision
+ f25217f68ab1b69930660a29a61e5433cbd84210. Demo producers explicitly built ab36.
+ Both revisions have the identical verified tree
+ dc691ccb384709520202274dd2db7904841048a6, confirmed by Git tree inspection and
+ an empty diff. Their behavior covers the published source tree; their producer
+ revision is not relabeled as ab36. Future final-head evidence must retain its
+ actual source/producer attribution too.
+
+## Retained replay correction and journal cutover
+
+- The native real-HTTP mutation regression reproduced the source-audited bug:
+ an original PATCH completed with HTTP 200, but its identical submission after
+ exact approval expiry returned `NotEntered gateway.effect.approval`. Stored
+ facts themselves remained retained. That projection did not satisfy §5.3.
+- `effect::NativeAuthority` now retains the bounded canonical native authority
+ projection that reconstructs the binding's existing normalized commitment.
+ It includes exact native proof/action digests, validates the portable
+ self-digest, stays private and is limited to 65,536 bytes. No new effect identity
+ algorithm or alternative verifier is introduced. The common journal validates
+ it against the binding and includes it in immutable fields for all protocols.
+- The native journal cuts over directly from `/15` to `/16`. Qualification
+ candidates, fixture generators, decoder fuzz seeds, compliance and the runbook
+ use the new schema; obsolete disposable records fail closed. There is no
+ migration/compatibility reader, runtime selector or second authority cache.
+- One native history matcher checks the partition/epoch, namespace, operation,
+ recipe, both committed inputs and the retained binding's exact approver
+ signature. Its historical signature check grants no current entry authority.
+ Mutation, content, recovery, candidate-check and disclosure entry APIs use
+ typed stored projections; additional resource/source/recipient inputs still
+ have to match the original signed action. Terminal replay requires no current
+ approval window, connection, credential lease or authenticated delivery.
+- Real native binding/signing fixtures stayed byte-identical under the retained
+ commitment algorithm. Post-verification reference commands now carry explicit
+ canonical oracle authority evidence instead of an all-zero digest; only those
+ synthetic IDs/signatures regenerate. No production path can use that oracle.
+- Historical approval matching verifies the original signature and binding
+ without sampling or inventing an evaluation time. Current-entry approval
+ verification still requires the signed window at the real sampled time.
+- A reserved operation can only continue through full current native compilation,
+ exact approval and the existing atomic writer fence. Current authorization
+ failure returns the unchanged retained reservation. A history value cannot
+ construct a sealed effect, approved effect, live channel or lost permit.
+- The four required fixture generators passed 4/4, zero ignored, in 4.90 seconds
+ against the real file store and independent TLS PostgreSQL fixture on port 55451.
+ They author 40 histories: ten mutation histories and thirty acquisition/disclosure
+ histories. Mutation checks run in the original and reopened native engines,
+ rejecting different valid proof/action inputs and a damaged approval. Read and
+ release histories reopen the native engine, preserve exact response/ACK/unknown
+ facts after expiry/revocation/configuration replacement, and reject changed
+ resource/recipient proposals. Actual provider counts, native execution witnesses
+ and journal bytes remain unchanged through replay. The release cases use real
+ authenticated Unix sockets and both exact ACK and lost-ACK outcomes.
+- The older required artifact generators passed 31/31, zero ignored, in 83.06 seconds
+ after the cutover, including both adapters' capacity, SIGKILL, dependency-order,
+ window-budget, real file I/O faults and real PostgreSQL failure/COMMIT reply-loss
+ matrices, plus native candidate/content/recovery/prerequisite/admission fixtures.
+ This is artifact-construction evidence, not a new complete hosted gate.
+- Assumptions remain explicit: fixed public test keys, maintained test clock,
+ local provider transport, same-UID recipient tests and trusted durable stores.
+ No human-presence, hostile-host, whole-store rollback, live-provider or managed
+ Google guarantee follows. Full A01–A22 acceptance and exact new-head CI remain
+ open. AP070 stays independent; AP073 remains untouched.
+
+- Release freeze 403 advances the generator-reported identity protocol 147,
+ public SDK contract 144 and public surface 403 subjects. The gateway source
+ closure is regenerated. No kernel source, proof statement, tool pin, security
+ deadline or test assertion was relaxed. The additional retained-header test and
+ regenerated matrices passed in the final selected generator run (8/8, zero
+ ignored, 4.89 seconds); four commissioning-floor tests also matched that filter.
+
+## Published formal/CI batch resolution
+
+- The CI agent's final report is `/tmp/ap074-ci-ab36-report.md`. On the published
+ revision `ab36e4b76420a37912d00b2c5b1fac1ad6c556c5`, all 85 successful checks
+ include authoritative Lean and final formal evidence. The final formal job is
+ `123460599133` in run `40867724155`. Eight conditional skips establish no live
+ qualification. The three failed badges are the original authoritative CLI
+ fixture root and its two rollups, already repaired by local commit `d7eea272`.
+- That integrated repair's 24 focused checks passed. The broad workspace
+ reproduction did not pass; its load-sensitive file conformance case passed
+ only in isolation with original deadlines and assertions preserved. Combined
+ source still requires exact published-head hosted verification. Final formal
+ evidence on `ab36` does not attest to subsequent replay sources or journal `/16`.
+
+## Native mutation SIGKILL and reserved-continuation checkpoint
+
+- `native_mutation_crash_tests.rs` exercises the real native mutation engine in
+ a separate OS worker with independently opened custody and the maintained
+ file/TLS PostgreSQL stores. A transparent test-only adapter wrapper pauses at
+ actual before/after reservation, entry and response commit boundaries. A real
+ local HTTP server pauses after a partial response. The parent sends SIGKILL,
+ verifies signal 9 and absence of a caller result, then reopens native history.
+ No native verifier result, decision closure, durable ACK or permit is replaced.
+- The matrix includes the maintained Airtable recipe and the closed
+ `stripe-plain` recipe in the maintained hostile-recipe corpus. Stripe authority
+ carries native count and sum commitments: one invocation and 500 minor units are
+ held in slot 0, with no slot 1 allocation, refund or second charge after restart.
+ Initial admission/slot existence is all-or-none. Every observed write body and
+ declared idempotency header equals the original native review. The provider
+ counts all requests and performs no idempotency deduplication.
+- Seven interruption points for each recipe and adapter give 28 crash histories.
+ A committed grant projects Unknown even when the provider received zero writes.
+ Loss during/before response persistence remains Unknown with the actual request
+ count retained; a committed response returns its HTTP 200 fact after caller loss.
+ Only absent or reserved operations with no prior grant may enter after full
+ fresh native authorization. A second replay changes neither journal/slots nor
+ the same native execution witness and performs no implicit observation.
+- Another 32 histories kill workers after reservation or before entry, then
+ independently expire approval/resource authority, revoke the native connection
+ or activate changed native configuration before reopening. All retain Reserved
+ and the exact original charges/revision; no fresh custody, write or read occurs.
+ Original proof/action/approval bytes and admission event times are not renewed.
+- The test authoring helper now accepts an explicit signed validity interval.
+ Existing callers retain their original intervals. New Stripe proofs are signed
+ from `NOW - 10` through `NOW + 3600` (or `NOW + 5` for isolated authority expiry), within the
+ recipe's declared idempotency retention. The initial long-lived fixture was
+ correctly refused with `gateway.idempotency.window-exceeds-retention`; no
+ verifier clock, retention rule, deadline or installed trust was changed to make
+ the positive control pass.
+- Required generator execution passed 4/4, zero ignored, in 12.01 seconds: the two
+ new adapter matrices plus the two existing acquisition crash generators. The
+ new matrices contain 60 histories; all fourteen earlier acquisition histories
+ and their committed artifacts stayed unchanged after sharing the transparent
+ test pause/process helper. PostgreSQL uses the independent canonical TLS
+ fixture on port 55451 with CA/server-name verification and synchronous commit.
+- Both new ignored PostgreSQL/file cases are explicitly registered in the hosted
+ workflow and compliance inventory. These are source/artifact-construction
+ results; exact new-head hosted verification remains required. Full Stripe
+ guard/prerequisite and disclosure interruption schedules, partial upload and
+ device/power failure, broad fault/concurrency histories, installed human UI,
+ differential/performance and the remaining A01–A22 gates remain open. Fixed
+ public test keys, maintained clock, trusted stores, local HTTP and same-UID
+ process assumptions are explicit; no live-provider or hostile-host claim follows.
+
+## Holistic CI repair after retained replay publication
+
+- The exact published revision `9ae2e2c7` finished with 64 successful checks,
+ thirteen failures and nineteen skips across all fourteen workflows. Two source
+ roots caused three failed jobs and ten fail-closed rollups: strict Clippy
+ diagnostics, and the same substituted-challenge assertion in both installed
+ Stripe/Airtable demos. Fast Lean feedback passed; required qualified formal
+ translation/Kani/Lean/evidence were preflight-skipped, not new passes.
+- The CI agent's single source batch `7ad003e423c6290599dee44842883f8dff210df0`
+ is integrated into this combined change. A differing offered native proof or
+ action is now an advisory history miss, allowing ordinary native verification
+ to preserve the existing `denied / challenge-mismatch` result. Matching history
+ still requires both digests and the retained exact approval signature. Corrupt,
+ unavailable, foreign or invalidly signed retained state remains a hard refusal.
+ No existing demo assertion, security deadline, tool pin or invariant changed.
+- New replay controls are stronger: each changed native proof/action is compiled
+ and independently exactly approved before the original entry. Live changed
+ bindings must reach durable operation conflict; current expiry, revocation or
+ configuration refusal has its explicit expected code. Reusing the original
+ approval on a different effect still fails before custody, with exact journal,
+ provider count/body and native execution witness preserved throughout.
+- The agent reproduced the actual failing full workspace strict Clippy command
+ successfully, and passed four retained replay matrices on file/TLS PostgreSQL,
+ the native header-corruption test and unchanged demo diagnostics: Stripe 23,
+ Airtable 34 semantic plus 2 crash cases. The demo diagnostic uses existing installed
+ artifacts and same-UID execution; it is not fresh Docker/SDK/isolation evidence.
+ Detailed report: `/tmp/ap074-ci-9ae2-report.md`; machine matrix:
+ `/tmp/ap074-ci-9ae2-evidence.json`.
+- Published PostgreSQL job `123481020880` in run `40883517998` actually passed all
+ sixteen explicit native cases, zero ignored in each, plus thirty existing
+ gateway PostgreSQL tests. The actual producer revision is PR merge
+ `d842ec5607c0d0c4fa915a732df78d5720b3aad7`, whose verified tree
+ `101b09d359ecba54d666b61d92f3b79d886c6ca8` equals the published source tree.
+ Preserve that producer identity; do not relabel it as literal source head.
+- The combined gateway library/tests also passed strict all-feature Clippy in
+ 12.59 seconds after repairing three new test-style diagnostics with equivalent
+ patterns/borrowing. No lint allowance was added and every assertion remains.
+- The combined source includes the sixty new mutation crash/continuation
+ histories, so fresh exact-head hosted verification is still required. Release
+ freeze 404 advances only the reported identity protocol 148, SDK contract 145 and
+ public surface 404 subjects. Journal schema remains `/16`; canonical effect
+ identity and existing native fixture bytes remain unchanged. AP074 is not
+ complete and no acceptance gate is waived by this source repair.
+
+## Native disclosure/ACK process crash checkpoint
+
+- `native_disclosure_crash_tests.rs` and its private testkit run separate native
+ gateway/recipient OS processes over real authenticated Unix sockets. The source
+ is acquired by the real native engine through local HTTP and retained privately.
+ Release uses an independent native proof, exact approval, recipient key/channel
+ and native fixed-window count/byte commitments. No verifier result, entry
+ decision, durable acknowledgement, live channel or permit is substituted.
+- Each adapter has 28 histories: eight reservation/entry/partial-delivery/receipt/
+ ACK-commit interruptions, plus twenty expiry/revocation/configuration controls
+ around reserved continuation and genuine late ACKs. The gateway is SIGKILLed at
+ every selected point. Five histories per adapter also SIGKILL the recipient
+ after full native receipt but before ACK transmission; its exact private ACK
+ was retained before death. Native receipt returns no partial plaintext/ACK when
+ the gateway dies after a content prefix.
+- The partial-send selector is compiled only into the native unit-test executable
+ with loopback/testkit features. It pauses after an actual native prefix write,
+ preserving frame bytes, deadlines and the real consumed permit. Ordinary
+ binaries, including installed testkit binaries, have no selector/branch. The
+ shared test adapter observer has a distinct ACK phase; response and ACK facts
+ remain typed separately in the sole shipping journal/reducer.
+- Durable grants reopen as Unknown with no new challenge or plaintext. An absent
+ or reserved operation continues only with current native authority and approval
+ through the original writer fence and a new real recipient handshake. Expired
+ approval, expired resource authority, native revocation and changed native
+ configuration retain the original reservation/revision without authenticating
+ another recipient, fetching a provider response or recharging capacity.
+- Both source and release lifetime allocations, native fixed-window count/sum
+ slot 0 rows and absence of slot 1 are checked. A source acquisition remains one
+ invocation/1024 response bytes; disclosure holds one delivery/full projected
+ byte amount even on partial/unknown delivery. Original signed bytes, source
+ journal and charge rows remain byte-identical throughout replay and late ACK.
+ Exact repeated ACKs preserve the journal revision. Current expiry/revocation/
+ configuration cannot erase a valid assertion for a previously granted delivery.
+- Thirty histories actually submit genuine late ACKs (including exact replay of
+ already committed ACKs). Before the positive ACK, each exercises ten native
+ historical API substitutions: effect, source, protected content, channel,
+ instance, projection, byte limit, acknowledged length, an independently signed
+ different recipient, and damaged signature. All 300 negative controls leave
+ history and charges unchanged. A valid recipient signature without any durable
+ grant also cannot manufacture ACK history. Public snapshots contain no content.
+- Required construction passed 6/6, zero ignored, in 33.33 seconds, covering all
+ 130 existing/new acquisition, mutation and disclosure process histories over
+ real file/TLS PostgreSQL. The two new matrices contain 56 histories. All older
+ acquisition/mutation crash artifacts stayed byte-identical after sharing the
+ observer. Combined strict all-feature gateway library/test Clippy passed in
+ 11.47 seconds; only equivalent future boxing/reference fixes were needed, with
+ no new lint allowances or weakened assertions.
+- Both new cases are explicitly registered in the PostgreSQL workflow and
+ compliance inventory. These local source/artifact results do not close all
+ A05/A06/A08/A09 gates: actual remaining I/O-error/ambiguous-ACK-commit races,
+ guarded-provider/prerequisite, concurrent native histories, independent
+ differential, installed authenticated human UI, retention/rollback and
+ performance/formal integration evidence remain required. Fixed public keys,
+ maintained clock, same UID, trusted host/stores and local provider are explicit;
+ no live-provider, distinct-UID isolation or human-presence claim follows.
+
+- Source identities are regenerated under freeze 405 (only generator-reported
+ protocol 149, SDK 146 and public surface 405 subjects advanced). Journal schema
+ remains `/16`; no normal transport, grant, ACK, clock or provider-retry contract
+ changes. Publication was held until the actual authoritative Lean job
+ `123512213089` and final evidence for published `9913d4c7` finished; no live
+ formal run was cancelled to publish another test fixture batch.
+
+## Published 9913d4c7 hosted verification checkpoint
+
+- Source `9913d4c702451c19de2b601cd634e397db20f951` is CI-qualified: all 96
+ checks finished, with 88 successful, eight conditional skips and zero failures.
+ All fourteen workflows are terminal: thirteen successful and custody-live
+ conditionally skipped. No repair or rerun was needed for this revision.
+- Authoritative clean Lean, 329-statement audit, four Rust refinement checks,
+ 49 Kani harnesses, final formal evidence and CI-qualified passed. Pinned
+ translation used validated same-PR evidence reuse from `ab36e4b7`; it was not
+ freshly translated. All eighteen explicit native PostgreSQL workflow cases
+ actually ran with zero ignored; installed packages, demos and isolation passed.
+- The [durable hosted evidence report](0074-hosted-evidence-9913d4c7.md) records
+ every workflow, exact jobs, reused translation provenance and the actual PR
+ merge producer identity with a verified identical source tree. Conditional
+ skips do not establish live-provider, production or human-review qualification.
+- This closes the CI blocker for `9913d4c7` only. The following disclosure test
+ commit `a81943a7` and later revisions still need exact-head hosted checks. Full
+ AP074 acceptance remains open; the separate reference AP070 is not migrated.
+
+## Native observer-expiry and threshold-selection checkpoint
+
+- The existing dependency matrix now has twelve histories per adapter (24 total).
+ Its original eight histories remain identical. New independent worker/operator
+ histories advance the maintained clock while entry waits behind the actual
+ file/TLS PostgreSQL connection-record writer fence. No verifier result,
+ admission/entry decision, signed event time or store snapshot is substituted.
+- Two signed-observation controls exercise the inclusive five-second freshness
+ boundary and stale evidence at six seconds. The latter retains Reserved with
+ one call/1024 bytes and no provider request. Two 1-of-2 threshold controls use
+ the same complete proof/action/observation bytes: both branches authorize at
+ five seconds; at six, the proof remains Authorized through the longer-lived
+ branch. Actual selected branch/observation counts and native authority
+ commitments are checked, and only the authority field of the compiled binding
+ changes. The old exact approval remains in time but cannot approve that effect.
+- The installed worker repeats the complete native check inside the actual
+ final writer fence. The changed selection yields no grant, HTTP entry or new
+ allocation; unchanged selection enters exactly once. These final-fence
+ schedules occur after initially authorized admission/custody. All four new
+ cases measure one earlier credential lease; refusals correctly measure zero
+ HTTP reads. No claim that a later refusal reverses an earlier lease is made.
+- Original signed inputs stay byte-identical, admission remains at the original
+ time, positive entry uses the post-wait time, and reopened history leaves the
+ journal unchanged. Fixed public keys, maintained test clock, trusted host/
+ stores, local HTTP and same-UID process assumptions remain explicit.
+- Required artifact generation passed both adapter generators (2/2, zero
+ ignored, 5.02 seconds). Final refactored source passed no-update generation
+ verification (2/2, zero ignored, 5.03 seconds) and the previously failed strict
+ gateway all-feature library/test lint (7.74 seconds). These are construction
+ and focused reproduction results, not exact new-head hosted qualification.
+ A13/A14 installed surfaces and full mutation/disclosure races remain open.
+- Gateway source closure was regenerated and its rebuilt no-update check passed.
+ Semantic-freeze generation reports no changed release subjects, so freeze 405
+ and every existing semantic version remain unchanged. This batch changes only
+ native test authoring/evidence and documentation, with no shipping behavior,
+ journal schema, clock rule, approval rule or verification limit change.
+
+## Native independent recovery actor checkpoint
+
+- New `native-independent-recovery-{file,postgres}.json` contains four histories
+ per adapter, with actual native writer and separate recovery-reader OS
+ processes and real local PATCH/GET transport. The operator activates a native
+ trust configuration withdrawing the writer root while retaining an independent
+ reader root and actor. No post-verification oracle, substituted record,
+ provider response injection or reconstructed permit is used.
+- The original writer proof and exact approval remain in time at the recovery
+ clock. Previously valid, exactly approved new write/read submissions are
+ refused with native `untrusted-root` after activation, without admission,
+ custody or HTTP entry. Their initial native positive controls passed before
+ withdrawal. The reader's distinct principal, native read proof, current exact
+ approval and binding to the original source effect are checked. Source write
+ approval is independently rejected as recovery consent.
+- Four outcomes distinguish recorded/lost write response and state equality/
+ declared echo linkage. The provider receives the exact reviewed PATCH body
+ once, records that state, and answers exactly one bodyless GET. A matching
+ value without a link leaves lost-response history Unknown/Inconclusive. A
+ supported link may add provider-linked evidence while original response status
+ remains absent. A recorded response with equality alone projects Observed
+ without provider linkage; these facts are never collapsed into one success.
+- Recovery preserves every source field except the sole reducer's supported
+ recovery facts/revision, including original identity, actor/proof/approval,
+ connection/configuration attribution, entry audit and charges. Both signed
+ input sets stay unchanged. Recovery replay preserves source/read snapshots and
+ execution witnesses; reopened historical writer replay preserves both journals
+ with zero lease/read/write calls even after current authority is withdrawn.
+- Required construction passed both adapter generators (2/2, zero ignored,
+ 2.36 seconds), and no-update verification passed (2/2, zero ignored,
+ 2.45 seconds). Combined strict all-feature gateway library/test Clippy passed
+ in 13.13 seconds without lint allowances or relaxed assertions. The provider
+ helper explicitly restores blocking mode on
+ accepted sockets; its existing three-second I/O deadlines remain unchanged.
+ This fixes inherited nonblocking behavior in the new test helper, without
+ changing any native deadline or assertion. Both explicit tests are registered
+ in the hosted PostgreSQL workflow and compliance inventory.
+- This is operator root withdrawal, not signed principal/grant status-service
+ revocation evidence. Fixed public keys, simulated reviewer signatures,
+ maintained clock, trusted operator/host/stores, same UID and local provider
+ assumptions remain explicit. Installed authenticated UI/language paths, other
+ revocation scopes, isolation and live-provider qualification remain separate
+ gates. AP070 remains independent and AP073 remains untouched.
+- The required gateway source closure is regenerated. Freeze 406 advances only
+ the generator-reported compliance-bound identity protocol 150, public SDK 147
+ and public surface 406 subjects. Journal schema `/16`, effect schema `/3`,
+ compiler `/8`, all native clocks/deadlines and shipping semantics remain
+ unchanged. This batch and local threshold commit `0e20c5f8` require fresh
+ exact-head hosted verification after publication. The authoritative Lean
+ job `123552204257` for published `ce9d6cf5` was preserved and completed; its twenty explicit
+ native PostgreSQL workflow cases passed with none ignored, including the newly
+ published disclosure matrices. That evidence does not attest to these later
+ unpublished sources or close complete AP074 acceptance.
+
+## Authenticated owner review UI checkpoint (authored; hosted execution pending)
+
+- The managed Airtable demo has an opt-in simulation-only owner origin. It
+ withholds all application exact approvals at startup, renders the complete
+ native literal output and signing request, and requires authenticated explicit
+ approve/decline before the operator signs that exact pending native request.
+ Native code still owns identity, verification, approval encoding, admission,
+ execution and recovery. No presenter hashes, reconstructs or authorizes effects.
+- A root-private per-run login file maps the owner to the installed approver.
+ The approver key remains in operator memory outside the application's UID.
+ Owner bearer/CSRF values stay only in owner-page memory and explicit same-origin
+ headers. Cookies are absent because localhost cookies cross ports and would
+ reach the untrusted application. No private authentication value enters URLs,
+ logs, application status, browser storage or source-controlled evidence.
+- The new channel fails closed on host/origin/session/CSRF mismatch, stale or
+ substituted pending IDs, unsupported renderer/envelope, truncated/mismatched
+ native projection, duplicate/malformed/oversized commands and declined/expired
+ consent. One pending review, eight HTTP handlers, bounded input/complete native
+ output, HTTP deadlines and separate session/consent limits are explicit.
+ Existing 1,024-byte recovery-source RPCs retain that bound; mutation proof/action
+ proposals use a separate bounded socket. Native clocks and final checks do not
+ change. Default simulation/live channels retain their previous behavior/claims.
+- Unit transport tests and a source-free installed browser/native journey are
+ authored and registered in hosted CI/compliance. The latter checks application
+ UID denial on the owner login file, empty startup exact approvals, no owner
+ authentication on the app origin in a shared browser context, exact complete
+ native display, decline/substitution with no entry, one approved write,
+ independently approved recovery and replay with unchanged execution witnesses.
+ Screenshots and a bounded manifest-linked JSON report are hosted outputs.
+- Per GitHub-first policy, new local unit/browser/build/lint gates have not run.
+ This is authored implementation and CI coverage, not passed hosted evidence.
+ Resource-proof manager approvals and initial simulation commissioning remain
+ labeled scripted setup; browser owner actions simulate consent and establish
+ no human-attention/production claim. Full managed acquisition/disclosure through
+ installed UI/language consumers, broader isolation/negative parity and all
+ other A01–A22 requirements remain open. AP070 remains independent.
+- P0's owner/consumer and finite-bound inventories include this demo consent
+ transport. Required release generation updates freeze 407 and only its reported
+ identity protocol 151, public SDK 148 and public surface 407 subjects. No Rust
+ runtime, gateway source closure, native schemas, formal predicates or tool pins
+ change in this UI batch.
+
+## Published ce9d6cf5 terminal hosted checkpoint
+
+- All 96 check records are terminal: 86 successful, two cancelled and eight
+ conditional skips. Main CI `40899439849`, authoritative implementation/Kani/
+ clean Lean, final formal evidence `123567518646` and CI-qualified
+ `123572202035` passed. The complete PR matrix is not all green: Python workflow
+ `40899439869` failed solely because CPython 3.10/3.11 macOS consumers never
+ acquired hosted runners, including one grouped recovery attempt. Neither
+ cancelled consumer ran a test. No source defect was identified, no healthy
+ execution was rerun, and no platform/assertion/deadline was removed.
+- PostgreSQL job `123537375377` executed all twenty explicit native selectors
+ with zero ignored, including the published disclosure crash matrices. Installed
+ Docker demos, TypeScript, SDK recipes, operator package, isolation and retained
+ independent Google reference passed. Conditional live/reviewer skips provide
+ no qualification evidence.
+- The [durable terminal report](0074-hosted-evidence-ce9d6cf5.md) preserves the
+ complete workflow matrix, exact producer merge/tree identity, validated
+ translation reuse chain and both original/retry macOS job identities and
+ runner-acquisition annotations. All useful formal work completed before the
+ next publication. Local threshold `0e20c5f8`, independent recovery `5eb761d2`
+ and owner UI `37a218c9` remain outside this older-head evidence and require
+ fresh exact-head hosted verification. Complete AP074 acceptance remains open.
+
+## Independent mutation differential checkpoint
+
+This batch starts from published `cb1ee6d15a8f728b2f6f9ce01021de69b67ac966`.
+It changes test-only native authoring/comparison, fixtures, CI wiring and
+documentation; no shipping authorization, admission, recovery, binding schema,
+compiler identity, provider recipe or AP070 source is changed.
+
+- Native generator
+ `observed_tests::effect_native_tests::differential_tests::native_independent_mutation_differential_fixture_is_current`
+ verifies actual signed native grants/actions/proofs and exact effect approval.
+ Airtable uses the independently reviewed fixed base/table pure expansion;
+ Stripe uses the maintained guarded platform recipe and a signed bounded
+ amount/count/sum policy. No admission, custody or provider call occurs.
+- The two-family native artifact has 28 request and 22 response cases.
+ Successful shared requests agree byte-for-byte with the unchanged independent
+ Airtable/Stripe mappings. Shared refusals, source-owned resource-pool limits,
+ fresh-guard versus compilation differences, full domain witness versus
+ generic observation, and distinct recovery locator capabilities have explicit
+ assertions. An injected read-only exchange runs each independent reference's
+ real discovery/read-back selection without a network or credential.
+- The native capability projection explicitly retains response/verified locator,
+ provider link, unknown reobservation, declared idempotency retention and
+ nonconditional-write facts. A declared Stripe idempotency key cannot fabricate
+ its missing recorded-response locator. Reference discovery is not a native
+ permit. Full contracts and remaining gaps:
+ [0074-differential-contracts.md](0074-differential-contracts.md).
+- Construction executed locally on this batch over the stated base:
+ `AUTHS_UPDATE_FIXTURES=1 cargo test --locked -p auths-gateway --features testkit-harness --lib observed_tests::effect_native_tests::differential_tests::native_independent_mutation_differential_fixture_is_current -- --exact`
+ passed one test, zero failed/ignored, 0.10 seconds after compilation.
+ `python3 qualification/reference/ap074_differential.py --update` constructs the
+ independently checked comparison report with native corpus and source hashes.
+ These are required artifact construction, not exact-head hosted qualification.
+- Eight adversarial comparator mutations are authored (request bytes, decision,
+ omitted/duplicate cases, response claim, missing-response locator, resource
+ scope and recovery capability); that unit test is not yet executed. The
+ existing reference unittest workflow will execute it. The Python source-contract
+ workflow now first executes the native generator without updates, then the
+ independent comparator and uploads both artifacts.
+- Semantic inventory construction advances freeze408 and only the reported
+ identity152/public-SDK149/public-surface408 entries. No formal source or
+ translation input was intentionally changed. New hosted execution remains
+ pending; the live cb1 CI/formal cohort is preserved while this batch is local.
+- A19 remains partial. AP070 acquisition/disclosure comparisons, native durable
+ contract overlap, and final hosted evidence remain required. AP070's
+ transactional release is not a recipient ACK, its task-lifetime budgets are
+ not native fixed-window/disclosure units, and its runtime lock is not native
+ multi-host durability. No managed guarantee is claimed for that reference.
+
+## Independent information comparison source (hosted execution pending)
+
+After local mutation comparison commit `91552fa5`, authored
+`qualification/reference/ap074_information_differential.py` without changing
+any AP070 file. This executes the retained application-owned SQLite state
+implementation at its existing synthetic post-authorization test boundary,
+using the same-checkout installed native SDK only for the unchanged canonical
+body-digest call. No Google call, managed credential/permit, reference migration
+or authorization/ACK equivalence claim is made.
+
+- Three count/byte/both scenarios use an initial admission and two independent
+ SQLite worker processes racing for the last unit; failed/closed work retains
+ all conservative call, content, metadata/provider-byte and two-attempt charges.
+- Four current/revoked/disconnected/expired release scenarios reopen state and
+ refuse repeat release/claim without refund. An additional successful release
+ followed by connection revocation provides the independent historical replay
+ comparison. The reference's release commitment stays distinct from ACK.
+- A real worker SIGKILL occurs after durable claim and provider-attempt start;
+ runtime reopen must mark the old claim abandoned, free only in-flight slots,
+ preserve all charges and refuse old-claim entry/release. It performs no I/O
+ against Google or a managed provider port.
+- The maintained PostgreSQL workflow preserves the output of the existing
+ native selector block without removing or repeating any of its tests. The
+ runner requires one actual pass, zero failed/ignored, for each of eight native
+ file/PostgreSQL capacity/replay/acquisition-crash/disclosure-crash selectors.
+ It checks their schemas and adapter identities, executes the reference, and
+ writes 22 paired comparisons with separate units, differing result meanings,
+ source/run/commit and artifact/log hashes. Synthetic log-parser controls
+ reject missing, duplicated, ignored and zero-pass executions.
+- New information runner, parser test and workflow extension are authored,
+ **not executed locally or hosted yet**. Existing cb1 native matrices do not
+ establish that the new comparison ran. The mutation report construction is
+ separate evidence. A19 remains partial pending exact-head hosted execution
+ and a full coverage audit. See the differential contract document for limits.
+- Required semantic inventory construction for this workflow/consumer extension
+ advances freeze409 and only its reported identity153/public-SDK150/
+ public-surface409 entries; no shipping Rust or reference implementation changes.
+
+## Differential response boundary hardening
+
+Review of the new mutation corpus found that its oversized response was also
+invalid JSON, so that negative alone could not isolate the byte-limit predicate.
+The follow-up over `f5f3a8ca` replaces it with complete valid domain JSON plus
+trailing whitespace at 16,384, 16,385 and 65,537 bytes. Native observation accepts
+exactly its configured 16,384-byte bound and refuses the next byte. The unchanged
+independent witness has a 65,536-byte bound and therefore accepts the valid
+16,385-byte response; both refuse 65,537. The difference is required explicitly,
+not hidden as equivalent bounds. A bounded padding count keeps the retained
+fixture readable; both implementations receive the actual materialized bytes.
+
+The updated native artifact generator passed one test, zero failed/ignored,
+0.11 seconds after compilation. Independent report construction also passed,
+now with 28 request and 26 response cases. Its nine adversarial comparator
+mutations include changing the exact-limit padding; that unit test and new
+hosted execution remain pending. Semantic inventory construction stayed at
+freeze409 without identity/hash changes. No shipping or AP070 source changed.
+
+## Resumed completion and retention checkpoint
+
+The user resumed the full completion prompt after the pause. The published
+CI-only repair `d9618a568c4db8fae0b8bc4cb80b77de81c8e7d1` is now an ancestor of
+this branch through a non-rewriting merge; the three A19 commits, duplicate
+local cherry-pick and all unfinished work were preserved. The durable goal
+above continues to mean all P0–P4/A01–A22, not merely the current patch.
+
+The exact published d961 Airtable run40942189355, package123688155935 and
+installed-demo123691155527 passed, including all 36 existing cases and the
+complete repaired owner-browser journey. Its verified artifact12928066979 is
+attributed to d961; production_qualification=false remains explicit. The saved
+[hosted report](0074-hosted-evidence-d9618a56.md) records the CI-only scope and
+pause as historical facts. On resumption, the commit-workflows API returned all
+14 d961 workflows terminal: 13 successful and custody live conditionally skipped.
+No unpublished A19 or retention behavior is established by those old-head runs.
+
+Retention implementation now removes the file adapter's full-directory
+collection/sort. Each call scans at most 4,096 directory entries, checks the
+common two-second cooperative deadline between entries and retains one
+nonauthoritative iterator for progress. It uses the same writer fence, only
+deletes expired count/sum slots and syncs safe deletions even if a later read
+fails. Restart loses only scan progress. Filesystem I/O latency remains an
+explicit host assumption. All other record kinds remain permanent; retained
+full attempts are the replay tombstones, with continuing storage growth.
+
+The engine sweep now uses the existing deployment/effect clock and fails before
+storage if required qualification lacks trusted time. A separately sampled wall
+clock could previously delete windows still current on that effect clock.
+No new clock-rollback or malicious whole-store rollback guarantee is claimed.
+
+Required fixture construction executed this batch over the reconciled branch:
+
+- `AUTHS_UPDATE_FIXTURES=1 cargo test --locked -p auths-gateway --features loopback-provider,testkit-harness --lib native_information_replay_tests::retention_tests::native_file_retention_fixture_is_current -- --exact`
+ passed one test, zero failed/ignored, 0.93 seconds after compilation.
+- The corresponding explicit `postgres_native_retention_fixture_is_current`
+ selector with `--ignored --exact` and the owned TLS PostgreSQL fixture passed
+ one test, zero failed/ignored, 1.63 seconds after compilation.
+
+The six histories use actual native signed bounded grants, exact approval,
+HTTP acquisition and authenticated Unix delivery/ACK loss. Cleanup preserves
+slots before expiry and removes two at exact expiry while approval is still
+current. Original replay cannot send again or allocate again; an independently
+signed new operation cannot consume the old approval. After reopening and full
+authority expiry, source/disclosure records and recorded allocations/charges
+remain byte-identical and provider reads remain one. Scope excludes independent
+OS isolation, mutation/reserved retention and epoch/rollback-floor evidence.
+
+File cursor scan bounds/corrupt-slot controls, required-untrusted-clock refusal,
+and both-adapter conformance over all five durable record kinds are authored,
+not executed locally. The PostgreSQL workflow adds the two native selectors
+(24 explicit selections total) without dropping existing cases. Hosted checks
+must execute these and the previously unpublished A19 comparisons on the new
+head before any completion claim. AP070/AP073 source remains untouched.
+
+The preserved information comparator also now uses the native artifact's actual
+`disclosure-acknowledged` label; the prior shorthand was not an accepted native
+protocol value. This is a correction to previously unrun test orchestration,
+not a runtime protocol change. Required semantic inventory regeneration advances
+freeze410 with only reported identity154/public-SDK151/public-surface410 inputs.
+Effect/compiler/attempt schemas and translated pure predicates stay unchanged.
+
+## Bounded inventory and first hosted differential results
+
+Published `76d008a020e5b8e6c89e70bcb86ebc34e2c0e29d` has real hosted progress,
+not complete qualification. Python source-contract job123730999956 executed
+the exact native mutation generator and independent comparison: 28 request and
+26 response cases passed; report artifact12934145521 was checksum verified by
+the CI subagent. PG job123730998172 executed all 24 selected native cases,
+including the two new retention generators, before the new information runner
+failed during spawned crash-worker startup. Its downstream restore stages were
+skipped and are not passed evidence.
+
+The CI repair retains a parent reference to the multiprocessing Event until
+the real crash worker exits; a temporary Event could lose its semaphore before
+spawn unpickling. The full 22-pair information runner passed a focused reproduction
+using the downloaded hosted native log and current hosted SDK wheel; no AP070
+source or SIGKILL/accounting assertion changed. Source-only reference discovery
+still requires no installed SDK; its existing 104-test suite passed. Strict
+repository preflight found one additional test-generator semicolon lint error;
+fixing it requires no allowance or predicate change. The CI subagent's focused
+all-feature gateway library/test clippy reproduction passed. Both repairs were
+committed together as `037cb97f` and integrated locally as `49a5c0f0`; publication
+and new exact-head hosted confirmation remain required. The independent mutation
+comparison report was regenerated because it hashes that generator source.
+
+The next bounded-read issue was file `claimed_keys`: it collected and sorted
+the entire directory before truncating. The new implementation keeps at most
+4,096 candidate keys in a max-heap, preserving exactly the original sorted
+smallest-key prefix when successful. It inspects at most 16,384 directory entries
+plus one overflow probe, with the shared two-second cooperative work budget.
+Irrelevant filenames count against work. Oversized requests or incomplete scans
+fail unavailable without a partial or apparently empty inventory. Direct
+known-operation loads, admission and replay are unchanged. The support consumer
+already distinguishes unavailable from empty inventory. Large file archives can
+exceed this explicit diagnostic-query limit; no records are discarded to fit it.
+
+Two controls are authored for exact prefix/cardinality, zero/excess limits,
+incomplete scans, nonclaim entries and unchanged direct lookup. They have not
+yet executed locally or hosted. Final A19/A21 closure still requires actual
+new-head report execution and the remaining scope-specific evidence; neither
+the local reproduction nor a parent workflow name replaces that evidence.
+
+The combined worktree's exact failed-check reproduction,
+`cargo clippy --locked -p auths-gateway --lib --tests --all-features -- -D warnings`,
+passed in 13.23 seconds; this checks the integrated lint repair and new inventory
+source but does not execute those unit tests. Independent mutation report
+construction passed again (28/26). Required source-closure construction emitted
+the same `a002dc42725b1e3e5234f8ba11eea3f8ba8141e2e64f052d950182d4f93a2825`;
+it is not a new Lean/Kani/translation result. Semantic inventory construction
+advances freeze411 and only reported identity155/public-SDK152/public-surface411.
+
+The prior cohort is now terminal: 96 checks = 65 success, 12 failure, 19 skipped;
+14 workflows = 11 success, CI/PG failed, custody conditionally skipped. The
+required implementation/formal rollups failed closed after the identified
+preflight root; formal-fast passed explicitly without qualification. All 24
+selected native cases passed in PG before the reference failure. Full provenance
+and exact PR-merge/tree identities are preserved in
+[the 76d report](0074-hosted-evidence-76d008a0.md). No healthy jobs were cancelled.
+
+Criterion-specific audit must avoid treating unrelated open gates as absent
+native evidence. A09 already has 56 actual disclosure crash histories and 30
+late-ACK histories across both adapters, with 300 independently re-signed or
+signature-damaged substitutions. Source `substituted_acks` exercises effect
+(including operation identity), source effect, content, channel, delivery
+instance, recipient/key, projection, byte ceilings/acknowledged length and
+signature, asserting unchanged raw journal and charge records after every
+refusal. The positive late/repeated original ACK controls refine once and
+preserve charges. These native generators passed on 76d; the next exact-head
+run must confirm them before current-head closure. Broader language/containment
+gates remain independently mandatory under A17/A18/A22 rather than disappearing
+from the full objective.
+
+## Native epoch control and exact-head evidence audit
+
+On published `74ce9f753faa6bf3f1cb065bc369f655e6c0b236`, PostgreSQL job
+123745460518 executed all 24 explicit native selectors with one matched,
+nonignored passing test each, then passed the independent information comparison
+(22 pairs across both adapters) and physical/WAL restore controls. Python source
+job123745459370 passed the native mutation generator and independent comparison
+(28 request / 26 response cases); source-only reference job123745461358 passed
+104 tests, including parser/contract adversarial controls. Artifact checksums and
+merge-tree provenance belong in the cohort report. These are executed evidence
+for that published revision, not verification of subsequent epoch changes.
+
+The current epoch regression first failed against the prior implementation:
+configuration keys included epoch, so a new epoch selected a separate active
+control. The fix keeps one control per namespace/partition, stores its active
+epoch and permits only nondecreasing explicit operator activation under the
+existing writer fence. Cached old hosts and old exact approvals cannot authorize
+new entry after activation. Historical operation keys remain epoch-scoped;
+retained facts confer no permit. Private configuration/key `/2` and installation
+manifest `/7` directly reject obsolete disposable state without migration.
+
+Both new native artifact constructors passed locally against file and TLS
+PostgreSQL (one test, zero failed/ignored each, 0.33s and 0.40s respectively).
+They use actual signed proofs, exact approvals, real HTTP and two engine/store
+instances, reject old-host entry and fresh-token epoch downgrade, preserve both
+epochs' histories and assert lifetime allocation slots `[0, 1]`. They establish
+this concrete activation order, not every concurrent schedule or malicious
+store-rollback detection. The unchanged trusted-store assumption and independent
+rollback-anchor exclusion remain explicit. Shared configuration conformance now
+also authors epoch advancement, retained revocations and lower-epoch refusal;
+its new behavior awaits hosted execution. Canonical configuration construction
+passed with obsolete-schema, missing/null/zero/out-of-range epoch refusals.
+
+The differential coverage audit confirms that the hosted A19 corpus exercises
+two mutation recipes with distinct recovery capabilities plus the declared
+acquisition/disclosure overlap. All twelve generator/reference/comparator source
+hashes in the downloaded information and mutation reports match the current
+checkout. The contracts document now distinguishes executed74ce evidence from
+pending subsequent/final-head verification instead of calling that runner unrun.
+
+The PostgreSQL workflow now explicitly selects both new epoch constructors in
+addition to the prior 24. Existing selectors and independent comparison parser
+requirements are preserved. The exact hosted strict-clippy failure reproduction
+passed after factoring the longer test history into focused helpers; no lint
+allowance or acceptance predicate was weakened. The stale gateway semantic
+closure was regenerated and passed a rebuilt non-update verification; this
+repairs the previous cohort's sole implementation root. This gateway closure is
+distinct from the formal translation source closure and proves no new theorem.
+
+Staged Gitleaks v8.28.0 found two public configuration digests. The fixture
+generator now uses explicit `sha256:` values and digest field names under private
+fixture schema `/2`; regeneration and the repeated staged scan pass without a
+new ignore. Semantic metadata construction advances freeze412 and only reported
+identity156/public-SDK153/public-surface412. Formal translation source closure
+remains unchanged; metadata generation is not fresh proof execution.
+
+Full AP074 completion remains open: installed authenticated managed UI/language
+journeys, remaining mutation/disclosure concurrency and adverse durability cases,
+complete negative/containment inventories, reconciliation limits and escalation,
+generated durable histories/formal linkage, supported rollback floors, bounded
+performance and all required hosted checks on the final completed head. AP070
+and the paused AP073 checkout remain unchanged by this batch.
+
+## Installed managed owner/agent journey (authored)
+
+The previous published74ce cohort is fully terminal: 85 passed, 3 failed and
+8 conditionally skipped checks across 14 workflows (12 succeeded, CI failed,
+custody conditionally skipped). The three failures share the stale gateway
+semantic closure root repaired in local epoch commit `41763d5a`. Actual formal
+aggregation passed after 329 compiled Lean claims, 49 Kani verifications and the
+four native Rust refinement tests. Translation evidence is validated reuse under
+unchanged a002 source, not a new translation. Full exact identities, artifact
+checksums and scoped behavioral evidence are in
+[the 74ce report](0074-hosted-evidence-74ce9f75.md). No healthy jobs were cancelled.
+
+The next source increment composes the existing native content capability into
+the installed authenticated owner journey. A separate local data-only managed
+recipe retains the original mutation declaration and adds a bounded fixed-record
+GET (response1,024/content512 bytes). The maintained live/reference recipe stays
+unchanged. Native recipe-review construction completed successfully and produced
+`demos/airtable-gateway/managed-review.json`; the hosted packaging workflow now
+reconstructs and compares it. This is actual pure compilation/review evidence,
+not an executed read/disclosure or new provider qualification.
+
+The private demonstration plan advances directly to `/2`; old disposable plans
+are refused rather than read through compatibility fallbacks.
+
+Installed Python and TypeScript clients each submit a separately authorized
+acquisition. A separate complete native owner review gates release to a pinned
+local recipient. The native CLI runs as the application UID and owns recipient
+identity, channel authentication, exact binding/bytes and ACK. The browser shows
+native acquisition/ACK facts and preserves the difference between sending an ACK
+and the gateway recording it. Authorized recipient output is confined to recipient-private
+files; public history does not return it. The actor is the same native agent for
+read, disclosure and write. The development proposal simulator retains a fixed
+root-private actor-key copy before app ownership changes, never loads a mutable
+app signer configuration as root, and remains explicitly separate from exact
+reviewer keys. A thin testkit seed import calls the existing native Ed25519
+implementation; no Python signing/hash/ACK or admission semantics are added.
+
+The source-free browser scenario is expanded to assert actual provider counters,
+review-before-entry, both installed language acquisitions, no content release on
+acquisition, separate disclosure consent, native recorded ACK, private0600 files,
+public sentinel exclusion and replay without new calls or files. It then performs
+the reviewed write, independently authorized observation and a distinct real
+response-loss/unknown control with no reentry. These new runtime/browser/SDK cases
+are authored and have not executed locally or hosted. Exact-head hosted results,
+remaining negative/containment scope, file/PostgreSQL adverse schedules, generated
+concurrent histories, rollback floors and bounded performance remain mandatory.
+AP070 remains independent and AP073 untouched.
+
+Semantic metadata construction advances freeze413 and only the generator-reported
+identity157, portable-ABI126, mechanism-profile13, public-SDK154,
+simplified-waist20 and public-surface413 identities. The gateway Rust source
+closure remains the repaired epoch commitment; the formal translation source
+closure remains a002. These metadata updates do not execute the new browser/SDK
+cases and are not a fresh formal proof.
+
+## Native mutation/disclosure capacity and retention continuation
+
+CI inspection, reproduction and repair belong to the existing CI subagent, per
+user instruction. The main agent continues core implementation and required
+artifact construction. The subagent's coherent installed-demo repair was
+integrated as `25c93709` from `31f71657`, preserving the identical pending provider
+ledger regression. It creates the owned counter log exclusively at startup,
+uses the existing native JSON-string projection at `/fields/PrivateNote` for
+an application/json response, and correlates browser completion/operation IDs
+and the native versioned disclosure contract correctly. Native media checks,
+all counter/consent/privacy assertions and the original reference recipe remain
+unchanged. The private recipient receives the exact selected string, not a
+reinterpretation of the complete provider JSON as text/plain.
+
+The subagent reports successful full source-free local reproduction against the
+exact hosted ab980 SDK/native package plus the repaired payloads: 36 semantic
+cases, ordinary browser and complete authenticated owner content/write/recovery/
+unknown journey. Package labels still identify the original hosted ab980 source;
+this is repaired local execution, not a passed new-head hosted job. Precise
+producer identity, artifact hashes, repairs and limitations are in
+[the repair report](0074-ci-repair-ab980d10.md). New-head hosted confirmation
+remains required; the main agent does not duplicate the subagent's CI checks.
+
+Four new native constructors execute 24 histories across file and TLS PostgreSQL:
+three final-unit mutation races and three disclosure races per adapter, plus
+three reserved-retention histories for each protocol/adapter. Each race admits
+a prefix then starts two distinct OS contenders against shared durable state.
+Exactly one contender wins the last count/amount-sum or delivery/byte-sum unit.
+Native signed proofs and exact approvals are used, with actual HTTP mutation
+capture and authenticated same-UID Unix recipient delivery/ACK. Counter records,
+permanent lifetime rows and attempt bytes are inspected directly; replay after
+reopen creates no custody/transport entry, recipient content or new recorded
+charge. The disclosure contenders each host gateway and recipient tasks in one
+OS process; this proves independent admission contention, not hostile-recipient
+isolation or separate gateway/recipient processes within each contender.
+
+A 60-second window lets cleanup occur while original proofs and approval times
+remain eligible. Full native recompilation remains authorized but produces a
+new window-bound identity, and the old exact approval refuses it. Only expired
+window slots disappear. Terminal and reserved attempts and permanent allocations
+remain byte-identical; reopening cannot remint entry or re-charge old operations.
+Mutation reservation cases use actual SIGKILL after durable reservation ACK;
+disclosure cases use native reservation-only admission, which returns no permit.
+No synthetic grant, alternative journal, callback or second decision owner is
+introduced. Counts and byte/amount units remain distinct.
+
+Required local artifact construction passed each exact selector with one matched,
+nonignored test: mutation file1.61s/PG3.34s and disclosure file2.83s/PG5.60s.
+Each constructor contains three race and three reserved cases. The PostgreSQL
+workflow preserves all previous 26 selected native cases and adds these four
+for 30 explicit selectors. This is not new-head hosted execution or full A04/A21
+closure: dependency-change/entry races, remaining adverse durability schedules,
+supported rollback floors, complete containment/binding negatives, generated
+concurrent histories/formal linkage and bounded performance remain mandatory.
+
+Native source-closure construction and its rebuilt non-update verification both
+passed after these Rust test changes. Semantic inventory construction advances
+freeze414 and only reported identity158/public-SDK155/public-surface414; other
+identity counters remain unchanged. Formal translation source closure remains
+a002. These construction results are not new hosted/formal acceptance.
+
+## Native valid-authority scope and recipe checkpoint
+
+Two new artifact constructors cover file and TLS PostgreSQL. Each independently
+verifies eleven acquisition and eighteen disclosure substitutions with the ordinary
+native verifier before requiring managed compilation/admission refusal. These are
+fresh, genuinely signed proofs under the installed trust: unsupported task/scope,
+participant/ancestor/charge fields, caller partition/epoch, and substitutions for
+the closed namespace/recipe/content or disclosure contract, source, recipient,
+channel, delivery instance and byte ceiling. The expected compiler binding refusal
+and unchanged raw attempt records/execution witness are asserted. No credential
+lease or provider request follows a refused call. Original exact acquisition and
+authenticated same-UID recipient delivery/ACK are positive controls.
+
+Each adapter also runs shared-store cross-partition and cross-recipe histories
+with identical actor, resource and logical operation names. Different operator
+partitions compile the same valid native proof to different effect identities;
+the old exact approval cannot cross them, and foreign history/disclosure source
+cannot resolve. The original approval still executes and replays without new
+entry. A replacement recipe requires explicit activation plus native connection
+rotation/CAS. Activation alone cannot bypass the old pinned descriptor. Old proof
+and approval cannot authorize the new recipe; even fresh proof/approval cannot
+overwrite the retained operation. A separately approved new operation succeeds.
+Original recipe history remains available under its original recipe.
+
+The cross-recipe constructor exposed a missing recipe check in the disclosure
+proposal getter: compilation and entry already rejected foreign provenance, but
+the getter could return a proposal. `GatewayEngine::disclosure_target` now compares
+the validated retained binding's recipe digest before returning provenance, using
+the existing binding and refusal code. The regression requires immediate refusal,
+with no journal or execution changes. This adds no authority owner or task system.
+The test-only engine loader can initialize an explicit credential generation;
+the replacement record is checked for exact equality with the ordinary native
+generation-incrementing rotation result before the existing connection CAS.
+
+Required local artifact construction passed both exact selectors with one matched,
+nonignored test: file0.79s and TLS PostgreSQL2.25s. The two artifacts contain 58
+valid-authority substitutions plus four partition/recipe histories. The PostgreSQL
+workflow retains all 30 preceding selectors and adds two for 32 explicit native
+selectors. These are constructor results, not new-head hosted acceptance. Same-UID
+local transport proves no hostile OS containment. Complete binding-dimension,
+installed language/UI and concurrent dependency/fault obligations remain; A02/A16
+remain partial. AP070 and the paused AP073 worktree are untouched. CI inspection,
+holistic repairs and publication timing remain with the requested CI subagent.
+
+Native source-closure construction and rebuilt non-update verification passed.
+Semantic inventory construction advances freeze415 and only the reported changed
+identity159/public-SDK156/public-surface415 subjects. Formal translation source
+closure remains a002; these construction results are not fresh formal proofs.
+
+## Native mutation entry-order checkpoint
+
+Sixteen new histories cover eight schedules on each of file and TLS PostgreSQL:
+connection revocation and shared configuration activation before versus after
+the final entry grant, plus independent approval/native-authority eligible and
+expired controls during an actual writer-fence wait. Separate native worker and
+operator OS processes share the maintained adapter. The operator holds or changes
+a physically distinct connection/configuration row; the worker's final native
+callback receives snapshots only under the same real adapter fence. Worker and
+operator PIDs differ, and the worker must reach that final fence explicitly.
+
+When operator withdrawal wins, the native result remains Reserved with original
+attempt facts and zero HTTP writes. When the grant wins, the counting provider
+receives exactly one original closed Airtable mutation body. Terminal reopen and
+replay retain byte-identical journal data and perform no new custody or transport
+calls. Eligible lock-wait controls record their advanced grant time; expired proof
+or approval produces no write. Each worker already acquired one credential lease
+before the pause, and that observed lease is counted explicitly even when final
+entry refuses. Signed proof/action/approval bytes and original admission time
+remain unchanged; no test refreshes authorization or fabricates a grant.
+
+The acquisition scheduling helper now accepts an explicit operation key and
+recognizes the existing typed mutation and acquisition grant shapes. No shipping
+runtime/store behavior, new fault switch, callback authority or lifecycle owner
+was added. The initial constructor exposed two fixture setup errors: the helper
+had assumed acquisition's nested phase shape, and an operator using a new
+approver needed its own matching `GatewayAttempts` policy. Both fixture errors
+were corrected without changing native authorization or reducing assertions.
+
+Required artifact construction passed each exact selector with one matched,
+nonignored test: mutation file1.92s and TLS PostgreSQL3.01s. The original acquisition
+constructors also passed without update mode after the shared helper change:
+file3.63s and TLS PostgreSQL4.90s, preserving their committed artifacts. The workflow
+preserves all 32 preceding native selectors and adds two for 34 explicit selectors.
+This is local constructor evidence, not new-head hosted acceptance. Selected
+observer/threshold timing for mutation, full disclosure races and installed
+language/adverse/containment obligations remain; A12/A13 stay partial.
+
+Integrated the CI subagent's exact documentation correction as `0609ff25`:
+backticks around PostgreSQL in the mutation-budget module comment. Its isolated
+reproduction of the actual failed strict Clippy check passed; the main agent did
+not inspect CI or duplicate that check. AP070/AP073 remain untouched.
+
+Native source-closure construction and rebuilt non-update verification passed.
+Semantic inventory construction advances freeze416 and only its reported changed
+identity160/public-SDK157/public-surface416 subjects. Formal translation closure
+remains a002; neither metadata construction nor translation reuse is fresh proof.
+
+## Native disclosure entry-order checkpoint
+
+Sixteen additional histories cover eight schedules per file/TLS PostgreSQL:
+connection revocation and shared configuration activation before versus after
+the final native disclosure grant, plus eligible/expired independent approval and
+native authority during actual writer-lock waits. An original native HTTP
+acquisition records the exact private source. Separate worker and operator OS
+processes race the maintained adapter fence; gateway/recipient tasks within the
+worker use real same-UID Unix sockets and exact native recipient authentication.
+The scheduling helper never constructs a grant, decision or journal bytes.
+
+If withdrawal or expiry wins, the original charged disclosure remains Reserved,
+with no recipient content result or ACK. If the native grant wins, the recipient
+receives exactly the retained projection and sends its native signed ACK. An
+opaque same-process socket relay independently captures every gateway byte: after
+the authentication challenge, refusal must produce zero bytes, including partial
+frame prefixes, and success must contain exactly one length-prefixed original
+content frame. The relay forwards bytes without changing native authentication or
+decisions. This strengthens the zero-content assertion beyond the receiver's
+complete-message result; it establishes no hostile-channel qualification.
+
+All schedules retain one lifetime delivery/byte allocation, original admission
+time and exact signed input bytes. Eligible wait cases record the advanced grant
+time. Release/replay performs zero credential leases or provider reads/writes;
+the sole HTTP read is the original acquisition. After reopen, native history and
+raw attempt bytes remain unchanged. Terminal replay emits no authentication or
+content frames, sends no recipient ACK and recreates no permit/charge. Reserved
+history inspection does not attempt continuation or create delivery authority.
+
+Required exact artifact constructors passed with one matched, nonignored test
+each. The initial file/TLS PostgreSQL cases passed3.04s/4.24s; the strengthened
+wire-capture constructors passed3.10s/4.28s. The workflow retains all prior34native
+selectors and adds two for36explicit selectors. No production runtime/store
+behavior changed in this batch. These are local constructor results, not
+new-head hosted acceptance. Gateway and recipient are tasks in one worker, not
+separate hostile OS processes. Complete selected observer/threshold/revocation
+scopes, installed binding/UI, adverse durability, containment and formal/performance
+obligations remain; A07/A12/A13 stay partial. AP070/AP073 are untouched.
+
+Native closure construction and rebuilt non-update verification passed. Semantic
+metadata construction advances freeze417 and only reported changed identity161/
+public-SDK158/public-surface417 subjects. Formal translation closure remains a002;
+this is not new hosted or formal qualification.
+
+## Native selected-evidence parity checkpoint
+
+The mutation and disclosure entry matrices each add four schedules per adapter:
+single-observer and genuine 1-of-2 threshold controls at the inclusive five-second
+freshness boundary and one second beyond. Each typed matrix now contains twelve
+histories per adapter, preserving the eight earlier order/expiry schedules. This
+adds sixteen histories and brings the mutation/disclosure entry corpus to 48.
+The existing acquisition authoring/selection/binding test helpers are reused
+through test-module visibility; no shipping semantic owner or decoder changes.
+
+The full ordinary native verifier establishes each proof's authority before and
+after the future wait. At five seconds, the signed observation and both threshold
+branches remain eligible, the exact effect stays unchanged, and actual native
+entry succeeds. At six seconds, single-observer authority is stale. The threshold
+proof remains Authorized through the longer-lived branch, but its selected
+authority commitment changes. Fresh gateway compilation is still successful;
+the canonical binding differs only in native authority. The original approval's
+signature and time remain valid for its original binding, yet cannot approve that
+changed effect. Original proof/action/observation/approval bytes are never renewed.
+
+Actual independent worker/operator OS-process lock waits reproduce that final
+grant check on file and TLS PostgreSQL. Refused mutation retains its reservation
+and sends zero HTTP requests; positive mutation sends its own exact native bound
+request. A newly signed proof changes the recipe's declared provider echo, so its
+expected captured body is taken from that proof's native request rather than
+reusing another proof's body. Refused disclosure retains its lifetime allocation
+and sends zero bytes after the authentication challenge, independently checked by
+the opaque relay, with no recipient ACK. Positive disclosure gets the exact
+retained bytes and native ACK. Native history/replay conserves the existing facts
+and creates no new transport, credential lease or allocation.
+
+Required constructors passed with one matched, nonignored test each: mutation
+file2.90s/TLS PostgreSQL4.79s and disclosure file4.75s/TLS PostgreSQL6.54s. The same
+36 explicit workflow selectors remain; four of them now cover the expanded
+matrices. These are local constructor results, not new-head hosted qualification.
+The observation attachments use public fixture observer keys, not live-provider
+attestation. All selected-authority/revocation scopes, installed language/UI and
+hostile-process/adverse durability obligations remain distinct. A13/A14 stay
+partial. AP070 and paused AP073 remain untouched; CI belongs to the subagent.
+
+Native closure construction and rebuilt non-update verification passed. Semantic
+inventory construction advances freeze418 and only its reported public-surface418
+subject; identity161/public-SDK158 and the other counters remain unchanged. Formal
+translation closure remains a002. These construction results are not new hosted
+or formal qualification.
+
+## Native mutation/disclosure PostgreSQL batch-abort checkpoint
+
+Seven new histories reject every actual admission insert prefix: three mutation
+rows (attempt/count/amount sum) and four disclosure rows (attempt/lifetime
+allocation/delivery count/byte sum). Existing owned-schema trigger/sequence controls
+run inside the maintained TLS PostgreSQL adapter transaction. An independent
+initialized worker OS process reaches each exact failure; a nontransactional
+server sequence survives rollback and equals the rejected ordinal. Complete
+gateway-table counts plus known slot/allocation keys prove that every attempted
+prefix row rolled back, with no hidden or duplicate charge.
+
+The shared witness now also requires PostgreSQL's `is_called` flag to be true.
+Without that assertion, an unused fresh sequence's initial `last_value = 1` could
+be mistaken for a first-row exception. The strengthened oracle applies to all
+eleven acquisition/mutation/disclosure prefixes; native gateway decisions and
+transaction behavior are unchanged.
+
+Failed workers perform zero custody or provider calls. Failed disclosure admission
+also sends zero wire bytes, including no native authentication challenge. Its
+original acquisition journal and source allocation rows remain byte-identical.
+After removing only the owned fixture trigger, a reopened native engine confirms
+the operation is absent and explicitly reauthorizes the unchanged signed inputs.
+Mutation sends one exact native compiled Stripe request to the counting provider,
+which performs no idempotent deduplication. Disclosure authenticates a separately
+spawned native recipient OS process over a real same-UID Unix endpoint, delivers
+the exact content and records its exact signed ACK. The original source is never
+refetched; one lifetime release allocation, one count slot and one byte-sum slot
+remain. Terminal replay performs no custody/provider work, emits no wire bytes and
+preserves all inspected attempt/charge rows.
+
+Required constructors passed with one matched, nonignored test each. After the
+sequence-witness strengthening: acquisition 1.70s, mutation 1.32s and disclosure
+2.83s. The workflow preserves all 36
+previous native selectors and adds the two exact ignored-PostgreSQL selectors for
+38. Alongside the existing four acquisition prefixes, the corpus now covers 11
+actual native admission aborts across all three protocols. New compliance anchors
+name both constructors. The testkit exposes the existing ready-marker reader and
+strengthens the sequence witness; shipping runtime/store behavior is unchanged.
+
+These are local artifact construction results, not new-head hosted acceptance.
+They cover actual server transaction aborts and fresh handles, not ambiguous
+COMMIT delivery, sync/power loss, hostile process containment or the full A06
+fault matrix. The recipient retry crosses an actual process boundary under the
+explicit same-UID fixture policy; it establishes no hostile OS qualification.
+A06 remains partial. AP070 and the paused AP073 worktree remain untouched.
+
+Native source-closure construction and rebuilt non-update verification passed.
+Semantic inventory construction advances freeze419 and only its reported changed
+identity162/public-SDK159/public-surface419 subjects. Formal translation closure
+remains a002; metadata construction is not fresh formal or hosted qualification.
+
+## Native mutation/disclosure lost-COMMIT-reply checkpoint
+
+Fourteen new histories extend the actual TLS commit-reply-loss harness: mutation
+admission/final grant/response persistence, and disclosure admission/final grant/
+ACK/Unknown persistence, each with separately observed committed and aborted
+outcomes. The native worker reaches a deferred trigger in its actual COMMIT; a
+separate direct TLS client checks that backend's SSL session and observes its
+advisory-lock wait. An opaque TCP proxy withholds replies while the observer either
+releases the transaction to commit or terminates that exact fixture backend and
+verifies rollback. TLS is never terminated by the proxy and no adapter error,
+decision, permit or record bytes are fabricated. A transparent store wrapper counts
+one actual selected adapter error in each history.
+
+The shared observer now takes a typed native record path and exact operation key.
+Its query filters the attempt record kind, and all-row snapshots retain kind plus
+key, preventing distinct record families from collapsing into one comparison.
+The original acquisition artifact constructor passed without update mode in2.45s
+after this helper change and its committed artifact stayed byte-identical. These
+are test-only observation/transport changes, not a new runtime or lifecycle owner.
+
+Mutation uses maintained plain Stripe recipe/proof/approval and a counting HTTP
+provider that performs no idempotent deduplication. A committed final grant whose
+reply was lost stays charged Unknown even when no HTTP write occurred. Absent or
+reserved work continues only through full native reauthorization on a reopened
+handle. Native response evidence remains distinct from uncertainty, and actual
+captured bodies equal the original native compiled request. Count and amount-sum
+rows retain their exact bytes, first slot and original admission time; no replay
+leases credentials, sends again, refunds or creates another charge.
+
+Disclosure runs native gateway and recipient in separate OS processes, with an
+opaque same-UID Unix relay independently measuring every gateway frame. A lost
+grant reply sends no content; terminal replay emits no authentication or content
+bytes. ACK commit success despite reply loss remains Acknowledged. An aborted ACK
+write follows the existing native Unknown transition; an aborted Unknown write
+leaves the physical grant and a conservative Unknown projection. The recipient
+stores a genuine native signed ACK only after receiving the exact retained bytes.
+That historical ACK may later refine uncertainty or replay an existing ACK without
+delivery, provider/custody work, changed source facts or changed charges. No ACK is
+invented for the committed/lost grant that delivered nothing. The original HTTP
+source remains fetched exactly once across all release and recovery actions.
+
+Required exact artifact constructors passed with one matched, nonignored test:
+mutation 3.40s and disclosure 5.88s (initial 2.72s/5.77s before explicit pre-existing
+charge-byte comparisons). The workflow preserves its prior 38 native selectors and
+adds two for 40. The combined acquisition/mutation/disclosure corpus now has 20
+actual committed/aborted reply-loss histories. Compliance anchors name both new
+constructors; native source closure and release identities are refreshed.
+
+These are local artifact-construction results, not new-head hosted qualification.
+They do not establish file fault parity, server/power-loss protection, hostile OS
+containment or live-provider qualification. Existing native authority, approval,
+grant, transaction and historical ACK owners are unchanged. A06/A08 remain partial
+pending the remaining adverse/installed/durability evidence. AP070 and paused
+AP073 remain untouched; the subagent alone handles CI inspection and publication.
+
+Native source-closure construction and rebuilt non-update verification passed.
+Semantic inventory construction advances freeze420 and only its reported changed
+identity163/public-SDK160/public-surface420 subjects. Formal translation closure
+remains a002; these construction results are not fresh hosted or formal evidence.
+
+## Native mutation/disclosure file-error checkpoint
+
+Thirty-two new native histories extend the maintained file-adapter error harness:
+fourteen mutation schedules and eighteen disclosure schedules. Mutation admission
+has three rows (attempt/count/amount sum); disclosure has four (attempt/lifetime
+allocation/count/byte sum), in addition to its separately acquired source. Tests
+simulate errors at batch data sync/publication, every target install and all three
+admission directory-sync calls. Grant and protocol outcome persistence each cover
+data sync, replacement and directory sync, including separate disclosure ACK and
+Unknown transitions. The transparent selector now accepts an exact operation key
+and typed native stage path and forwards every native decision unchanged.
+
+Workers run in separate OS processes and report exactly one reached adapter
+failure site. Error returns are simulated; actual preceding writes, private file
+permissions, native verification/approval, writer fences, all-or-none rollback and
+transport execute unchanged. Pre-grant work retains no entry permission and can
+continue only through full native reauthorization after reopening. A grant that
+was installed before a lost directory-sync acknowledgement stays charged Unknown
+even with zero actual send. Response or ACK persistence that did complete retains
+its stronger fact. Terminal replay changes no journal bytes, execution witness or
+charges and performs no new I/O.
+
+Disclosure uses a native recipient in another OS process and the existing opaque
+Unix relay. Failed admission emits no challenge/content; failed final grant emits
+no content. The recipient can deliberately withhold its network ACK only after
+checking exact received content and storing its genuine native signed ACK. Failed
+ACK persistence follows the existing native Unknown path, while a persisted ACK
+cannot be overwritten with Unknown. Historical ACK submission can refine retained
+uncertainty without new delivery, provider/custody access or changed source/charge
+bytes. No ACK is invented for a lost grant that sent nothing. The sole provider
+read remains the original acquisition. First-slot count/sum/lifetime records are
+retained once, with no hidden duplicate or refund. Mutation has no disclosure
+source-lifetime allocation.
+
+Required exact artifact constructors passed with one matched, nonignored test:
+mutation 3.00s and disclosure 7.75s. The original acquisition file-error artifact
+passed without update mode in 3.40s, and the original native disclosure file-crash
+artifact passed without update mode in 13.34s after the recipient helper extension;
+both committed artifacts stayed byte-identical. The workflow retains all prior 40
+native selectors and adds two for 42. The modeled file-error corpus now has 47
+histories across acquisition/mutation/disclosure. Compliance anchors name the two
+new constructors. No shipping runtime/store or decision owner changed.
+
+The CI subagent's two-site import-order repair is integrated as `a2129593` before
+closure refresh. Both imports moved ahead of statements; assertions, behavior and
+fixtures are unchanged. Its isolated reproduction of the actual failed strict
+Clippy check passed. Main did not inspect CI or duplicate that check.
+
+These are local constructor results, not new-head hosted qualification or proof
+of actual kernel/device EIO, power loss, whole-store rollback, hostile OS isolation
+or live-provider behavior. Remaining adverse/installed/runtime-refinement and
+durability assumptions stay explicit. A06/A08 remain partial. AP070 and the paused
+AP073 worktree are untouched; all CI inspection/publication remains delegated.
+
+Native source-closure construction and rebuilt non-update verification passed.
+Semantic metadata construction advances freeze421 and only reported changed
+identity164/public-SDK161/public-surface421 subjects. Formal translation closure
+remains a002; metadata construction is not new hosted or formal qualification.
+
+## Current phase and acceptance gates — final coding batch
+
+The [final acceptance audit](0074-final-acceptance-audit.md) is the current
+P0–P4, E01–E05, R01–R10 and A01–A22 producer/requirement map. It supersedes the
+stale outstanding-work descriptions in earlier checkpoint tables, without
+removing any spec requirement or test. The historical checkpoints above and below
+remain provenance, not instructions to reimplement later completed behavior.
+
+| Phase | Current implementation and executed evidence | Unclosed gate |
+| --- | --- | --- |
+| P0 | Native owner/consumer map, ownership ADR, abstraction and independent reference contracts documented | Final source inventory and architecture/compliance audit |
+| P1 | Immutable binding, exact approval, complete literal review and installed native request path; signed binding/review corpus and installed consent/substitution controls execute on eb44d8f7 | Final repaired-head native/SDK/UI gates |
+| P2 | Sole journal/transition and admission/entry fence; all-protocol races, crashes, faults, guarded prerequisites, partial upload, strict journal/revision and recovery cases execute in the 63-selector file/TLS PostgreSQL cohort | Final broad repaired-head gate and conditional proof/runtime audit |
+| P3 | Native acquisition, separate recipient disclosure/ACK, retained history and thin wrappers; adverse read/release and complete installed owner journey execute on eb44d8f7 | Final repaired-head installed/platform confirmation |
+| P4 | Two-recipe/information comparisons, distinct-UID inventory, public/private audit, native cost and generated histories pass on eb44d8f7; formal links implemented | Two obsolete fuzz positives repaired through native generators; clean formal qualification, actual Lean reuse and final-head gates/requirement audit remain pending |
+
+The [eb44d8f7 hosted report](0074-hosted-evidence-eb44d8f7.md) records actual
+source-specific execution, including all 63 required selectors exactly once with
+one pass and zero ignored tests, 302 passing broad gateway tests, installed
+journeys and authenticated artifacts. Those are executed evidence, not merely
+authored producers. They cannot qualify the subsequent corpus-repair revision
+without its applicable gates. A18's stale committed corpus is an actual failure;
+A20's full formal qualification and final P4 acceptance remain open. The final
+audit names a producer and expected observation for every row. Live-provider
+qualification remains a separate, unclaimed deployment property. AP070 stays an
+independent reference; AP073 remains untouched.
+
+## Deployment assumptions retained
+
+Durability relies on acknowledged filesystem sync/rename and PostgreSQL commits.
+Neither an in-memory test nor a Rust permit establishes malicious-process
+isolation. Provider state matching does not establish exclusive causation.
+Recipient authentication is a transport assumption. Whole-store rollback is
+outside the guarantee without an independent monotonic anchor; a restored
+database cannot attest to its own freshness. No new live-provider qualification
+or deployment claim is made by AP074 source work.
+
+## Native recovery lifetime reporting and adverse HTTP checkpoint
+
+- `auths.gateway-recovery/2` now projects retained observation count and declared
+ response-byte reservations from all sixteen permanent source allocations under
+ the existing source/read writer fence. It validates canonical row identity and
+ the selected read's allocation; malformed or discontinuous retained usage cannot
+ produce a zero-usage report. No recovery ledger or decision owner was added.
+- The projection includes the selected complete response receipt time, a finite
+ 16-attempt / per-read response-byte ceiling, the maintained 15-second observation
+ deadline, its 240-second aggregate transport allowance, and the existing
+ 30-second evidence age. The aggregate excludes separate prerequisite transports
+ and is not a wall-clock completion promise. Reservations are not actual bytes
+ received and never imply a refund.
+- Native escalation distinguishes incomplete, unusable, stale and inconclusive
+ evidence; unresolved exhaustion takes precedence. Next actions are inspection,
+ requesting a separately authorized observation if capacity remains, and
+ escalation. None grants retry, disclosure, non-entry or accounting authority.
+- New file and TLS PostgreSQL constructors execute sixteen independently signed,
+ exactly approved reads after withdrawing the writer root. Actual HTTP returns
+ 404/no match, multiple records, unrelated echo, incomplete JSON, truncated HTTP,
+ an oversized body and state equality without causal linkage. Every source
+ record and its original signed inputs remain byte-identical and unknown.
+ Replaying every read adds no lease, request or charge. Read seventeen is
+ durably not-entered with no local allocation, while its public report retains
+ all sixteen earlier allocations and 262,144 reserved response bytes. Reopening
+ the adapter preserves exhaustion and authorizes no new I/O.
+- The constructors also exercise age 30 inclusively and age 31 as stale without
+ renewing receipt time, plus refusal of an out-of-range deployment clock. They
+ do not claim an actual elapsed transport-timeout experiment, unavailable store
+ or every revocation scope. Those remain explicit acceptance work.
+- Construction results: new file constructor 1 passed / 0 ignored (3.34 s), new
+ PostgreSQL constructor 1 passed / 0 ignored (5.79 s); maintained independent
+ reader constructors passed on file (1.55 s) and PostgreSQL (2.49 s); the exact
+ synthetic recovery report constructor passed (3.22 s); CLI/daemon content
+ journey constructor passed (3.21 s). These construct the changed artifacts,
+ not fresh hosted qualification. The initial mistyped synthetic selector
+ matched zero tests; its corrected fully qualified selector executed one test.
+- Both new selectors join the explicit adapter workflow, increasing the native
+ selector inventory from 42 to 44. Bindings return the native report bytes and
+ contain no parallel recovery calculation. AP070 remains independent with its
+ narrower application-owned claims; AP073 is untouched.
+- The CI subagent's seven-diagnostic durability-helper repair is integrated as
+ `a3835fa5` (source commit `ca1125bf`). It changes five test files only and
+ preserves assertions, recipient modes and returned slot bytes. The subagent
+ reports the specific failed strict Clippy reproduction passed; the main agent
+ did not inspect CI. Full AP074 completion is not claimed.
+- After the final helper cleanup, both new artifact checks passed unchanged:
+ file 1 passed / 0 ignored (3.36 s), TLS PostgreSQL 1 passed / 0 ignored
+ (5.12 s). The ordinary provider double still requires successful response
+ writes; only its explicit oversized-response fault permits a bounded client
+ to close early. Gateway semantic closure is
+ `d9b45da9ef5aa30a6601a673c3243b473c01c50fda922d602b3d4fe7d59f8b4e`;
+ its exact constructor and rebuilt non-update check each passed. Semantic
+ freeze 422 updates identity 165, public SDK 162, vocabulary 28 and public
+ surface 422 as required by the inventory generator. These source/metadata
+ construction results do not replace the exact-head hosted gate.
+
+## Exact retained recovery history and elapsed timeout checkpoint
+
+- Fixed a production discrepancy in `GatewayEngine::recover`: it previously
+ recompiled the reader's current authority for every request, including exact
+ replay. `acquire_recovery` already supported exact historical retrieval. Both
+ entry points now use one acquisition helper and the same native journal match.
+ The read operation comes from either the compiled binding or a journal-validated
+ canonical-action hint whose proof/action digests, complete binding, installed
+ scope and pinned approval signature have matched. Hints alone expose no facts.
+- Reader grant/approval expiry, operator withdrawal of the reader root and a
+ disabled connection cannot erase completed, ambiguous or reserved read facts.
+ A retained reservation is returned unchanged when fresh authorization fails;
+ it cannot resume or create a permit. New inputs still need current native
+ authority, exact approval, admission and the final durable grant. Historical
+ refinement preserves the original source actor, entry, charges and response
+ receipt time; stale unresolved evidence stays stale.
+- New file and TLS PostgreSQL history constructors execute four scenarios each:
+ linked response, inconclusive response, pre-grant reservation and actual
+ elapsed transport timeout. A provider control channel withholds every response
+ byte while the maintained 15-second native deadline fires; the test then
+ releases the provider. It checks the actual elapsed interval, the provider
+ still waiting at timeout, one GET only, unknown acquisition and retained
+ capacity. No synthetic transport timeout result is supplied.
+- The pre-grant scenario lets native admission and grant validation run through
+ the real adapter, then returns a controlled error before the grant is written.
+ Later expiry/withdrawal cannot resume this reserved read. Additional controlled
+ errors at attempt load and inside the actual adapter writer fence return typed
+ unavailability without a public zero-usage report, source/read mutation or I/O.
+ These are adapter-port refusal tests, not a claim about a physical database
+ outage, filesystem/device error or power failure. Earlier syscall/transaction
+ fault artifacts retain their separate claims.
+- Each scenario preserves the source/read and all sixteen allocation-key
+ snapshots across expiry, withdrawal, unavailable-store probes and connection
+ disable. Altered proof, action, source hint or approval is refused. Previously
+ approved inputs for a different, never-entered read cannot acquire custody or
+ contact HTTP after expiry or root withdrawal.
+- Exact constructor results: file history 1 passed / 0 ignored (17.07 s); TLS
+ PostgreSQL history 1 passed / 0 ignored (17.92 s). Existing independent-reader
+ artifacts passed unchanged on file (1.73 s) and PostgreSQL (2.43 s). These are
+ required artifact construction results, not hosted qualification. Both new
+ selectors are explicit in the adapter workflow, increasing the inventory
+ from 44 to 46. No AP070 or AP073 files were changed.
+- After restricting the pre-grant fault to this exact recovery operation key,
+ both new artifacts passed unchanged: file 1 passed / 0 ignored (17.15 s),
+ TLS PostgreSQL 1 passed / 0 ignored (17.90 s). Existing exhaustion artifacts
+ passed unchanged on file (3.31 s) and PostgreSQL (4.69 s); the exact
+ CLI/daemon content journey artifact passed unchanged (3.27 s). No broad
+ local CI gate was run, and no current-head hosted completion is claimed.
+- The same batch expands the existing actual-HTTP lifetime corpus with a
+ complete 16,384-byte response at the exact declared limit, an empty successful
+ response and duplicate response members. The limit response is completely
+ recorded and compared without implying causation; empty/multiple/duplicate
+ responses are unusable. The existing 16,385-byte case remains unknown, with
+ its full reservation retained. Updated constructors passed on file (3.29 s)
+ and PostgreSQL (5.10 s), each 1 passed / 0 ignored.
+- Final exact-limit assertions passed unchanged on file (4.59 s) and TLS
+ PostgreSQL (5.63 s), each 1 passed / 0 ignored. Gateway semantic closure is
+ `7371aee21694d1783a11ce8fc8977b8db6c921df83ca6407cfb01012136525fd`;
+ its constructor and rebuilt non-update check each passed. Semantic freeze
+ 423 updates identity 166, public SDK 163 and public surface 423, the only
+ changed subjects reported by the generator. The runtime orchestration change
+ adds no reducer transition, permit constructor or accounting authority.
+
+## Bounded admin decoder and managed interface checkpoint
+
+- Fixed native admin decoding that first collected JSON into a map and thereby
+ accepted last-key-wins duplicates. A bounded visitor now rejects duplicate
+ schema, command and argument members, including escaped aliases, unknown
+ members and non-string values. The existing command enum still owns valid
+ argument combinations and every refusal keeps `gateway.admin.invalid-frame`.
+ `MAX_ADMIN_REQUEST_BYTES` is 4,096. The admin socket checks the advertised
+ first-frame length before reading/allocating its body; separately bounded
+ secret/native-submission second frames retain their current protocol.
+- The native application enum and admin enum now generate their own catalogs,
+ with seven application schemas and eleven operator commands. One native
+ Clap metadata constructor inventories every command and argument, including
+ generated help (52 entries). Catalogs grant no authority and add no dispatcher.
+ A new native boundary corpus checks every operator command, identical/changed
+ duplicate members, escaped duplicates, unknown/nested members, trailing data,
+ 4,096 accepted / 4,097 refused and an oversized advertised length with no body.
+- New actual installed-daemon Unix constructor uses a fresh random credential,
+ real counting HTTP and retained private content. Its untrusted client exercises
+ all seven application schemas in 32 frame cases, all eleven operator commands
+ injected on the app channel, obsolete schemas/routes, raw HTTP and oversized
+ advertised frames. It reaches help for all 52 CLI entries. There is one
+ authorized acquisition only; four direct provider attempts without the fresh
+ credential return 401. Native custody/read/write counters do not advance.
+- Disclosure with actual signed authority/approval but no recipient authentication
+ stays reserved without content. The trusted native recipient then authenticates,
+ receives exact bytes and records its ACK. A later hostile replay gets history
+ without redelivery or new source I/O/charges. The source journal remains
+ byte-identical. Captured public results, CLI errors, daemon logs, post-ready
+ frames and decoded base64 fields are checked for content/credential plaintext,
+ hex/base64 encodings and raw/encoded SHA-256 digests. Ordinary provider fixtures
+ still require the exact credential; intentional unauthorized HTTP is enabled
+ only for this fresh-credential fixture and never mutates provider state.
+- Linux-only distinct-UID execution is authored in the existing isolation workflow,
+ preserving its two existing deployment tests. It drops capabilities/groups and
+ privileges, retains only the gateway's primary group for app-socket access,
+ probes every private fixture file, recipient outputs and gateway process
+ environ/memory, and verifies it cannot acquire the gateway UID or chmod state.
+ It intentionally makes the relocated admin inode traversable/writable to
+ exercise the accept-time UID gate for every operator command. Four privileged
+ CLI examples are refused. The workflow publishes an exact-source actual report.
+ This Linux execution is pending hosted evidence; the local Unix run uses the
+ same UID and does not claim OS containment. CLI help coverage is not a claim
+ that all privileged CLI execution paths were exercised.
+- Required artifact construction passed: native admin boundary 1 passed /
+ 0 ignored (0.01 s), native CLI inventory 1 passed / 0 ignored (0.01 s),
+ managed Unix interface constructor 1 passed / 0 ignored (2.31 s). The original
+ complete CLI/daemon content journey passed unchanged (2.86 s). The generated
+ help entry is included in the final count; the earlier unbuilt catalog had 51.
+- A17 remains open: positive configured signed mutation observations and legacy
+ public mutation evidence still require the raw body-derived commitment audit
+ and projection hardening. Private audit/evidence binding must be preserved.
+ No exit criterion was relaxed, no AP070/AP073 file was touched, and the main
+ agent did not inspect CI. Broader A18 binding parity, A20 formal/concurrent
+ evidence, A21 rollback schedules and final exact-head gates remain required.
+- Final gateway semantic closure is
+ `87d516a91c79770a0db7ce4e066e304b2358274bc5af99248ce21b1abef48efb`;
+ the exact constructor and rebuilt non-update check passed. Semantic freeze
+ 424 updates identity 167, public SDK 164 and public surface 424, the only
+ changed subjects required by the generator. No complete acceptance row or
+ new hosted/Linux result is claimed from these construction checks.
+
+## Public mutation evidence hiding checkpoint
+
+- `GatewayEvidenceSummary` now uses `auths.gateway-public-evidence/1`, exposing
+ channel, a native record-scoped hiding commitment and the original receipt
+ time. The result no longer exposes the action-derived echo or raw response
+ digest. The protected journal retains full evidence, digest, attribution,
+ immutable binding, grant and all charges. No record schema or reducer edge
+ changed, and persisted facts still cannot construct an entry permit.
+- The sole journal owner computes HMAC-SHA256 using the existing private immutable
+ 128-bit attempt nonce, a versioned domain, length-framed exact canonical
+ binding, a closed field label and the raw digest. Temporary decoded key bytes
+ are zeroized. Bindings receive this value and never compute it. Snapshot and
+ provider-evidence Debug now redact private locators, echoes and commitments.
+- Native file/TLS PostgreSQL constructors perform real native writes and separately
+ approved reads whose responses contain a private sentinel. Original-writer
+ replay after withdrawal returns the public hiding summary; reopening preserves
+ it without I/O. Two records with identical private response bytes have distinct
+ public commitments. Raw evidence remains complete, original source fields stay
+ fixed, and public/debug output excludes plaintext, encoded values and raw
+ evidence/action digests. Final constructors passed, each 1 passed / 0 ignored:
+ file 0.70 s, TLS PostgreSQL 0.90 s.
+- Python and TypeScript cut over directly to the new schema. Both reject old
+ evidence members, unsupported schemas, malformed commitments, unsafe times,
+ duplicate object members, non-integer number spellings and non-UTF8/BOM JSON.
+ Data parsing grants no authority. Fifteen native wire vectors are consumed by
+ both binding suites; those suites await hosted execution. The explicit native
+ adapter inventory increases from 46 to 48 selectors.
+- Integrated the CI subagent repair as `09838a19` from `1f731115`: scope the
+ permissive admin inode to untrusted probes, restore owner-only modes before
+ trusted controls, require acknowledged within the existing bounded wait, move
+ the included module below tests/content_journey to prevent Cargo auto-discovery,
+ and fix related strict lints without allowances. Compliance paths follow the
+ module move. The subagent reports actual Linux UID65534 execution passed with
+ all 32 app cases, 11 admin refusals, 52 help routes, 4 private CLI refusals and
+ 36/40 private-file probes; main did not inspect CI. Combined-head hosted
+ verification remains required.
+
+- Gateway source closure is `9f470c3eed04ddaf76e41bfb9a3565fc73ef75db51578c9b14a2bd9e04744824`; construction and rebuilt check passed. Semantic freeze 425 updates identity 168, portable ABI 127, public SDK 165, vocabulary 29 and public surface 425. The required metadata constructor completed successfully. Hosted checks remain the CI agent’s responsibility.
+
+### Saved unfinished privacy cutover
+
+- Signed outcomes and public pre-entry observations in the PR still expose
+ private facts; this checkpoint does not close A17 or the specification.
+- Branch `codex/ap074-public-output-checkpoint` preserves the in-progress direct
+ cutover to public signed outcome `/3`, redacted pre-entry history and app
+ observe `/4`, plus binding, corpus and operator export changes. Its resume note
+ distinguishes implemented source from incomplete consumers and stale metadata.
+- Public facts use a stable hiding effect reference and a hiding commitment to
+ the complete canonical journal record, excluding its nonce from the message.
+ Entry/fence/authority and accounting fields are included in the record seal.
+- The saved implementation improves the earlier proposed journal opening: the
+ operator exports the existing complete signed private outcome fact set and
+ protected pre-entry bytes from one loaded native snapshot. A separate closed
+ private outcome schema retains the existing audit decoder, presence predicate,
+ counter/pre-entry/basis checks and signature pin. No journal nonce is exported
+ and no second journal interpreter is introduced. This remains unfinished.
+- Complete SDK/demo/example/template and operator-bundle consumers atomically;
+ private outcome or pre-entry bytes must never return through an application
+ socket, UI, binding or public artifact. Preserve all current audit checks,
+ including negative/tamper cases. Reject superseded public schemas directly.
+
+## Committed pause checkpoint: bounded CI repair
+
+- Integrated the CI subagent's final four-file batch `2e31c2d7`: TypeScript API
+ declaration snapshot, closed public hiding-summary assertions in both demos,
+ and Stripe retention of the existing operator-private native exact binding.
+ Expected echoes are derived only from the private authored native binding;
+ raw public echoes or digests were not restored and linkage checks remain.
+- Fixed native redacted Debug builders with `finish_non_exhaustive`, and added
+ explicit redacted messages to private byte/digest equality assertions. No
+ sensitive values enter failure diagnostics. These address the CI subagent's
+ exact strict-lint findings without allowances or predicate changes.
+- The subagent reports that both actual installed Docker reproductions passed:
+ Airtable 36 cases and Stripe 23 cases, with production qualification false.
+ It reports 48 native selectors, 15 parser vectors, hosted Linux UID isolation
+ and exact producer/artifact hashes verified. Main did not inspect CI. Its
+ report is `/tmp/ap074-ci-1769-report.md`. Final combined-head hosted
+ qualification remains pending; no complete acceptance row is inferred.
+- Required source-closure construction and rebuilt non-update check each passed
+ (1 passed / 0 ignored). Only the semantic-freeze metadata constructor was run
+ afterward. No broad local suite or new CI-monitoring loop was started.
+- Final checkpoint source closure is
+ `93ef3b915cc48f0c0d4c0d4de7e03d0ed394e83526fbdfdc2ad19443dec5ec20`.
+ Semantic freeze 426 updates only the generator-required identity 169,
+ evolution contract 83 and public surface 426; the metadata constructor passed.
+
+## Signed public history and private audit cutover — resumed implementation
+
+- Reconciled saved checkpoint `f55aa61a` over published `3cb2caad`, preserving the
+ completed CI repairs. App observe `/4` now returns signed public outcome `/3`
+ or redacted pre-entry history. Action, recipe and full journal-record references
+ are hiding HMACs scoped to the complete binding and private immutable nonce;
+ raw body-derived digests, pre-entry values and relative basis stay protected.
+ Recipe references are hiding because a recipe can include fixed private data.
+- Operator-only `audit-export` uses the existing peer-UID gate and one native
+ snapshot. Private outcome `/1` retains all original full-fact presence, signature,
+ counter/recount, pre-entry and basis audit checks. Native bundle `/3` assembly
+ discards supplied public history and obtains private facts per operation;
+ absence remains unverified. It grants no admission or audit verdict. Bounded
+ sequence decoding, canonical base64 and aggregate caps apply before retaining
+ extra evidence. Serialization buffers and attached private export data are
+ zeroized on drop where owned by the new assembler.
+- CLI export modes select one operation or one submission inventory, run under
+ the existing 60-second operator deadline, print only success metadata, and
+ create a new 0600 file in an owner-private directory. They cannot overwrite.
+ Both demos, maintained Python/TypeScript journeys and generated scaffold use
+ operator-side completion; the app receives only public projections. Detailed
+ private bundles/reports are not copied to public demo evidence.
+- Cut over SDK types/parsers and native public/private vector corpora directly.
+ Removed the superseded outcome-v2 fixture and renamed affected binding tests.
+ Both SDKs consume 15 native mutation-result vectors plus 35 public-history
+ vectors covering malformed/obsolete/private schemas, byte/count/time limits,
+ duplicates, fractional/exponent numbers, BOM and noncanonical base64. Their
+ complete suites await hosted execution. The TypeScript declaration build and
+ required API snapshot constructor completed successfully.
+- Artifact construction passed: file privacy 1 passed / 0 ignored (0.73 s), TLS
+ PostgreSQL privacy 1 passed / 0 ignored (0.92 s); admin/app/public/private/code
+ constructors 5 passed / 0 ignored (0.03 s); built native CLI inventory 1 passed /
+ 0 ignored (0.01 s); installed Unix interface constructor 1 passed / 0 ignored
+ (2.73 s). Two fixture setup defects were fixed: activate observer configuration
+ before compiling approvals and use the silent activation CLI's exit status.
+- The Unix constructor now performs an actual approved mutation in addition to
+ acquisition/disclosure, inspects configured signed public history, completes a
+ private audit with one independently verified entry, checks 0600 output and
+ overwrite refusal, and preserves no-new-entry/custody and source-byte checks
+ under hostile app probes. Its inventory has 33 app cases, 12 operator commands
+ and 53 CLI help routes; the distinct-UID probe adds a fifth private CLI route.
+ Linux execution remains for hosted CI; this local constructor uses one UID.
+- Added private-export substitution/duplicate/encoding/collection-limit tests
+ without weakening existing private-audit tamper assertions. These tests await
+ hosted execution. Redacted protected counter, relative-basis and pre-entry Debug
+ implementations as well as the native private-export Debug.
+- Source-closure construction and rebuilt check each passed (1 passed / 0 ignored,
+ 0.11 s and 0.06 s). Exact combined-head hosted evidence remains required. No
+ complete acceptance row is closed by artifact construction alone. Main has not
+ inspected CI, migrated AP070, touched AP073 or expanded delegation.
+- Final privacy-batch source closure is
+ `899847a8307da6574f1b83a8fa9d85e9d4d54fa4f94c2952f696dafacb2e8a07`.
+ Semantic-freeze metadata is 427, with generator-required increments only:
+ formal closure 106, identity 170, portable ABI 128, external custody 7,
+ public SDK 166, vocabulary 30, evolution 84 and public surface 427.
+ The prior metadata-constructor process handle expired before its final output
+ could be recovered; generated files are present, and hosted verification must
+ establish freshness on the committed revision.
+- Both demos now retain detailed audit reports and independent read-back bodies
+ in private 0600 files. Public evidence carries bounded result counts and
+ checked-state flags. A recursive public-artifact scan checks raw, hex and
+ base64 provider/body-derived digests against retained private evidence.
+ These final Python changes await hosted execution.
+
+## Native bounded-cost evidence producer — 5 November 2026
+
+Added a test-only pass-through measurement cohort over the maintained file and
+TLS PostgreSQL adapters. Each backend executes eight real native mutation,
+independent recovery, acquisition and authenticated disclosure/ACK operations.
+The harness measures native verification against installed trust, compilation,
+exact approval verification, execution, every actual adapter call and native
+fenced callback, buffer sizes and increasing native inventory/decode costs.
+It asserts real HTTP/recipient counts, outcomes and existing per-buffer/fence
+limits. Linux RSS observations are process-wide; adapter elapsed includes waiting
+and I/O and is only an upper bound on the lock-held interval. No shipping decision,
+retry, storage acknowledgement or security deadline changes.
+
+The PostgreSQL workflow preserves the 48-selector differential evidence log and
+adds a separate two-selector cost step/artifact containing exact source and
+checkout revisions plus source closure. Full measurement scope and limitations
+are in [the native cost evidence plan](0074-native-cost-evidence.md). Actual cost
+runs await hosted CI; no performance result or completed P4 gate is claimed.
+
+Required source-closure construction passed (1 passed / 0 ignored, 0.08 s) and
+its rebuilt non-update check passed (1 passed / 0 ignored, 0.07 s). Final closure
+is `0581155d4d8b1b9ebb10fa4cd294e7f72e7440ba6482d30d45ae6ed38eef496e`.
+The semantic metadata constructor passed with no release identity changes:
+this batch changes testkit/evidence sources rather than the public contract.
+No local benchmark, broad suite, lint, formal or CI inspection was performed by
+main. CI publication and holistic repairs remain with the existing subagent.
+
+## Closed submission-result decoding and complete variant corpus
+
+A direct A18 source audit found native decoders accepted invalid HTTP statuses,
+unbounded refusal strings and an omitted optional status that SDKs reject.
+Python counted refusal characters rather than UTF-8 bytes. Native results now
+use the existing bounded refusal decoder, HTTP status 100–599 and a mandatory
+nullable provider-linked status. A unit-variant Serde regression also accepted
+extra fields on Unknown despite enum-level deny_unknown_fields. An empty-map
+native decoder closes that loophole without changing the Rust variant or its
+serialized representation. No authorization, transition or recovery rule changes.
+
+Both maintained adapter artifacts now contain 159 additional native-decoded
+vectors over all eight result variants, with 29 accepted and 130 refused cases.
+They cover every required field, duplicates/escaped aliases, unknown members,
+UTF-8 and astral-byte boundaries, invalid surrogate sequences, safe revision
+limits, HTTP boundaries, boolean/null/type substitutions, fractional/exponent
+numbers and BOM. The existing 15 evidence and 35 history vectors remain intact:
+209 shared cases are now consumed by both SDK suites. SDK execution awaits CI.
+
+Required artifact construction passed the exact native file and TLS PostgreSQL
+selectors (1 passed / 0 ignored each, 0.76 s / 0.86 s). The initial constructor
+correctly caught the Unknown extra-field discrepancy before its native fix.
+Source-closure construction passed (1 passed / 0 ignored, 0.07 s); its final hash
+is `e6112e97c83f4e4b422f62828b3840fa3db9a9a7a9421a0f090a401613e06f81`.
+Rebuilt closure and semantic metadata results are recorded before publication.
+These constructors do not close all A18 or establish installed package parity.
+Rebuilt non-update source closure passed (1 passed / 0 ignored, 0.07 s).
+Semantic metadata construction passed at freeze428, portable ABI129 and public
+surface428; no other semantic identity increment was required. Main did not run
+SDK, lint, broad test or formal gates and did not inspect CI.
+
+## Integrated holistic CI repairs after the private-history cutover
+
+Integrated the CI subagent's `80b89755` as `14cdc0f7`, preserving main's cost and
+strict-result work. [Its source/evidence report](0074-ci-repair-7807d04b.md) records
+five actual roots and precise focused reproductions. Native audit semantics and
+all existing assertions remain intact. The installed Python/TypeScript north-star
+journeys now preserve operator-private signed history before intentional rollback;
+a separate post-rollback export proves lost entries remain absent. The scaffold
+supplies its configured private admin socket. Authoritative generators refreshed
+Stripe decision-record pins and ten native differential effect IDs after the
+observation-contract change. The independent28request/26response comparison is
+unchanged. Clippy repairs use helpers, borrowing/import placement and a method
+reference, with no new allowances or removed assertions.
+
+The subagent reports all focused repaired journeys and the failed native/reference/
+Clippy checks passed in its isolated checkout. These are repair reproductions,
+not acceptance on the combined head. Its exact-source7807 hosted audit did verify
+48 nonignored native selectors,22 information pairs, both installed demos and the
+Linux33app/12admin/53CLI/5privateCLI inventory. Main did not inspect CI; new cost
+and209-vector binding work still require the next integrated hosted cohort.
+
+Combined source-closure construction passed (1 passed / 0 ignored, 0.08 s).
+The final combined closure is
+`f943f5421fbaca04419bbf535f6da621cb349884fc54813695788f7f14f110db`.
+Release metadata advances to freeze429, portable ABI130 and public surface429,
+reflecting the repaired generated Python consumer; no other increment was required.
+Rebuilt source closure and metadata construction must finish before publication.
+Combined rebuilt closure check passed (1 passed / 0 ignored, 0.05 s), and the
+semantic metadata constructor passed. The combined revision is ready for the
+CI subagent's exact-head publication and holistic verification; no completed
+acceptance or formal qualification claim is made for this new source.
+
+## Generated concurrent native histories and assurance boundary map
+
+Added test-only deterministic event generation with an independent set-of-entered
+operations oracle. Each durable adapter has four seeded histories,32 positive
+concurrent rounds and four fresh-operation/old-approval negative rounds.72 fresh
+OS workers reopen the actual store, synchronize at a bounded start barrier and
+call the ordinary installed native engine. Every round checks real HTTP entry
+counts and native transport witnesses, typed journal state, exact earlier record
+bytes and unchanged original source. The model chooses only expected entry counts;
+it supplies no authority, journal decision, grant, acknowledgement or retry.
+
+The new explicit workflow selectors preserve the existing48 adapter selector log
+and two cost selectors and upload actual scheduling results in separate exact-
+source artifacts. No canonical scheduling outcome is fabricated. Details and
+limitations are in [generated histories](0074-generated-histories.md). This adds
+mutation single-grant/replay evidence; existing all-protocol formal, capacity,
+crash, dependency, disclosure and recovery producers remain mandatory.
+
+Added the [six-property assurance map](0074-assurance-map.md), naming shipping pure
+predicates, their actual runtime callers, adapter/crypto/transport premises and
+concrete evidence families. It identifies the pinned translation/refinement,
+compiled theorem audit, bounded Kani and mutation source links without treating
+nonempty model evidence as authenticated runtime truth. No security invariant or
+acceptance row was weakened or declared complete.
+
+Source-closure construction passed (1 passed / 0 ignored,0.09 s); it compiled the
+new cohort without executing it. Actual generated-history and cost runs await
+hosted CI. Main did not inspect CI, run local broad suites or change AP070/AP073.
+Rebuilt source-closure verification passed (1 passed / 0 ignored,0.07 s), and
+semantic metadata construction passed without release identity changes. Final
+source closure is `0057a01cccebd4fbc6b16c08a1a042f4af83ceee5e21e67a72c0c08c4bce88a1`.
+No local generated-history execution is claimed; the new explicit CI producers
+must execute before their evidence is accepted.
+
+## Integrated cost-artifact and private-audit fixture repairs
+
+Integrated all five changes from the CI subagent's `bbb391a9` directly into the
+current generated-history source: the parent module ordering is formatted with
+the new generated-history module retained; the other four path diffs apply
+unchanged. Cost and generated-history outputs now use workspace-absolute paths.
+The Python audit argument-forwarding stand-in targets private-bundle.json and
+poisons public bundle.json to detect fallback, retaining every verdict, pin and
+option assertion. Two Clippy repairs change only a literal separator and unit
+closure pattern. No runtime semantics, vectors or acceptance assertion changes.
+
+[The subagent report](0074-ci-repair-46989b7d.md) records actual source46989 hosted
+results:48 ordinary native selectors and both new cost selectors each executed
+once,1 passed/0 ignored; cost selector durations3.88s and4.88s. Their metric files
+were not uploaded due to the now-repaired path, so measured distributions remain
+unavailable. The independent22 comparison and restore were skipped after upload
+failure and are not claimed as passed. Both installed demos, owner journey,
+28/26 mutation differential and real SDK journeys passed, as reported by the
+subagent. Final current-head qualification remains required.
+
+Main ran only source artifact constructors: updated closure1 passed/0 ignored
+(0.07s), rebuilt closure1 passed/0 ignored(0.05s), and semantic metadata
+construction passed with no release identity changes. Final source closure is
+`c46116f44bcab92337597ef95fc499dd79488a214584135c60135fb82c44150a`.
+Generated concurrent histories remain unexecuted until the next exact-head CI
+cohort. Main did not inspect CI or rerun local SDK/Clippy/cost suites.
+
+## Native signed review meaning and approval corpus
+
+Added five native-signed approval witnesses over maintained Airtable and Stripe
+refund example recipes. Changed label-as-literal-data, consequences, destination,
+financial units and the 80-byte unit boundary each change the recipe/effect. The
+original signed native action refuses the changed recipe. Fresh native authority
+still cannot reuse the original exact approval; original and freshly approved
+controls pass. Complete native literal/private-review equality and refusal of a
+summary/truncation renderer are asserted. Native proofs, policy scopes, currencies,
+amount sums and exact Ed25519 approval use the existing owners.
+
+Thirteen financial review source negatives reject missing/null/empty/whitespace/
+over-limit/control/bidi units, missing fields, unsupported review versions,
+missing/oversized/excess consequences and case-insensitive label collisions.
+The canonical [review/approval artifact](../../bindings/fixtures/gateway/review-approval-boundary.json)
+retains the native literal and signing request, with no provider/durability/human
+claim. [Scope and composition](0074-review-approval-evidence.md) keep the installed
+owner-channel and generic literal-renderer evidence separate and mandatory.
+Compliance names the exact native constructor; no public API or shipping decision
+changes. Changed units do not themselves qualify new provider-domain meaning.
+
+Required artifact construction passed (1 passed /0 ignored,0.09s). Updated source
+closure passed (1 passed /0 ignored,0.07s) and rebuilt non-update check passed
+(1 passed /0 ignored,0.07s). Closure is
+`a00efd11f811f045119cfef58ea049b73bdb6848f3eabfbfed9dc94e9d55d3b8`.
+Semantic metadata advances freeze430, identity171, public SDK167 and public
+surface430, only as required by the changed compliance evidence inventory.
+Hosted execution and complete acceptance closure remain pending. Main has not
+inspected CI or run local SDK, lint, broad test or formal suites.
+Semantic metadata construction passed for this review-evidence batch. No complete acceptance row is closed by the constructor alone.
+
+## Complete final hardening batch authored — verification pending
+
+The finite [coding checklist](0074-final-coding-checklist.md) is written before
+execution, following the owner's batch instruction now recorded in AGENTS.md.
+No test, formatter, fixture constructor or new CI cohort ran while authoring it.
+
+- Full maintained Stripe mutation: four installed prerequisite acquisitions
+ (account, two denied reads, relative ceiling), 15 parent schedules and 28 child
+ schedules per durable adapter. Onboarding probe remains candidate-owned.
+ Completed evidence stays byte-identical. Lost child grants preserve the parent
+ reservation, do not refetch evidence or create transport, and never refund.
+- Actual TCP partial-upload relay forwards one approved body byte, kills the
+ gateway worker, then disconnects. Reopen retains Unknown and charges, with no
+ new custody/transport on replay; no kernel-backpressure or provider-atomicity
+ guarantee is claimed.
+- Thirteen independent real-signed binding substitutions plus original/fresh
+ approval controls. Unsupported namespace/configuration refuses compilation.
+- Epoch activation added to all three independent worker/operator entry-order
+ matrices on both adapters. Native generation-two positive, database-only
+ regression, same-generation substitution and missing/malformed/obsolete
+ independent floors retain facts and refuse before custody/entry.
+- Shared corrupt-journal corpus over mutation, acquisition, acknowledged release
+ and lost-ACK release. Each variant preserves writer encoding except explicit
+ encoding negatives, so canonicalization cannot mask independent checks.
+ Obsolete, duplicate, malformed, zero/unsafe/fractional/bool revision, changed
+ immutable bindings/approval/authority and inconsistent evidence refuse before
+ custody or transport. Exact restored positives pass without replay I/O.
+ Oversized replacement refuses at the adapter and preserves original bytes.
+
+The shared journal decoder now requires its existing writer encoding and a
+nonzero JSON-safe revision/time. Corrupt history yields typed indeterminate
+state rather than manufacturing non-entry. No compatibility reader or second
+production interpretation is added. New explicit CI selectors collect every
+independent hardening failure before exiting, retaining separate logs without
+changing the existing independent-reference log inventory.
+
+Construction, metadata refresh, combined verification and final hosted audit
+are pending. AP070 remains independent and AP073 remains untouched.
+
+## Final batch construction and aggregate verification
+
+All planned new artifacts and changed native histories are constructed. Both
+maintained adapters passed the 43 guarded parent/prerequisite schedules, both
+partial-upload recipes, all corrupt-journal protocol corpora, independent floor
+cases and three-protocol epoch races. The 13 signed binding dimensions pass with
+original/fresh approval controls. Its fixture now collects independent case
+failures rather than stopping at the first one. The epoch-only acquisition race
+retains the same approver, keeping epoch and approver changes independent.
+
+The initial 22-selector construction aggregate passed21/failed1; collected repairs
+corrected test-only action/grant validity and a header outside the existing closed
+registry, plus five Clippy diagnostics using focused helpers and an enum. No
+shipping authority rule or test assertion was weakened. The five previously
+reported stale native compilation constructors passed with authoritative outputs.
+
+Aggregate verification: gateway library298passed/3failed/63ignored; all-target,
+all-feature gateway Clippy passed; workspace formatting passed; semantic freeze
+passed62entries/43publicRustpackages. Two failures were transitive application
+fuzz/frame artifacts using the regenerated native approval. Their constructors
+pass after regeneration. The remaining file-store raw race returnedUnavailable
+under concurrent local I/O, then passed its unchanged exact selector in isolation.
+No production deadline or assertion was relaxed; the full concurrent local run
+remains recorded as failed, not silently replaced by a narrow success.
+
+The final six-selector repair cohort passed both application constructors, both
+dependency adapters, unchanged file-store conformance and final source closure.
+All new/affected behavioral constructors pass. Metadata construction advances
+freeze431 and only reported identity172/publicSDK168/publicsurface431 subjects.
+Final hosted broad, adapter, platform, installed, containment, differential,
+formal/refinement/release and performance evidence remains the CI agent's job.
+The goal remains incomplete until its exact-head completion audit passes.
+
+## Final source-audit correction batch — authored
+
+Final requirement inspection found the P0 table's64-event promise was not enforced
+by the private journal's wider JSON-safe revision check. The complete correction
+is authored before any runs: one shared native revision advancement function,
+header/read-link bound checks, every legal increment and terminal overflow unit
+cases, and exact64/65 native replay on both adapters across all four retained
+protocol outcomes. At the cap, terminal history remains available while any
+further event fails before persistence or grant. P0's stale lock/inventory table
+values are reconciled to actual stricter maintained constants. No kernel/formal
+source, public result encoding, production deadline or existing assertion is
+relaxed. Artifact construction and verification run only after this entire batch.
+The CI agent is collecting977ae61f's actual roots and holds publication so these
+changes can join one combined repair cohort.
+
+The complete revision-bound correction cohort passed six exact selectors:
+all legal/terminal arithmetic boundaries, file/TLS PostgreSQL retained information
+and mutation journals, and source closure. Every selector matched once with one
+pass and zero ignored tests. Exact64 histories produced no custody, provider or
+recipient entry;65 refused and preserved bytes. All-target/all-feature gateway
+Clippy passed (10.02s), and semantic metadata construction passed without further
+identity changes. Final correction closure is
+`71ff1dcdd7464474a0d1b12f661f7277d665775a6a192128bbbc5870ff9231eb`.
+The correction is committed for integration with the CI agent's combined actual
+977 repair roots; final hosted acceptance remains unverified until that cohort.
+
+## Current hosted checkpoint and bounded final repair
+
+The complete combined implementation and CI repair source is
+`eb44d8f7082db6ed9ab46c1bf5319e8e736d7b28`. Its merge producer
+`8335af6a6f9170a12f4b63feae0f755e87290308` has the same tree,
+`195e1ae19c0491af1c7e8fdd6d0c63b87655fe89`. The
+[source-specific hosted report](0074-hosted-evidence-eb44d8f7.md) retains actual
+artifact provenance and replaces earlier merely authored status with executed
+evidence. All twelve applicable non-main workflows passed, including the complete
+63-selector durable adapter cohort, installed owner journey, differential,
+generated history, native cost and Linux distinct-UID containment producers.
+The broad gateway library passed 302 tests with no failures and 63 ignored tests;
+required adapter selectors execute separately rather than treating ignored tests
+as passing. Conditional live custody supplies no qualification.
+
+The remaining actual implementation failure is two stale committed fuzz positives:
+an old public outcome and observe/3 frame. The dynamic campaign generators already
+use current schemas. The complete repair uses the existing native canonical
+signer/frame constructors, current positives/refusals and preserved obsolete
+rejection controls. All six corpus tests and both native generators/unchanged
+checks pass in the bounded repair verification. No runtime decoder, signature,
+assertion or security invariant is relaxed. Hosted repaired-head confirmation is
+still required; the failed implementation/derivative rollup remains recorded.
+
+Repeated Lean work on each push came from the PR-wide formal planner and fresh
+jobs. Both Lean phases now support a successful, same-PR, clean authoritative
+producer with authenticated immutable provenance and the native complete-input
+digest; current Rust/source checks still run, and any missing or mismatched
+evidence falls back to fresh verification. Fast bootstrap Lean, two translation reproductions and all49Kani harnesses
+have passed. Clean authoritative Lean126929527493 also passed with329compiled
+statements/transitive-axiom audit and four Rust refinement tests. Its authenticated
+artifact13373617020 attests the complete proof key below. Final formal aggregation also passed. The overall old CI remains failed solely
+for the known corpus root and its dependent rollups (85success/3failure/8skips).
+Consumer-head qualification remains pending at publication.
+No warm hit is claimed before actual execution. The seed repair changes no proof
+inputs: the native digest is unchanged at
+`c523f465540e306ac9eaaf49ef460365b467be4bccaafd2160552ce23dc83432`.
+Preserve that successful producer and combine seed repair, evidence report and
+this acceptance reconciliation into one follow-up publication for warm verification.
+The exact-head [PR213 checks](https://github.com/auths-dev/auths-proof/pull/213/checks)
+and immutable final validation records will record consumer results without a
+recursive documentation-only push; no pending gate is represented as passing.
+
+The physical-WAL restore test explicitly reports `replay_continuity=false`.
+Connection-generation/exact-record floors protect their stated state, not an
+arbitrary rollback of attempt/counter history. It establishes before-target
+retention and after-target exclusion, not safe automatic managed resume after
+whole-journal rollback. File sync/rename, PostgreSQL participating-writer/TLS,
+clock, OS isolation and external-provider assumptions remain explicit. AP070
+stays an independent, narrower application-owned reference; AP073 is untouched.
+The goal remains incomplete until final applicable gates and the P/E/R/A audit
+close. No merge, deployment or live-provider qualification is claimed.
+
+
+## Final reuse dependency-order correction
+
+The b54 consumer passed all twelve applicable non-main workflows and retained
+all63 required native selector passes, differential/installed/containment/cost
+artifacts. Its first real Lean reuse attempt safely refused because the native
+warning audit required the pinned Aeneas checkout before workflow setup created
+it. This was an optimization defect; no proof or runtime security check was
+bypassed. The [reuse contract](0074-lean-reuse-contract.md) records the exact
+retained refusal and complete correction: unconditional pinned setup before both
+validators, expensive proof compilation still conditional, fresh-checkout/order
+regressions and unchanged package/provenance/fingerprint requirements.
+
+The corrected executor has a new complete proof key and needs fresh hosted
+qualification followed by real reuse in both jobs. Local affected checks pass;
+no warm result is inferred. Final acceptance remains open until those actual
+corrected-head gates and implementation results pass. Publication of this source
+repair is necessary; it is not a documentation-only retry of an unchanged build.
diff --git a/docs/plans/0074-lean-reuse-contract.md b/docs/plans/0074-lean-reuse-contract.md
new file mode 100644
index 000000000..60fdf3daf
--- /dev/null
+++ b/docs/plans/0074-lean-reuse-contract.md
@@ -0,0 +1,150 @@
+# AP074 follow-up Lean verification reuse
+
+The owner identified repeated expensive Lean work on every PR push. The CI
+planner compares the entire PR to its protected base, so a follow-up native or
+documentation commit continues to require formal gates. Previously both Lean
+jobs rebuilt unconditionally; authoritative Lean explicitly deletes repository
+outputs. Only Aeneas translation had validated reuse.
+
+The combined repair retains all required gates and adds reuse of a successful
+clean same-repository/same-PR authoritative Lean record to both jobs. The full
+provenance and complete-input contract is in [.github/ci/README.md](../../.github/ci/README.md).
+Native code recomputes the complete digest and reruns current Rust/source checks;
+GitHub metadata, exact producer attempt and archive integrity are independently
+validated before that call. Fallback is always fresh verification. No build cache,
+unsigned local fixture or old record without complete inputs proves anything.
+
+The first rollout is deliberately cold: source977's existing phase schema has no
+complete-input attestation, and the reuse executor/control plane itself changed.
+The running977 Lean job is preserved while authoring this batch but cannot seed
+the new optimization. Future native/docs-only edits leave the proof key unchanged;
+proof/generated/import/toolchain/assurance/executor changes invalidate it.
+
+Validation is batched after all implementation and negative controls are written.
+It covers exact key sensitivity and native/docs stability, transitive local data
+closure, every native phase/input/provenance field, failed/wrong-attempt jobs,
+foreign repository/PR/workflow, expired/missing/oversized/tampered archives,
+duplicate JSON and archive paths, legacy/chained evidence, current-check refusal,
+and mandatory fresh-workflow fallback. Synthetic controls are unit-test evidence,
+not an observed hosted reuse hit. The repaired head must bootstrap with real
+hosted clean Lean evidence; a later legitimate unchanged-input follow-up will
+establish an actual hosted hit. The required final evidence/acceptance-ledger documentation commit is the planned warm-hit verification; no speculative code publication is added.
+
+The actual977 run also exposed a second source of redundant pushes: source-closure
+regeneration left59 existing assurance citations stale. The bounded updater
+published `ee7975be` automatically while manual publication was held, changing only
+those source-closure evidence hashes. That commit is preserved in the combined
+ancestry. `semantic-freeze --update` now asks the native assurance owner to refresh
+only existing citations of `formal/qualification/aeneas/source-closure.json` after
+regenerating that file. This is metadata construction, not proof qualification.
+Typed negative controls preserve the complete statement/axiom/claim/review/semantic
+binding structure and every other evidence item, reject kind/path substitutions,
+and require the unchanged fresh compiled audit before qualification.
+
+
+The final affected verification passed: Python9 provenance tests; planner35 tests;
+21 native phase/assurance tests;25 qualification-policy tests; strict xtask/planner
+Clippy; phase ownership and production source-closure checks. Actual logs are in
+`/tmp/ap074-ci-lean-reuse-final`. The metadata constructor refreshed exactly59
+source citations and preserved the entire remaining typed assurance structure.
+Release inventory requires freeze432, assurance bytes62, source-closure bytes107,
+evolution85 and public surface432; other identities remain unchanged. Hosted
+fresh bootstrap and a subsequent genuine warm hit remain required evidence.
+
+A post-publication read of actual GitHub discovery metadata exposed a bounded
+lookup defect: each push creates14 workflows, so the first ten repository-wide
+runs contained only the current head. This safely fell back to fresh Lean but
+could prevent every warm hit. Discovery now queries the current CI workflow ID
+before applying the ten-candidate limit. Regression controls reproduce the14
+unrelated runs, prove the exact endpoint, and preserve the tenth/eleventh
+candidate boundary. This controller correction invalidates the initial proof key;
+the corrected head, rather than f5e9, must establish the fresh bootstrap.
+
+The complete lookup correction passed all10 offline provenance tests; the native
+metadata constructor passed with zero citation changes and byte-identical release
+inventory. The actual workflow-scoped GitHub endpoint returns current f5e9,
+previous ee797 and original977 CI runs, unlike the repository-wide first ten.
+No early f5e9 failed check was observed before this correction. Real bootstrap
+and warm-hit results remain pending; no hosted proof success is inferred.
+
+The requested audit against actual retained GitHub metadata found that a run's
+associated PR object updates its head SHA when the PR advances. Successfulce9
+run40899439849 and original977 now show the current20b PR snapshot. Historical
+identity instead matches across the immutable run, successful exact-attempt
+job123552204257, artifact12910849462 and its single lean.json member. Its verified
+ZIP SHA-256 is94562f1057bddcaae2aae0be22adc645c39830f498a72cf9a62c1057fb629cd1.
+The PR association still validates the number and both repository identities;
+only its mutable snapshot is excluded from historical source comparison. Tests
+accept this observed shape while refusing any mismatched immutable source field.
+The oldce9 record still lacks complete attestation and cannot qualify reuse.
+
+The complete metadata correction passed all11 provenance tests and the metadata
+constructor with no generated changes. Running the corrected transport against
+the retained actualce9 run accepts its same-PR association, then explicitly rejects
+its old phase for missing complete-input attestation. No native validator or
+qualified result was invoked or synthesized for that legacy artifact. No early20b
+failed check was observed before this correction.
+
+## Verified cold producer and pending consumer
+
+The corrected controller's real bootstrap is source `eb44d8f7`, CI run41924977969,
+attempt1. Fast Lean126908887765 passed; translation126913912807 reproduced twice;
+Kani126913912984 verified49harnesses. Authoritative Lean126929527493 completed its
+clean build and329-statement compiled audit successfully at23:38:21UTC. Final
+formal evidence126942117729 also passed; the three old CI failures all derive
+from the known corpus root (85success/3failure/8conditional skips). The
+[producer evidence report](0074-hosted-evidence-eb44d8f7.md) records immutable
+source/merge identities and authenticated artifact hashes. The old overall CI
+retains its separate, repaired corpus failure; a successful clean producer job
+is the reuse predicate, not a false claim that the entire old CI passed.
+
+Clean Lean artifact13373617020 contains only `lean.json`, with ZIP SHA-256
+`f2a256d1be842c86af7cf057b0e46e989e2029f5a4504ca18b2ecf093fd6818c`.
+It attests `executed-clean`, no reuse chain, source eb44/run41924977969/attempt1,
+and complete inputs
+`c523f465540e306ac9eaaf49ef460365b467be4bccaafd2160552ce23dc83432`.
+The actual native digest function returns this same key for the held corpus
+repair and final documentation. No producer result was synthesized locally.
+
+At publication of this report, real warm hits in both consumer Lean jobs and
+complete final-head qualification are still pending. The legitimate follow-up
+combines the required corpus repair with final evidence documentation. Both jobs
+must independently validate the retained producer, rerun current Rust/source
+checks and retain current-head provenance before skipping only the expensive
+Lean rebuild. Missing/mismatched evidence must still fall back to fresh work.
+Final results belong to the exact head in [PR213 checks](https://github.com/auths-dev/auths-proof/pull/213/checks)
+and its immutable validation records; no recursive docs-only publication is
+needed merely to record the outcome of the commit containing this report.
+
+## Actual first-consumer dependency-order correction
+
+The b54 consumer exposed a real warm-path miss: fast Lean completed reuse
+validation and then entered fresh setup. The native current-head validator audits
+`formal/.lake/packages/aeneas` for the exact pinned commit, clean status and
+reviewed warning inventory, but the workflow previously created that dependency
+checkout only after a reuse miss. A fresh hosted checkout therefore could not
+satisfy the intended audit. Both jobs now prepare pinned dependencies before
+validation; the expensive repository proof build remains conditional. No warning,
+commit, signature, provenance, current-source or refinement check is removed.
+Fresh-checkout and both-job order regressions reject omitted, late, conditional
+or ignored setup. The complete executor fingerprint is retained: this correction
+changes proof inputs and requires a new successful clean producer before another
+real warm verification. The eb44 successful proof remains valid for its exact
+inputs; it is not silently reinterpreted under this corrected executor.
+
+The retained b54 fast-job log confirms the refusal at00:03:19UTC: `git rev-parse
+HEAD` could not run in the absent Aeneas package directory. The affected batch
+passes27 native qualification tests, workflow ownership and strict xtask Clippy;
+metadata construction/non-update verification and formal source-closure check
+pass after advancing only the generator-reported counters. Freeze433 advances
+assurance bytes63, source-closure bytes108, evolution86 and public surface433;
+runtime protocol/SDK identities are unchanged. Exactly59 existing source-citation
+hashes changed, with all remaining manifest bytes identical.
+
+Corrected formal source closure is
+`09fb135d91c47ff1b2a873b583564db3467e76973c9763090bac2affd73b6837`;
+the actual complete native proof-input key is
+`9b5a7049779384207399e43d927cef5ccd856dce925ba80242c1bc0a25f929aa`.
+No local Lean/Kani run or corrected hosted warm success is claimed. The healthy
+old fallback may be superseded after remaining current implementation results
+are collected: its old executor key cannot qualify this correction.
diff --git a/docs/plans/0074-native-cost-evidence.md b/docs/plans/0074-native-cost-evidence.md
new file mode 100644
index 000000000..f3ac240e5
--- /dev/null
+++ b/docs/plans/0074-native-cost-evidence.md
@@ -0,0 +1,98 @@
+# AP074 native cost evidence
+
+AP074 §11 requires measurements as well as finite work bounds. The native test
+cohort in `native_recovery_actor_tests::cost_tests` executes eight separately
+authorized operations per path on each maintained adapter. Its four paths are
+Airtable mutation with a lost response, independently authorized linked recovery,
+bounded content acquisition, and separately approved authenticated disclosure
+with a real recipient ACK. It uses actual local HTTP and Unix transport and
+checks external call counts, received bytes and recorded native outcomes.
+
+The PostgreSQL lifecycle workflow runs both exact selectors after the existing
+adapter histories and uploads `ap074-native-cost--`. It preserves
+the original 48-selector differential log. The artifact contains
+`native-cost-file.json` and `native-cost-postgres.json`, each with the exact PR
+source revision, checkout revision and native semantic source closure. This is
+an additional evidence producer; existing fault, differential and formal gates
+remain required.
+
+Each sample records:
+
+- Native verification time against the engine's actual installed trust context.
+- Native installed compilation time, including its own verification and durable
+ configuration reads; it is not a subtraction-based compiler estimate.
+- Exact approval verification time, followed by ordinary runtime execution that
+ independently repeats its required admission and final-entry validation.
+- Every pass-through adapter call's elapsed time, input/output byte counts and
+ successful/error result. Each real fenced callback also records decision time,
+ number of snapshot records and aggregate snapshot bytes. Instrumentation
+ forwards the original callback, bytes, result and durability acknowledgement;
+ it does not choose authority, re-run a decision or add a retry.
+- Proof, action, context, binding, review and approval byte sizes. Actual binding,
+ review, approval and fenced snapshot bounds are asserted against native
+ constants. Only sizes and timings are exported; private bytes stay private.
+- Linux process RSS and cumulative high-water RSS when available. These include
+ the test runtime, native engine and local provider task. They are not allocator
+ accounting or an independent peak allocation for each path.
+- Native inventory enumeration **and record decoding** over increasing real
+ retained histories: 3–17 mutation/recovery attempts and 2–16 acquisition/release
+ attempts, with the production 4,096-entry inventory limit.
+
+Adapter elapsed time includes lock/pool waiting, reads, callback, persistence and
+unlock. It is a measured conservative upper bound on the lock-held interval,
+not an isolated lock-hold measurement. The callback time measures native work
+while the real adapter holds its writer fence. No network/provider or custody
+operation runs inside that callback. Filesystem blocking and database server
+scheduling remain operating-system/deployment assumptions.
+
+The source bounds remain: 262,144 bytes per opaque record, 64 records per batch
+or fence, two-second lock wait and fence deadlines, at most eight crash-batch
+recovery rounds, 16,384 scanned file directory entries and 4,096 retained
+inventory keys. A file recovery batch is bounded by
+`MAX_GATEWAY_RECORD_BYTES * 2 * MAX_GATEWAY_BATCH_ENTRIES` encoded bytes.
+These bound particular owned buffers/work units; they do not claim a total
+process heap cap. Native proof/action/context bounds and each recipe's response
+and content ceilings remain separately enforced.
+
+Eight small test-profile samples characterize these representative paths, not
+production p95 latency, maximum-population scan cost or worst-case heap usage.
+The raw distribution must remain available; no latency optimization may weaken
+native dependency verification, fresh fence validation or durability. The CI agent verified source `5acd62b0` measurements and retained artifact hashes;
+the measured table below is earlier-revision evidence. The final changed head
+remains unverified and complete P4 closure is still pending.
+
+
+## Native cost measurements
+
+Producer [PostgreSQL41903640108/job126832350307](https://github.com/auths-dev/auths-proof/actions/runs/41903640108/job/126832350307), attempt1. Downloaded artifact13358744149 ZIP SHA256 `c2f22d2ed1be9a81a6b9491fe06f88e551acb58c271970288c3fd99318e52825` verified. Raw files below preserve source5acd, merge4906 and gateway closurec461 identities above. Eight small test-profile fixtures per path per adapter; no production SLO or worst-case claim.
+
+All duration cells are **minimum / median / maximum milliseconds**, calculated directly from retained integer nanoseconds. Compilation includes native verification. Fenced callback and adapter per-call ranges pool the real execution callbacks/calls across eight samples; adapter totals sum elapsed calls within each execution sample. Adapter duration includes wait/load/decision/persistence/unlock; it is not isolated lock-hold time. Callback duration includes native validation and encoding, with no provider/custody I/O.
+
+| Backend/path | Verification | Compilation | Exact approval | Execution | Fenced callback, per call | Adapter, per call | Adapter total/sample |
+|---|---:|---:|---:|---:|---:|---:|---:|
+| file/mutation | 0.682 / 0.759 / 0.790 | 2.389 / 2.461 / 2.524 | 0.186 / 0.187 / 0.204 | 30.973 / 35.151 / 52.722 | 3.701 / 4.471 / 5.377 | 0.012 / 0.022 / 24.431 | 11.935 / 14.936 / 31.409 |
+| file/recovery | 0.691 / 0.722 / 0.800 | 6.650 / 6.764 / 6.888 | 0.175 / 0.177 / 0.193 | 62.791 / 63.183 / 82.912 | 7.612 / 10.348 / 18.844 | 0.011 / 0.027 / 19.982 | 33.828 / 34.604 / 44.607 |
+| file/acquisition | 0.746 / 0.757 / 0.772 | 3.602 / 3.664 / 3.779 | 0.182 / 0.189 / 0.199 | 47.814 / 54.784 / 65.110 | 5.845 / 7.281 / 9.007 | 0.012 / 0.022 / 15.214 | 18.172 / 18.794 / 25.689 |
+| file/disclosure | 0.716 / 0.753 / 0.782 | 3.216 / 3.323 / 3.369 | 0.187 / 0.189 / 0.200 | 68.578 / 70.135 / 102.108 | 10.345 / 14.997 / 20.797 | 0.011 / 0.025 / 36.314 | 33.468 / 34.663 / 58.944 |
+| postgres/mutation | 0.733 / 0.767 / 0.941 | 3.255 / 3.374 / 5.180 | 0.186 / 0.194 / 0.214 | 38.226 / 38.814 / 40.923 | 3.705 / 4.496 / 5.337 | 0.363 / 0.449 / 7.956 | 18.008 / 18.683 / 20.254 |
+| postgres/recovery | 0.694 / 0.714 / 0.902 | 8.021 / 8.125 / 10.153 | 0.174 / 0.181 / 0.202 | 76.172 / 78.377 / 88.508 | 7.647 / 10.447 / 12.775 | 0.363 / 0.484 / 18.087 | 47.319 / 48.335 / 59.375 |
+| postgres/acquisition | 0.731 / 0.769 / 1.950 | 4.489 / 4.547 / 5.954 | 0.182 / 0.185 / 0.206 | 52.586 / 68.293 / 83.657 | 5.900 / 7.351 / 8.877 | 0.325 / 0.454 / 11.710 | 25.783 / 31.355 / 36.357 |
+| postgres/disclosure | 0.709 / 0.748 / 0.775 | 4.151 / 4.177 / 4.728 | 0.191 / 0.192 / 0.211 | 77.029 / 78.328 / 85.115 | 10.530 / 14.824 / 21.274 | 0.376 / 0.476 / 24.747 | 41.644 / 42.753 / 45.029 |
+
+Snapshot bytes/records are maximum observed boundary values across compilation and execution, not allocator totals. RSS is Linux whole-process memory, including runtime/provider; high-water is cumulative for the test cohort, not an independent peak per path.
+
+| Backend/path | Execution adapter calls / callbacks | Max snapshot bytes | Max snapshot records | RSS KiB min/median/max | Cohort high-water KiB |
+|---|---:|---:|---:|---:|---:|
+| file/mutation | 88 / 16 | 4529 | 3 | 32984 / 33720.0 / 34048 | 34048 |
+| file/recovery | 104 / 24 | 12022 | 18 | 33324 / 33778.0 / 34084 | 34084 |
+| file/acquisition | 136 / 16 | 7332 | 4 | 34416 / 34970.0 / 35272 | 35272 |
+| file/disclosure | 96 / 16 | 16693 | 5 | 34708 / 35024.0 / 35336 | 35336 |
+| postgres/mutation | 88 / 16 | 4529 | 3 | 37860 / 38720.0 / 39036 | 39036 |
+| postgres/recovery | 104 / 24 | 12022 | 18 | 38308 / 38786.0 / 39080 | 39080 |
+| postgres/acquisition | 136 / 16 | 7332 | 4 | 41240 / 41652.0 / 41952 | 41952 |
+| postgres/disclosure | 96 / 16 | 16694 | 5 | 41460 / 41696.0 / 42016 | 42016 |
+
+| Raw artifact member | SHA256 of archived member |
+|---|---|
+| native-cost-file.json | `22f49ae6d0071091a0dbf09923440ab875bb64b4ae29a88550e3654c60083df0` |
+| native-cost-postgres.json | `777c78520adcc125a83a770e1f332015ed7ad2130ecb250e28aec34e96400842` |
diff --git a/docs/plans/0074-public-output-checkpoint.md b/docs/plans/0074-public-output-checkpoint.md
new file mode 100644
index 000000000..8783c07b5
--- /dev/null
+++ b/docs/plans/0074-public-output-checkpoint.md
@@ -0,0 +1,54 @@
+# AP074 public-output cutover checkpoint
+
+The user requested a committed, pushed pause on 30 October 2026 and resumed the
+complete AP074 goal on 5 November 2026. AP074 remains incomplete.
+The original checkpoint branch preserves unfinished work at `f55aa61a`. It has
+now been reconciled onto `codex/effect-recovery-foundation` over `3cb2caad`; the
+implementation ledger records subsequent integration and evidence. The original
+checkpoint alone is not a qualification or merge candidate.
+
+## Saved work
+
+- Native hiding effect and full journal-record commitments under the existing
+ protected immutable nonce; the nonce is excluded from the sealed message.
+- Public signed outcome `/3`, redacted pre-entry history and app observe `/4`.
+- Operator-only `audit-export`, writing a new private 0600 file. The private
+ signed outcome `/1` retains the complete existing outcome presence predicate
+ and audit checks. This avoids exporting the journal nonce or adding a second
+ journal decoder. No journal schema or entry transition changes are intended.
+- Python/TypeScript public pre-entry types and parser changes, native public `/3`
+ and private `/1` reference-vector constructors, and native sentinel assertions.
+
+## Construction evidence and limits
+
+The actual file-backed signed-public-output privacy constructor passed (1 passed,
+0 ignored, 0.74 s), including the existing private evidence and native-signature
+checks. The new public `/3` vector constructor passed (1 passed, 0 ignored,
+0.04 s). A later small source edit consumed the public stage field and adjusted
+Debug output; it has not been rebuilt. No SDK suite or broad gate ran locally.
+These are artifact-construction checks, not complete acceptance or CI evidence.
+
+## Required before integrating this cutover
+
+1. Reconcile this checkpoint against the completed CI-repair commits on the PR.
+ Preserve the redacted Debug/assertion repairs and updated SDK API snapshot.
+2. Finish native private export integration and all SDK/demo/example/template
+ consumers. Public app history must never carry raw protected pre-entry bytes
+ or private outcomes. Complete private audit bundles at the operator boundary;
+ keep existing offline audit, recount, basis, pre-entry and tamper checks.
+ Do not restore raw public echoes/digests or weaken any existing assertion.
+3. Finish the direct schema/fixture cutover. Remove superseded public `/2`
+ fixtures/readers; keep full private fact fixtures explicitly private. Regenerate
+ all affected native catalogs, CLI inventory and Unix interface artifacts.
+ The new command must be covered by Linux wrong-UID and private-output probes.
+4. Update the specification, compliance inventory, acceptance ledger, binding
+ documentation, source closure, semantic freeze and public API snapshot
+ atomically. The existing generated metadata is stale on this checkpoint.
+5. Generate the required file and TLS PostgreSQL artifacts, commit the coherent
+ change, and give the CI subagent the exact SHA to publish and qualify.
+6. Continue the remaining full AP074 acceptance work from the main status ledger.
+ Do not claim A17 or the entire specification complete from this checkpoint.
+
+AP070 remains the independent application-owned reference. AP073 remains paused
+and untouched. No additional delegation is authorized. The user has resumed the main goal; continue the full acceptance work recorded
+in the implementation ledger.
diff --git a/docs/plans/0074-review-approval-evidence.md b/docs/plans/0074-review-approval-evidence.md
new file mode 100644
index 000000000..7caa0ae74
--- /dev/null
+++ b/docs/plans/0074-review-approval-evidence.md
@@ -0,0 +1,29 @@
+# Native review meaning and exact approval evidence
+
+The `review_approval_tests::native_review_changes_require_fresh_authority_and_approval`
+constructor uses the maintained Airtable and Stripe refund example recipes,
+ordinary native MCP proofs and exact Ed25519 approvals. Five cases independently
+change a displayed label, consequences, destination, financial units or the
+80-byte unit boundary. Every case proves the original native action pins its
+original recipe and refuses the changed recipe; a newly authorized action still
+cannot reuse the old exact approval. Original and freshly approved controls pass.
+
+The complete native literal is retained beside its native signing request in
+`bindings/fixtures/gateway/review-approval-boundary.json`. The constructor compares
+the complete parsed literal to the native private review without truncation and
+refuses an unsupported summary renderer. Markup in a label remains literal data.
+Thirteen financial-review source negatives cover missing/null/empty/whitespace/
+over-limit/control/bidi units, omitted fields, unsupported review version,
+missing/oversized/excess consequences and case-insensitive label collisions.
+
+Trust, keys, connection/account and clock are explicit synthetic fixtures. This
+is signed native compilation/review/approval evidence, not provider execution,
+financial interpretation, durable-store behavior or human attention. A changed
+unit declaration does not establish that a provider honors that unit; qualified
+domain meaning still requires the independently reviewed recipe contract.
+
+The installed owner channel's separate authentication, review/signing correlation
+and complete literal-display tests remain required. The native constructor and
+the installed generic renderer provide complementary evidence; this artifact
+does not manufacture an installed frontend result. Exact-head hosted execution
+remains pending after artifact construction.
diff --git a/docs/product/SELF_HOSTED_CLAIM_LEDGER.md b/docs/product/SELF_HOSTED_CLAIM_LEDGER.md
index 4c31ffc13..53e0c225b 100644
--- a/docs/product/SELF_HOSTED_CLAIM_LEDGER.md
+++ b/docs/product/SELF_HOSTED_CLAIM_LEDGER.md
@@ -158,10 +158,9 @@ The gateway can now hold an observer signing key and sign two kinds of
observation for the application to attach to its next action:
`auths.gateway-readback/1` (the value the gateway read from the recipe's
observed field, and the echo field when present) and
-`auths.gateway-outcome/2` (the stored record of one logical operation: its
-commitment, stage, evaluation time, recipe digest, and, as recorded, the
-counter-set digest, refusal, response status and digest, read-back
-comparison, evidence digest, pre-entry digest, and relative-ceiling basis). A grant's observation requirement can then require, for example,
+`auths.gateway-outcome/3` (hiding effect/recipe/record references, conservative
+stage, evaluation time and bounded status/refusal). Protected full outcome facts
+and pre-entry bytes remain operator-only. A grant's observation requirement can then require, for example,
that the record held the action's `expected` value, or that step N−1 is
`observed-by-provider`, before the verifier authorizes step N. The gateway
verifies at its own clock and exposes top-level verified MCP arguments as
@@ -209,9 +208,11 @@ observation never satisfies; and an observer that is also in the authority
chain is refused. **The live Airtable run has not been done**, no hosted CI
result is cited here, and no document may yet describe gateway grants as
conditioned on observed provider state. The Python and TypeScript gateway
-clients expose the observation request (`observe_read_back`/`observe_outcome`,
-`observeReadBack`/`observeOutcome`) and return the exact signed bytes and
-media type, parsed as a closed result. Both SDKs can now attach a returned
+clients expose local outcome/pre-entry history and independently approved
+recovery. AP074 removes the unapproved application read-back request. Synthetic
+read-back observations for the reference journey now come only from the private
+harness control socket; they do not establish managed acquisition or disclosure.
+Both SDKs can now attach a returned
observation to the next action (`attach_observations`/`attachObservations`,
or `observations` on the authoring functions). The descriptors are bound by
the action signature, and the SDK's final verification reaches the gateway's
@@ -293,13 +294,13 @@ multiplying either, all-or-none reservation, fixed-context tightening, and
decider soundness, with the chain leaves translated and refined. No hosted
CI result is cited here.
-### Outcome and audit `/2` with the provider result (AP-SPEC-063 §8, repository-local only)
+### Public outcome `/3` and private audit with the provider result (AP-SPEC-063 §8, repository-local only)
**Claim.** For every claimed submission the gateway records the stage, the
gateway evaluation time, the refusal code when it refused, and the HTTP
status and response digest when a complete response exists. When an
observer key is provisioned, it signs that record on request as
-`auths.gateway-outcome/2`, with the counter-set digest, the pre-entry
+operator-private `auths.gateway-private-outcome/1`, with the counter-set digest, the pre-entry
digest, and the relative-ceiling basis when recorded; the gateway refuses
to sign, and the auditor refuses to accept, any fact set its stage could not
carry. The offline audit (`auths.gateway-audit-report/3`) re-verifies each
@@ -326,7 +327,8 @@ proof whose outcome was removed. Until AP-SPEC-038 Epic 4 supplies a custody
client, a production gateway signs no outcomes, so every entry audits as
`unverified` and the audit fails.
-Current evidence is `bindings/fixtures/gateway/outcome-v2.json`: Rust
+Public evidence is `bindings/fixtures/gateway/outcome-v3.json`; native private
+full-fact cases remain in `private-outcome-v1.json`. Rust
verifies every accepted vector with its exact facts and refuses every
refused one, and Rust, the Python SDK (native), and the TypeScript SDK
(WASM) reach the same verdict on every case of a grant requiring
diff --git a/docs/research/domains/abstraction-cases/0007-declarative-credential-isolated-request.md b/docs/research/domains/abstraction-cases/0007-declarative-credential-isolated-request.md
index 0ded36cba..48fed3f36 100644
--- a/docs/research/domains/abstraction-cases/0007-declarative-credential-isolated-request.md
+++ b/docs/research/domains/abstraction-cases/0007-declarative-credential-isolated-request.md
@@ -180,7 +180,7 @@ identical subset is extracted; the rest stays with the recipe author or the
vertical. "Vertical" means `product/integrations/auths-stripe` unless a row
names another consumer. The vectors for every capability are in
`bindings/fixtures/gateway/` (`hostile-recipes-v2.json`,
-`attempt-scenarios-v3.json`, `bounds-aggregate.json`, `outcome-v2.json`,
+`attempt-scenarios-v3.json`, `bounds-aggregate.json`, `outcome-v3.json`,
`key-identity.json`, `codes.json`), and tests in
`product/runtime/auths-gateway/src/pending_vectors/` show that current code
does not satisfy them.
diff --git a/docs/research/domains/abstraction-cases/0009-bound-effect-and-recovery.md b/docs/research/domains/abstraction-cases/0009-bound-effect-and-recovery.md
new file mode 100644
index 000000000..cb0ad54b4
--- /dev/null
+++ b/docs/research/domains/abstraction-cases/0009-bound-effect-and-recovery.md
@@ -0,0 +1,970 @@
+# Case 0009: Bound effect identity and durable recovery
+
+Status: AP074 implementation contract; behavioral and performance evidence pending.
+Owner: product gateway/kernel, per [ADR 0017](../../../adr/0017-bound-effect-and-recovery-ownership.md).
+
+## Baseline consumers and dispositions
+
+| Surface | Current semantics | Disposition |
+| --- | --- | --- |
+| Native `VerifiedAction` | Canonical action, proof/context/plan, selected branches and evidence | Retain sole native authority source |
+| Gateway `verify_detailed` / `VerifiedCommand` | Native projection, bounds, compiler output | Consolidate immutable effect here |
+| `recipe.rs` / kernel `construct` | Closed deterministic bytes, credential slot, observation plan | Retain compiler; bind entire output |
+| `submit.rs` / kernel `order` | Write ordering, credential/preflight reads, replay observation | Extend current driver; make all external entries explicit |
+| Gateway store / kernel `transition` | Claim and counters atomic; `Attempting` conflates reservation and entry | Replace conflation with explicit durable grant |
+| Kernel `recovery` / `outcome` | Response/state match/provider link are distinct; no lost-claim retry | Retain truth and no-retry semantics |
+| File attempt adapter | Private files, flock, fsync/rename, batch rollback, CAS | Retain durability; add bounded transactional validation/fencing |
+| PostgreSQL gateway records | Atomic batch, byte CAS, commit acknowledgement | Retain mechanism; extend shared transaction boundary |
+| Lifecycle `operation` / `sealed` | Preparation commitments, validated store ACK, profile-specific lifecycle | Reference semantics; no second managed gateway owner |
+| AP070 `AdmissionStore` / `ProviderAttemptLedger` | Application-owned SQLite task counters, challenge/release decisions, per-GET records | Retain independent reference and narrower claims; never authorize managed gateway effects |
+| AP070 `execute_read` / `ReceiptStore` | Application-owned fetch, release orchestration and receipt claims | Preserve broker/reference; do not promote its facts to managed entry/recovery evidence |
+| TS `gateway.ts`, Python native runtime, gateway app/admin | Socket/binding routes and views | Expose native construction and state; no local effect hashes |
+| Managed Airtable owner interface | Authenticated complete native presentation and exact pending consent; separate operator/app origins and UIDs | Demo-only consent transport; no effect hash, signature codec, native decision or lifecycle owner. Exact signing remains operator-held and native admission remains mandatory |
+| AP070 Drive adapter | Application-owned metadata/content GETs and provenance checks | Retain independent reference; any future managed Google integration needs separately reviewed native capabilities |
+| Gateway observer / audit / harness / commissioning | Signed observations, audit facts, local fixtures, bounded operator authority | Update identities and stages atomically; do not bypass commissioning bounds |
+
+## Comparison and extraction classification
+
+| Dimension | Stripe refund | Airtable update | AP070 read/disclosure | Lifecycle reference |
+| --- | --- | --- | --- | --- |
+| Exact action | Verified MCP refund fields | Verified MCP record fields | Verified API read scope | Concrete profile command |
+| Request | POST form, derived idempotency | PATCH JSON, no idempotency | Metadata/content GET then recipient bytes | Domain-defined |
+| Recovery locator | Response-derived refund ID | Verified record ID | Source effect + protected content + recipient channel | Domain-defined |
+| Unknown write recovery | No locator without response | Exact echo at verified locator may add evidence | No automatic refetch/redelivery | May support concrete non-effect proof |
+| Approval | Exact compiled mutation | Exact compiled mutation | Bounded upfront consent then exact release binding | Profile-specific |
+| Accounting | Count and currency-partitioned sum | Count | Calls, provider attempts/response bytes, disclosure bytes | Reservations/obligations |
+| Outcome | HTTP response is not refund proof | Matching state is not causation | Acquisition is not delivery; ACK is recipient fact | Proven partial/applied/not-applied |
+| Shared subset | Immutable identity, atomic claim, single grant, immutable evidence | Same | Same | ACK validation pattern only |
+
+This is an identical-mechanism extraction only for the last row. Provider
+semantics and concrete approvals are divergent. No provider tag, callback or
+untyped map selects a transition. Smaller primitives alone cannot atomically
+couple replay identity, admission capacity and grant publication; the reducer
+and journal transaction therefore share one explicit contract.
+
+The retained oracles are Stripe's reference integration, the Airtable recipe
+fixtures and independently expressed AP070 read/release scenarios. Differential
+tests must compare observable commands, refusals, charges and evidence, not just
+two calls to the new implementation. Source paths include
+`examples/stripe-refund-approval/recipe.json`, `demos/airtable-gateway/recipe.json`
+and `demos/google-employee-delegation/broker/`.
+
+## Encoding, limits and state contract
+
+Canonical effect encoding is native-owned and versioned; duplicate/unknown
+fields, noncanonical encodings and unsupported versions fail closed. Private
+effect identity is domain separated from public receipt identity. Exact approval
+excludes its signature from the value signed. Replay key is namespace, epoch,
+logical operation; effect identity is compared after lookup, never used to evade
+that lookup. Optional task-free scope has an explicit tag distinct from task scope.
+
+| Surface | Existing bound retained / AP074 bound to enforce |
+| --- | --- |
+| Proof / action | Gateway 4 MiB / 64 KiB; native configured bounds also apply |
+| Trusted context | Native configured bound; hard maximum 16 MiB, default 2 MiB |
+| Effect canonical metadata including compiled output | 65,536 bytes; reject before decoding |
+| Namespace / logical operation | Existing canonical ASCII 64 / 128 bytes |
+| Resource / recipient / channel / scope label | One per operation, each at most 512 UTF-8 bytes |
+| Compiler write body | Existing 16,384 bytes |
+| Recipe source / lock | Existing 65,536 bytes each |
+| Path / template | Existing 16 segments, 64 nodes, depth 6; encoded path at most 8,192 bytes, with the separately validated origin bounded by the 65,536-byte recipe source. Content transport additionally caps the full URL at 8,192 bytes. |
+| Headers | Existing recipe maximum two provider headers; total compiled headers at most 8, names 64 and values 1,024 bytes |
+| Selected branches/evidence | Native configured collection bounds; total effect encoding bound also applies |
+| Reservations | Existing maximum 34 counters; typed units and exact window |
+| Provider response / private retained bytes | 65,536 bytes per operation; content not retained by default |
+| Pre-entry observations | Existing 4, each native maximum 4,096 bytes |
+| Journal record / atomic batch | Existing 262,144 bytes / 64 entries |
+| Events / ACK | 64 durable operation revisions, including admission; one fixed typed snapshot, no unbounded event list / 4,096 canonical bytes per ACK |
+| Exact human approval | 4,096 canonical bytes; at most 3,600 seconds, with exclusive expiry |
+| Installed effect policy | `MAX_EFFECT_POLICY_BYTES`: 4,096 canonical bytes before decoding; partition at most 512 bytes; explicit nonzero safe-integer epoch; valid nonweak Ed25519 approver |
+| Private human review | 131,072 bytes including exact request and binding; never a public receipt |
+| Demo owner HTTP interface | One pending review; eight handlers; 1,024-byte commands; native 131,072-byte envelope; five-second I/O, 900-second session and 600-second consent wait; no native expiry extension |
+| Demo private review sockets | Existing source-only requests remain 1,024 bytes; separate mutation proof/action frame at most 4 MiB, followed by native proof/action limits; finite closed labels/attempt counts |
+| Observation attempts | At most 16 explicitly admitted linked operations |
+| Transaction lock / fenced decision | Separate two-second lock and cooperative decision bounds (`auths_stores::{GATEWAY_LOCK_TIMEOUT,GATEWAY_FENCE_DEADLINE}`); sample native time after lock; refuse before publication on deadline |
+| Native verification work | Native configured work-unit limit plus transaction deadline |
+| Driver | Existing 64 steps |
+| Transport | Existing 15-second maximum; no redirects or write retries |
+| Claim contention / file recovery | Existing 32 claim rounds / 8 recovery rounds |
+| Recovery inventory / expiry maintenance / public evidence | At most 4,096 inventory keys, 16,384 scanned file entries plus one overflow probe; expiry scans at most 4,096 entries and runtime sweep requests at most 1,024 deletions; all stop under the cooperative two-second bound. Signed native observations are at most 4,096 canonical CBOR bytes; the enclosing application protocol has its separate 8 MiB frame ceiling. |
+
+Tighter existing limits win. New limits above are implementation requirements,
+not claims that baseline code enforces them. Bounds must be checked before
+allocation, including persisted files (baseline `fs::read` checks too late).
+
+The reducer permits reserve → refuse-before-entry or reserve → unique grant.
+Provider acquisition/mutation may then record response or unknown; disclosure
+may record exact ACK or unknown. Late exact evidence is additive/idempotent.
+No unknown→not-entered, refund, new grant or automatic retry edge exists.
+Independent observation adds a typed fact without replacing response/ACK or
+original actor. Conserved capacity is available + held + consumed = allocated;
+unknown keeps charges. Pre-entry refusal only releases declared releasable units.
+
+Retained submission lookup is another shared mechanism across the three typed
+protocols. It uses the same canonical native authority projection and retained
+binding/approver pin, rather than recompiling expired authority before returning
+history. The exact proof and action digests are both required; each endpoint
+also binds any additional resource/source/recipient proposal to the historical
+action. The journal's canonical projection reconstructs the existing authority
+commitment and cannot mint a sealed effect or permit. These common checks do not
+merge mutation response, acquisition content and disclosure ACK facts. Reserved
+continuation still uses each protocol's full fresh native compilation and the
+shared writer fence. The direct journal schema cutover is `/16`, with a 65,536
+byte private authority projection limit and no reader for obsolete state.
+
+## Cutover and evidence
+
+The retained replay/maintenance boundary has two separate bounds: native policy
+windows and adapter maintenance work. Expiring count/sum slots may be reclaimed
+without deleting attempts, configuration, commissioning or lifetime allocations.
+File maintenance uses a nonauthoritative iterator (4,096 inspected directory
+entries and a cooperative two-second work budget), not a second journal or a
+full-directory sort. PostgreSQL retains its indexed LIMIT and common writer
+fence. The engine samples the existing deployment/effect clock for both stores.
+File inventory preserves its sorted-prefix contract using a bounded heap and
+separate output/scan budgets (4,096 keys, 16,384 names plus one overflow probe).
+If it cannot finish, it returns unavailable instead of a partial inventory;
+direct journal lookup and all admission/recovery decisions remain independent.
+The source and native retention fixtures are recorded in the ledger; they do
+not establish clock rollback detection, all reserved/mutation histories or
+malicious whole-store rollback protection.
+
+Replace affected formats and call sites in one bounded PR. No production
+compatibility path, migration or dual journal. Reference models remain explicitly
+test-only or independently application-owned with explicit narrower claims.
+They cannot exercise managed credentials or become a second managed authority.
+Formal work must update translated shipping predicates and refinement
+links, not claim a hand-written model proves runtime code. Adapter assumptions
+include lock/commit durability, clock trust, recipient authentication and provider
+observation integrity; whole-store rollback without an external anchor is excluded.
+
+Required executable/formal properties are E01–E05 and R01–R10. Measurements must
+cover verification, compilation, transaction/lock time, bounded memory and scans
+for both recipes and read/disclosure. No measurements have yet been executed.
+Actual results and exact revisions belong in the
+[acceptance ledger](../../../plans/0074-implementation-status.md).
+
+## Adapter fence mechanism
+
+The existing attempt adapters now have one bounded snapshot-and-replace
+operation. Its decision callback is in-process native validation over opaque
+snapshots, not a provider/profile dispatcher or an extension plugin. It runs
+once while the same lock excludes connection replacement, allocation insertion
+and expiry sweeping; it returns one record replacement. No provider or custody
+call may run under the lock, and no error is retried automatically. File
+publication and PostgreSQL COMMIT must acknowledge before returning success.
+
+Both maintained adapters share a 64-record read bound and two-second lock and
+decision ceilings; tighter installed PostgreSQL deadlines are preserved.
+The source conformance suite orders replacement, insertion and sweep writers
+against a held decision and checks failure atomicity. This source has not yet
+been exercised in hosted CI. The mechanism does not itself interpret effect
+identity, grant entry or establish multi-host revocation freshness. The shared configuration record now participates in the fence, retains bounded
+permanent qualification memory and requires explicit operator CAS activation.
+Connecting its snapshot to the one-use grant owner remains P2 work, as specified
+in ADR 0017. Early host checks are not final-entry evidence.
+
+
+The mutation source now connects this primitive to the existing transition and
+submission owners: reservation and durable grant are separate, fixed grant audit
+bytes cannot change, and transport consumes a capability issued only after the
+fenced replacement acknowledges. The same retained command supplies the request.
+Slot membership is checked without reallocation. The native synthetic grant
+corpus proves its narrow file-store sequence; it does not establish installed
+multi-process fencing, all fault boundaries or the still-open recovery cutover.
+
+
+Reservation resumption now remains in that same owner: equal reserved bindings
+reopen unchanged, and grant audit `/2` appends the fresh approval/read facts while
+preserving admission history. Attempt `/6` advances exactly one revision per
+replacement. The synthetic file corpus exercises restart after the initial
+approval/deadline, retained count/sum slots at exhausted capacity and one fresh
+grant. PostgreSQL equivalent source is authored, not executed. These mechanics
+do not establish acquisition, disclosure, independent observation authority or
+installed end-to-end guarantees.
+
+
+Review completeness is now a closed presentation mechanism shared by the
+maintained recipes: typed complete field coverage, bounded labels/units and a
+literal native renderer. Domain consequences and units remain operator-owned
+recipe data; the gateway neither converts units nor infers meaning by provider.
+Recipe/effect identities bind that declaration. Native negative and rendering
+corpora cover missing/type/version/label/duplicate cases and hostile text.
+Installed reviewer authentication, frontend parity and domain qualification
+remain separate gates; the OpenAPI mapper deliberately emits an unusable draft
+until its review facts are supplied.
+
+
+### Replay projection evidence checkpoint
+
+Both mutation families now return granted/terminal history without replay I/O.
+The pure order predicate stops on replay; the runtime compares the full native
+binding before projecting journal state. A new `reserved` result retains the
+pre-grant distinction and bounded revision. Ambiguous acknowledgement never
+constructs a permit, including when a successful reread finds a committed grant.
+
+`bindings/fixtures/gateway/replay-projection.json` is generated by the native
+submission driver and real file adapter with synthetic verified commands. It
+covers concurrent reopened-store replay, no-commit/committed-ack-loss and failed
+reread, exact external counters and malformed reservation projections. It is
+narrow mechanism evidence; installed authority and PostgreSQL fault evidence
+remain open. Existing explicit operator observation tests retain their original
+scope and are not evidence of independent AP074 recovery authorization.
+
+
+## Recovery compiler checkpoint
+
+The native recovery compiler reuses the verified-action, canonical binding,
+exact approval and complete literal-review owners. It does not reuse mutation
+authority or its request type. A protected journal target supplies the source
+identity and closed observation GET; a distinct native reader authorizes that
+exact target and its count/byte/time limits. The reserved native read tool cannot
+be installed as a mutation recipe.
+
+`recovery-compilation.json` is generated from signed native proofs and an actual
+file-backed source record. It covers an expired original writer, a different
+reader, exact read approval, source/contract/limit substitution refusal and no
+source mutation. Its source entry uses a test oracle and sends no provider
+request. It proves compilation only: conserved acquisition charging, fresh
+transactional read entry, credential replacement, installed review channels and
+bounded observation execution remain open evidence gates.
+
+
+## Recovery acquisition mechanism checkpoint
+
+The shared journal now encodes typed mutation/acquisition payloads under one
+header, codec and transition predicate. The concrete recovery-read payload
+binds the full source, request and capacity plan. It does not reuse mutation
+observation facts or introduce a second transition reducer. Permanent allocation
+slots conserve one call and its maximum bytes together; the source effect, not
+the read operation ID, owns capacity. The final existing adapter fence compares
+the exact source/claim/allocation and fresh authority before granting a permit.
+
+`bindings/fixtures/gateway/recovery-entry.json` is constructed by the native
+recovery fixture using signed proofs and the real file adapter. Its entry check
+and provider response are synthetic. It exercises same-reservation replay,
+stale grant refusal, permit loss/reopen, 16 concurrent contenders for 15 remaining
+slots, permanent retention, private response history and mixed-protocol stage
+projection. It establishes neither provider execution nor installed revocation,
+PostgreSQL parity, process crash behavior, source-evidence freshness or disclosure.
+Those remain required independent evidence and cutover gates.
+
+
+## Evidence refinement comparison
+
+The concrete recovery-read response is evaluated against its original closed
+source contract inside the shared source/read/allocation fence. No response field
+selects a provider implementation. The existing typed read-back distinction is
+retained: exact echo plus expected value can link provider evidence; equality
+without that echo cannot resolve an unknown mutation. Duplicate object members
+are rejected instead of silently choosing one value. This parsing hardening is
+a bounded observation leaf, not a new provider-semantic evaluator.
+
+`recovery-refinement.json` covers unusable/missing/multiple/duplicate responses,
+unrelated echo and noncausal equality, freshness boundaries, concurrent/idempotent
+recording, source attribution/charge retention and report privacy. It also freezes
+the closed replacement operator-frame grammar. These are native file-journal
+results with synthetic response, clock and entry validation; independent process,
+installed transport, PostgreSQL and final differential evidence are still needed.
+
+
+## Submission and independent observation boundary
+
+Mutation submission now ends at a recorded response or conservative unknown
+outcome. The shared order leaf contains no automatic read-back, observation
+lease mode or operator resume entry. The driver cannot promote a response into
+provider evidence. Historical observation-writing helpers are crate-private and
+compiled only for tests; the shipping recovery fence owns evidence refinement.
+
+The retained record/store oracle requests observation explicitly and makes no
+AP074 acquisition-authority claim. The native replay corpus separately measures
+one response-producing write with one lease and zero provider reads, no provider
+evidence, and a replay that adds no I/O or journal revision. The scenario corpus
+removes the former second lease, repeated credential checks and post-write read.
+Installed journeys, ancillary read admission and final translated/refinement
+artifacts remain required before this boundary has complete acceptance evidence.
+
+
+## Prerequisite declaration comparison
+
+The Stripe reference contributes account and denied-capability checks plus a
+relative-ceiling read; Airtable contributes the pre-entry read. Compiler `/5`
+retains the existing closed requests rather than reconstructing provider paths.
+The shared fields are exact method/URL/headers, one-call and response-byte
+allowances, transport/freshness bounds and absence of disclosure. Evaluation
+purposes remain tagged: account equality, expected refusal, selected pre-entry
+requirements and relative arithmetic are not interchangeable response meanings.
+
+`prerequisite-compilation.json` uses signed native proofs and exact approvals to
+freeze both plans, including account-scope header placement, resolved bound
+values, exclusion of the onboarding probe, native recompilation and obsolete
+compiler refusal. The Airtable case has no pre-entry observation requirement;
+it tests compilation, not admission eligibility. The fixture has no provider,
+custody, durable acquisition or human-channel claim. Runtime accounting,
+independent differential execution and installed evidence remain required.
+
+
+## Prerequisite journal cutover
+
+The same acquisition transition/adapter mechanism now handles the two compared
+prerequisite plans. Their authority differs from recovery: a prerequisite uses
+its exact parent's approved enumeration while the parent remains reserved;
+recovery uses an independent read proof/approval and an eligible unresolved
+source. A closed plan variant and source check preserve that distinction.
+Neither path decodes audit data into a transport request or grants disclosure.
+
+`prerequisite-entry.json` records file-journal histories for both compiled plans:
+missing reads, recorded responses, lost permits and stale responses. Equal
+reservations retain their charges, only one concurrent handle wins entry,
+reopening cannot remint a permit, response times are preserved, and final write
+entry requires fresh recorded child facts and their original allocations.
+Membership-witness substitution is rejected. Grants use an explicit-time oracle
+and responses are synthetic; no provider calls occur. The Airtable compilation
+case still has no native pre-entry requirement and does not establish installed
+mutation eligibility. Differential semantic evaluation, actual transport,
+process faults, PostgreSQL and tighter observer freshness remain exit gates.
+
+
+## Prerequisite evidence at entry
+
+The final gate does not add an alternate response evaluator: it calls the same
+account/denied predicates, recipe ceiling reader, ratio kernel, signed observation
+verifier, requirement selection and native freshness/condition predicates already
+used by the maintained paths. Journal membership provides the response and its
+original time; the installed gate checks the resulting evidence under the writer
+fence. Existing synthetic store oracles explicitly omit installed authority and
+remain test-only.
+
+`prerequisite-evidence.json` adds 15 native signed Airtable cases, including a
+2-second selected freshness bound within the 30-second acquisition ceiling.
+`prerequisite-guard-evidence.json` adds 15 Stripe response/checkpoint cases. The
+positive Stripe input now has a known synthetic account commitment, allowing an
+actual matching account check rather than an unopenable placeholder. These are
+executed predicate fixtures; their scope excludes installed process/transport,
+transaction lock waits, PostgreSQL and full differential qualification.
+
+
+## Ordered initial admission
+
+Mutation count/sum claims and recovery/prerequisite acquisitions now use one
+bounded snapshot-and-insert mechanism under the existing writer fence. The
+adapter owns atomicity only; the gateway reuses its native entry validator to
+establish authority before publishing any claim or charge. Mutation sum
+predecessors and acquisition source eligibility remain typed, distinct checks.
+No schema, allocation algebra, retry owner or transport reconstruction is added.
+
+The file artifact `effect-admission.json` covers expired-approval refusal with
+no claim/capacity writes, admission-time stamping, immutable replay after expiry,
+retained count/sum charges, missing/changed dependencies, batch collisions and a
+writer-before-admission dependency order. Authority uses the explicit-time
+approval oracle; opaque dependency bytes exercise adapter ordering, not native
+revocation semantics. The same conformance source is authored for PostgreSQL.
+Installed authority races, process faults and performance remain required.
+
+
+## Application provider-read cutover
+
+The application `read-back` request previously had neither native proof nor
+exact approval and could make credential/account checks and a provider read.
+It is removed from production, the shared application frame, CLI and SDKs.
+The existing native recovery engine now also serves a closed application frame
+with separate proof/action/approval. It preserves source/read distinctions and
+returns only the already-defined redacted report. The same bounded decoder and
+engine serve operator and application callers; no second decision owner exists.
+
+`application-recovery-frame.json` constructs real Unix socket decoder/dispatch
+histories with a counting refusal oracle, including obsolete schemas, omitted
+or malformed approvals, invalid source IDs and transport substitution.
+`application-recovery-cli.json` captures the production CLI's exact request and
+response over a real socket. These are transport artifacts, not installed proof,
+provider, disclosure or end-to-end acceptance evidence. The independent old
+observation evaluator remains confined to test-only reference/control paths.
+
+### Managed read reservation extension
+
+Mutation count/amount and recovery call/response-byte budgets share checked
+fixed-window arithmetic, ordered slot candidates, immutable cumulative sums,
+atomic insert and final membership validation. They do not share units or counter
+identities. Recovery sums mean the complete approved response ceiling; mutation
+sums keep their recipe-defined argument. Source-lifetime recovery allocation is
+an additional independent limit. Prerequisites charge their declared child
+allowance without repeating the parent's managed spend reservation.
+
+This extension reuses the existing store mechanism rather than adding a budget
+owner. The admission transaction includes every managed slot and the source
+allocation; replay and contention create no partial charge. The native recovery
+corpus must cover count/sum exhaustion, retained unknown charges, window changes,
+source/allocation collisions and missing/substituted membership. Full adapter
+concurrency and performance acceptance remain open until executed evidence is
+recorded in the AP074 ledger.
+
+### Credential prerequisites of independent recovery
+
+The account/denied-capability subset is identical between mutation and recovery:
+closed recipe reads, expected installed account, refused HTTP statuses, version
+checks, one invocation and full response ceiling, durable receipt, and 30-second
+conservative evidence age. Both paths now use the same compiler subset, journal,
+permit-consuming acquisition helper and native evidence predicates. Mutation
+keeps its additional pre-entry/relative-ceiling semantics. First-install probes
+and candidate credentials remain an independent authorization obligation.
+
+A recovery parent's request commitment authenticates its complete prerequisite
+witness. The existing child membership decoder accepts the closed mutation or
+recovery parent protocol and refuses disclosure; the source fence refuses a
+child as a parent. Final entry requires all child evidence under the same fence.
+The six-case recovery prerequisite corpus records the native predicate result
+separately from oracle journal grants, so synthetic wrong-account/capability
+histories cannot masquerade as installed entry-denial evidence. Installed
+transport, PostgreSQL and final formal/performance gates remain open.
+
+
+## Candidate credential compilation checkpoint
+
+Candidate first-install and rotation checks now reuse recovery's native MCP
+verification, branch/observer separation, managed read-bound admission and
+bound-effect construction. The target comparison stays distinct: recovery pins
+a retained mutation, while candidate compilation pins a staged descriptor and
+exact predecessor/absence. The existing recipe owns probe/account/capability
+meaning. Ordered child plans reuse the closed prerequisite declarations; the
+last actual request is the primary acquisition. No callback or aggregate
+provider response is introduced. Probe parsing now uses the same bounded,
+duplicate-rejecting decoder as the other native evidence predicates.
+
+`candidate-compilation.json` compares first install, active/disabled rotation,
+HEAD-final and probe-only plans using signed native actions. It records full
+review and substitution refusals, exact approval/reverification, stable lifetime
+keys under operation/time changes, and refusal through the existing mutation,
+recovery and orphan-child routes. It is compiler/predicate and route-refusal
+evidence only. The candidate journal admission, staged custody, ordered current
+or absence validation, transport and publication cutover remain required; this
+checkpoint does not promote an operator attestation into native read authority.
+
+
+## Candidate acquisition runtime checkpoint
+
+The candidate plan is now another closed source alternative of the existing
+acquisition owner, with connection/predecessor eligibility kept distinct from
+recovery's source mutation. Physical dependency deduplication preserves both
+source and native checks on the same snapshot. Native authority, qualification,
+exact approval and workload/profile checks remain in `entry.rs`; disabled
+candidate rotation is explicit and cannot authorize an ordinary mutation.
+The shared acquisition runtime consumes the same permits and retained-response
+predicates. There is no new reducer, provider callback or public content path.
+
+`candidate-journal.json` records actual native fenced validation with signed
+proofs and real file persistence for first install, active/disabled rotation,
+HEAD-final and probe-only plans. Histories cover absent/changed predecessors,
+changes before child/main grants, missing/bad/stale child evidence, lost main
+permits and restart. Cross-proposal capacity checks preserve the credential's
+first byte ceiling. Configuration and time are test-controlled; provider bodies
+are synthetic. These artifacts do not establish installed provider execution,
+PostgreSQL conformance, final-check acceptance or credential publication.
+
+
+### Retained candidate evidence at publication
+
+The candidate publication change composes the existing connection insert/replace
+fence, typed acquisition evidence and concrete credential-purpose predicates.
+It adds no acquisition phase, callback-based lifecycle or provider-specific
+publication policy. First install checks absence; rotation checks exact previous
+bytes and preserves administrative state. The common native fence validates the
+same proof, approval, configuration and qualification used at entry, while
+retained primary/child grants bound evidence age. Exact already-current state is
+reported without asserting fresh evidence or attributing publication to a retry.
+
+The native file corpus adds successful publication/replay and refusals for bad
+final purpose/version, unknown primary, stale primary/children, expired approval,
+changed configuration and changed predecessor. Acquisition snapshots and charges
+remain unchanged across publication. This evidence remains narrower than process
+faults, installed custody/provider execution or PostgreSQL conformance.
+
+
+### Staging and local custody serialization
+
+The staged rotation path preserves the existing concrete connection/credential
+relationship. It records the cleanup generation, stores without overwriting a
+candidate, and hands its private proposal to the shared native read owner.
+No provider semantics move into custody. The operator CLI directly uses native
+call compilation, review, acquisition and publication; unapproved rotation is
+limited to recipes with no network checks. Published-secret join is exact
+credential confirmation without acquisition.
+
+The persistent file credential store now orders operations with a stable
+cross-process sidecar lock and reloads bounded state. This is an existing store
+mechanism correction needed by independent operator/runtime processes, not a
+new lifecycle owner. The production CLI corpus drives two competing staging
+processes and a held-lock refusal. Native staging evidence checks stale handle
+refresh and conflicting successor refusal. Broader fault and adapter evidence
+remains required before AP074 acceptance closes.
+
+
+### First-install source cutover
+
+Installation now uses the same sealed candidate and native acquisition owner as
+rotation. Its concrete difference is the expected absence of a live connection
+and an explicit generation-one host floor. The private draft is the existing
+canonical connection record, not another lifecycle format. The cleanup journal,
+custody reference commitment and floor retain their existing owners. Only the
+operator staging action initializes missing installation state; managed candidate
+execution checks it and cannot repair it.
+
+The raw onboarding I/O wrapper is removed from production exports and CLI calls;
+its pre-cutover evaluator is retained under `cfg(test)` as the closed reference.
+The production `install` command stages read-bearing recipes without publication.
+Native file evidence and independent CLI restart evidence cover exact restaging,
+conflicting secret refusal, explicit absence, cleanup ordering and floor checks.
+No provider predicate is inferred from successful staging, and these artifacts
+do not close the installed approved provider, fault or PostgreSQL obligations.
+
+
+### Approved candidate process evidence
+
+The production CLI now has a six-case process/HTTP artifact, generated by
+`tests/candidate_journey.rs`. Native test authoring creates real proof/action
+bytes with explicitly public fixed keys; the runtime still performs native
+verification, exact review/approval, durable acquisition and fenced publication.
+Separate CLI processes exercise first install, active/disabled rotation, replay,
+competing acquisitions, invalid/reused approval, failed probe/account/final
+capability and response loss. Actual external request counts and byte-preserved
+journal/allocation snapshots provide evidence independently of intermediate
+runtime flags. The local provider remains a domain-specific test reference;
+no provider branching or test authoring enters the shipping default runtime.
+
+This strengthens the existing mechanism comparison without adding a production
+abstraction or claiming human presence, deployment isolation, PostgreSQL faults
+or live provider qualification. Maintained deployment/demo consumers still need
+the exact authority/approval and qualification bootstrap cutover.
+
+
+## Recipient evidence and disclosure reducer checkpoint
+
+The shared predicate now admits only the disclosure edges reserved→not-entered,
+reserved→entry-granted, entry-granted→unknown/acknowledged and
+unknown→acknowledged. Fixed allocation bytes and grant bytes survive every edge;
+ACK is a separate terminal fact. Exact repeated evidence is a read-only lookup.
+Acquisition/mutation cannot carry ACK bytes, and disclosure cannot carry HTTP or
+observation facts. Existing provider outcome signing refuses the ACK stage.
+
+The native Ed25519 ACK contract binds every release dimension and exact protected
+content length, using the existing raw-key principal derivation, canonical codec
+and domain-separated commitment. It adds no token system, recipient registry or
+provider-specific branch. Recipient evidence is historical and independent of
+current execution authority; the future entry fence must separately establish
+that authority before releasing any content.
+
+`disclosure-acknowledgement.json` is generated from real native signatures and
+the shipping transition predicate: 576 same/cross-protocol stage pairs, four
+histories, changed field/signature/encoding cases and read-only replay. Its
+retained content identities and journal projections are synthetic. It is neither
+AP070 differential equivalence nor a durability, actual delivery, authenticated
+channel or process-isolation claim. Those remain required for the direct consumer
+cutover. Kani source is extended, not executed; translated Lean/refinement work
+remains pending.
+
+
+## Content acquisition compiler comparison
+
+AP070 separates exact resource selection, bounded UTF-8 response validation and
+later release. Its provider-specific metadata predicates (ID, trash state,
+download capability and MIME-dependent endpoint) are additional domain semantics
+and are not erased by the new content declaration. They still need reviewed
+closed declarations/reference evidence before the AP070 runtime cutover.
+
+The identical subset implemented here is a fixed-origin GET over verified
+resource/account values, literal query parameters, and bounded UTF-8 or one
+JSON-string projection. Path/header/percent construction and JSON duplicate
+rejection reuse the existing gateway mechanisms. The new declaration has one
+resource field, sixteen path segments, eight sorted unique query pairs, 64-byte
+ASCII query names, 512-byte query values, the existing 8,192-byte URL bound and
+65,536-byte response/content ceilings. It rejects response locators and operation
+IDs as content resources. Query values are encoded data; no callback interprets
+them. The ADR's narrow primitive exception applies to this construction subset,
+not to Google metadata semantics or to a general provider qualification claim.
+
+`content-compilation.json` uses native signatures, exact approval and recompile
+for both projections. It exercises resource/installation/limit changes,
+operation-independent and administrative-metadata-independent capacity keys,
+expired/reused approval refusal, operator/native review, malformed source/query/
+pointer rejection and all six typed response failures. Projection copies exact
+bytes without truncation into a zeroizing buffer and gives no disclosure permit.
+This evidence does not establish durable source allocations, actual provider
+transport, content custody, standalone read-only deployment, AP070 differential
+parity or a recipient release. Those remain part of P3/P4.
+
+
+## Content acquisition durable comparison
+
+The shared mechanism is the existing acquisition admission/entry/replay order,
+with the connection row as source and the compiler's resource capacity keys.
+Content keeps a distinct response carrier: status, version agreement, raw bytes,
+receipt time and bounded media/encoding facts. Credential-purpose evaluation and
+mutation observation cannot consume this carrier. It establishes neither an
+AP070 metadata predicate nor recipient delivery. Fixed reviewed query parameters
+are sent only through a sealed content permit, using the same pinned-origin,
+no-proxy/no-redirect transport policy and explicit identity encoding.
+
+`content-journal.json` is constructed by native signed authority/approval,
+installed entry validation and the file adapter, with eight real local HTTP
+requests. It checks exact query/Accept/encoding headers, raw-body bounds,
+duplicate/missing/encoded/wrong-media response metadata, retained HTTP denial,
+one-use grants, reopen replay without record changes, permanent charges after
+lost permits and refusal of attempt seventeen. A typed-record downgrade is
+rejected. Synthetic credential/provider data stays private; the artifact contains
+only public snapshots and synthetic response header facts. This is not an
+independent-process, PostgreSQL, full engine, protected-content/disclosure,
+AP070 equivalence or isolation claim. These evidence gates remain open.
+
+
+## Protected projection comparison
+
+The durable content owner now invokes the same bounded projection leaf as the
+compiler corpus after checking retained version and encoding facts. The domain
+boundary remains unchanged: no Google file-ID/trash/canDownload/MIME predicate is
+inferred. JSON-string projection is data-only and duplicate rejecting; UTF-8 is
+exact and never lossy or truncated. Ready content is reconstructible from the
+immutable raw response and contract, while refusal stays a distinct retained
+fact. This performs no additional provider read or recipient delivery.
+
+The hiding commitment is a direct composition of the pinned HMAC-SHA256 primitive,
+native canonical tuples and native domain commitments. Its private per-source
+key and source/projection/length/content binding are explicit. It does not reuse
+the unrelated MCP execution-reference format or turn a receipt-protection port
+into a semantic owner. The protected journal remains a deployment assumption.
+`content-journal.json` version 2 covers fifteen actual HTTP responses plus lost
+permit/exhaustion, typed response refusal, key/tag/length/body/metadata mutation,
+foreign-source material, old-schema refusal and unchanged replay. Random keys
+and tags are never emitted into the canonical fixture; only checked facts and
+public redacted statuses are stored. This is still narrow file-adapter evidence,
+not PostgreSQL, independent-process or AP070 differential qualification.
+
+### Retained source to release compiler
+
+The native disclosure compiler now consumes sealed content provenance loaded
+from the acquisition journal. Unlike provider acquisition, it has no credential
+prerequisite or provider request. Unlike a mutation, it binds an exact retained
+projection to a recipient/channel/instance and uses disclosure count/byte units.
+The identical mechanisms remain native verification, selected dependencies,
+canonical binding, complete review, exact approval and full recompilation.
+Managed policy arithmetic is reused under closed, disjoint protocol domains.
+
+The source lifetime is the retained acquisition, with sixteen delivery counts
+across recipient/channel/operation changes and exact-length byte declarations.
+This is a bounded implementation contract, not an inferred provider semantic.
+The content journal corpus provides actual source transport plus separate
+release compilation/ACK comparisons for empty and nonempty content. Runtime
+entry, capacity enforcement and recipient-channel behavior remain unqualified;
+no generic workflow/callback or new ledger is introduced by this extraction.
+
+The source allocation leaf is now physically shared: protocol owners supply
+count/byte units and source commitments; the existing capacity module owns the
+append-only row encoding/key and occupied-row equality check. No policy or
+source-eligibility callback is introduced. Required content/recovery native
+artifacts regenerate byte-for-byte unchanged after this extraction; disclosure
+compilation uses the same bounded reservation arithmetic with disjoint units.
+
+### Disclosure journal instance
+
+Disclosure now supplies typed state/plan projections to the same journal protocol
+and kernel reducer as acquisition and mutation. Source provenance is checked under
+the existing writer fence, alongside selected native dependencies and permanent
+source/fixed-window slots. The count/sum arithmetic, immutable allocation encoding,
+replay-key comparison and CAS transition owner remain shared. Provider evidence and
+recipient ACK bytes remain divergent typed fields; no generic success outcome is
+introduced. The native engine reserves and delivers through the sealed Unix
+channel boundary below.
+
+The file corpus adds real-source admission/exhaustion/replay and decoder mutations.
+A synthetic retained-grant assumption enables real ACK persistence/concurrency and
+late-history comparisons; it is explicitly not live recipient authentication or
+transport evidence. The original corpus does not establish transport; complete differential and
+performance gates remain outstanding.
+
+
+### Authenticated local disclosure instance
+
+The new Unix transport reuses kernel UID observations and pinned cryptographic
+primitives, adding a closed fresh challenge over the exact native delivery
+statement. It does not repurpose provider response evidence or generic exchange
+server authentication as recipient possession. A proposed channel policy and a
+serialized historical transcript are data; only the owned authenticated stream
+can participate in the final native fenced decision. Only its acknowledged grant
+constructs a one-use content permit. The source projection and protected-content
+validator remain the acquisition owner; recipient code cannot supply a second
+plaintext or destination to the send function.
+
+The identical shared leaves remain native verification/recompile, exact approval,
+writer fencing, source/counter allocation membership, durable acknowledgement and
+immutable entry/evidence transitions. The distinct disclosure instance fixes a
+local UID/key/channel handshake and exact byte length with explicit bounded
+framing and deadlines. It adds no provider dispatch or policy callback. The real
+socket-pair/file artifact exercises ACK loss, late evidence, fresh-authentication
+replay, dropped permit, one-winner concurrency and dependency changes before the
+final fence. Independent processes/adapters, complete recipe differential and
+performance evidence remain required by the extraction and AP074 exit gates.
+
+
+### Application and thin-client cutover
+
+Content application inputs no longer supply a resource map beside the signed
+native action. The acquisition compiler's existing body is shared below sealed
+verification, allowing content target reconstruction from that same verified
+command. This is one compilation owner; old explicit-target callers and new
+native-action callers produce identical bindings in the native corpus. Native
+history and ACK recording are exposed without capability reconstruction.
+
+The application control switch invokes existing native admission and then hands
+one owned socket to the existing authenticated delivery owner. It does not infer
+readiness, authorization or receipt from a transport label. Rust client/CLI code
+alone executes recipient authentication and signs ACKs after receipt. Python and
+TypeScript carry bounded envelopes and native result bytes; no shared semantic
+predicate, hash, signature or capacity rule moves into those wrappers. Frame,
+CLI refusal and native-client/file evidence have explicit narrow scopes. Full
+installed consumer, reference differential and isolation gates remain required.
+
+
+### Native authoring and process composition evidence
+
+The content/disclosure authoring commands reuse the native target constructors
+and canonical MCP encoding. The existing bounded duplicate-free JSON decoder is
+reused only for mechanical parsing; resource meaning remains in the reviewed
+recipe capability. No new recipe operation dispatch, hash owner, policy callback
+or transition rule is introduced. Test-only trust authoring now accepts an
+explicit bounded same-service tool set, while each real signed action/grant has
+its exact permission and still passes native verification.
+
+`content-process-journey.json` adds real CLI/daemon composition over the file
+adapter: separate acquisition/release, actual recipient ACK, restart no-redelivery,
+crash after recipient receipt before ACK, late evidence, approved write and
+independent read-back, and separate-process disable before release. The artifact
+records three actual provider requests and redacted native facts. Public fixed
+keys simulate actors; neither installed UI/language integration, isolated hostile
+agent containment nor PostgreSQL/full fault coverage is inferred. Differential
+reference and bounded performance exit gates remain open.
+
+
+### Content-only installation
+
+The absence of a mutation declaration is now represented by a closed compiled
+`ContentOnly` alternative containing no `WritePlan`. It shares existing source
+credential/header/resource validation and native content compilation; it does
+not derive a read from a dummy write or add a provider operation branch. Only the
+reserved native read tool is accepted, content is required, and all mutation
+adjuncts are refused. Mutation construction is impossible for this compiled kind.
+Review/audit expose absent mutation method/recovery rather than a fictitious
+recorded-write capability. Existing mutation canonical source/review identities
+are preserved by the native artifact comparisons.
+
+A generated negative corpus and real local CLI/daemon install/read/disclosure
+journey establish the narrow capability and its mutation refusal. The installed
+trust in that journey contains only read/release permissions. Existing mixed
+read/write process composition still passes. No new provider capability, Google
+metadata interpretation, AP070 policy semantics, formal theorem or adapter
+assumption is extracted by this change; those complete gates remain pending.
+
+
+### Concrete Drive metadata reference before extraction
+
+The AP070 Drive reference now separates bounded metadata validation from token
+custody, transport, admission and release. The existing reader invokes that same
+pure domain function and selects its existing export/media path from a closed
+representation enum. No native capability, account approval, budget rule or
+provider callback is inferred. Missing/mismatched identity, download restriction,
+MIME selection and optional size/trashed behavior are frozen in
+`demos/google-employee-delegation/contracts/examples/google-metadata-reference.json`.
+The pure function rechecks the existing 16 KiB bound and returns typed denial for
+malformed MIME values or decimal conversion overflow rather than runtime errors.
+
+The reference records historical last-member-wins JSON, alternate encoding,
+nonboolean trashed values and Unicode decimal acceptance as explicit divergences:
+native bounded UTF-8/typed decoding must refuse them. It does not authorize
+weakening native parsing to match a weaker historical reference. Status/media,
+lease, native resource authority, prerequisite freshness and actual content entry
+remain separate gates. Native metadata capability/refinement, request selection,
+refresh custody and AP070 migration are deferred future integration. This
+provider reference is retained independently, not proof of managed Google access.
+
+
+### Owner decision: retain AP070 independence
+
+AP074 closes the general-purpose managed gateway contract. Its single-owner
+requirement is per managed operation, not a requirement that every independent
+application use the gateway. Preserve AP070's Python broker, SQLite journal,
+credential custody and narrower receipts as an independent application-owned
+reference. Only shared, explicitly comparable contracts participate in
+differential evidence; document differences instead of forcing equivalence or
+rewriting the oracle to call the implementation it tests. Native read/disclosure
+must still have an independent reference corpus; this decision does not waive it.
+
+The observed source separation is concrete: AP070's compose starts its broker
+and relays with its own SQLite state and token vault; its agent transport calls
+its own execute_read, not GatewayClient/application sockets. The gateway does not
+load the Google adapter. These are source/configuration observations, not proof
+of every deployed isolation arrangement. Required managed hostile-process tests
+must enforce credential/channel separation and reject alternate managed paths.
+Sharing provider credentials, trusting AP070 receipts as native evidence or
+letting it authorize the same managed effect would violate the boundary.
+
+Native acquisition/disclosure, full installed managed journey/UI, two mutation
+recipe differentials, every file/PostgreSQL concurrency/crash/fault case and all
+formal/performance/hosted gates remain mandatory. Google metadata, OAuth refresh
+and migration are deferred until explicitly productizing managed Google access.
+
+### One active epoch per managed partition
+
+Epoch activation reuses the existing canonical configuration record, operator
+review token and bounded writer-fence CAS. Configuration key `/2` excludes epoch,
+so changing an epoch cannot create a second independently active control for
+the same namespace/partition. Record `/2` binds the active epoch and rejects a
+lower candidate even with a fresh review token. Permanent qualification memory
+is retained; operation replay keys remain epoch-scoped. Private manifest `/7`
+rejects obsolete disposable installations without compatibility readers.
+
+The native file and TLS PostgreSQL constructors exercise the same signed proof
+across epoch activation with two engine/store instances and real HTTP. They
+refuse cached old-host entry and old approval under the new epoch, accept a new
+exact approval, preserve both histories and all replay charges, and assert source
+lifetime allocation slots `[0, 1]` across epochs. They do not establish every
+race schedule, hostile-process containment, or an independent rollback anchor.
+No new semantic abstraction, provider branch or admission journal is added.
+
+### Installed owner/agent content composition
+
+The local Airtable owner demonstration consumes the already admitted content
+capability through a separate data-only `managed-recipe.json`. Its fixed-record
+GET shares the native partition/control/journal with the reviewed PATCH; it does
+not turn the independent reference into managed execution or change that
+reference's recipe. Native recipe review construction bounds response/content
+to 1,024/512 bytes. No new shared semantic extraction is proposed.
+
+Public SDK frozen dataclasses represent native-authored content/release calls;
+they do not interpret provider fields or select lifecycle transitions. Both
+installed language clients call native acquisition. The native recipient CLI
+receives into app-UID private files and owns key authentication, channel, binding,
+exact bytes and ACK. History polling distinguishes ACK sent from ACK retained,
+performs no external I/O and cannot issue another permit. Public HTTP exports
+only native redacted facts and the already reviewed proposal.
+
+The browser scenario authors same-actor acquisition/disclosure/write, separate
+authenticated owner consent, real provider counters and native Unix delivery,
+private sentinel checks, replay without new calls/files, independent recovery
+and actual unknown response-loss presentation. It is not yet hosted evidence.
+The development simulator retains a fixed actor key copy before app ownership
+changes, separately from exact reviewer keys; production/exclusive signing
+custody, human attention and hostile-OS claims remain excluded. The maintained
+AP070 reference and every remaining AP074 gate are unchanged.
+
+### Native mutation/disclosure capacity and retention histories
+
+New test-only vertical histories consume the existing approval, admission,
+transaction, grant and sweep owners. They introduce no shipping capability or
+semantic extraction. Plain Stripe mutation count/amount-sum and content-release
+count/byte-sum each have count-only, sum-only and combined cases on file/TLS
+PostgreSQL. A real prefix precedes two independent OS contenders. Raw records,
+exact native mutation bodies, actual authenticated recipient content/ACK, and
+provider/custody counters establish the corresponding unit conservation.
+
+Each protocol also has reserved-state expiry controls. Mutation kills the native
+worker after durable reservation ACK; disclosure calls native reservation-only
+admission. Original proof and approval times remain eligible after a short window
+expires, but the new window changes the exact binding and requires fresh approval.
+Only expired slots are removed; attempts and lifetime allocations remain intact.
+Reopen/replay creates no grant, resend, redelivery or new recorded charge. These
+are actual local constructor histories, not new-head hosted evidence, malicious
+rollback detection or hostile-process containment. The disclosure contenders
+each host gateway/recipient tasks in one process over real same-UID Unix sockets.
+
+The pending acceptance gates still include dependency-race/adverse schedules,
+supported floors, complete public binding/containment negatives, generated
+concurrent histories/formal links, bounded performance and final hosted results.
+AP070 remains independently application-owned; AP073 is untouched.
+
+Native scope histories additionally compare ordinary verifier authorization with
+managed compiler refusal. Fresh signed acquisition/disclosure calls carrying
+unsupported delegation, participant or charge fields never reach admission;
+independent operator partitions sharing one adapter cannot exchange approvals,
+history or disclosure provenance. Original exact acquisition and recipient ACK
+are positive controls. These test-only histories use existing owners and create
+no reusable semantic mechanism, task implementation or second decision system.
+The evidence covers file/TLS PostgreSQL engines and local HTTP/same-UID sockets;
+installed language boundaries and hostile OS containment remain separate gates.
+The cross-recipe history exposed a missing provenance-getter recipe check.
+The engine now refuses a foreign recipe before returning its disclosure proposal;
+existing compilation/entry checks remain. The positive replacement uses native
+generation-incrementing connection rotation and the existing CAS, checks that
+activation alone is insufficient, and requires a new operation rather than
+overwriting retained facts. No new semantic owner is extracted.
+
+Mutation entry-order histories reuse only the test scheduling helper from the
+native acquisition matrix. The helper now accepts an exact operation key and
+recognizes the existing typed mutation/acquisition entry-granted wire shapes.
+Native compilation, approval, admission, time checks, dependency snapshots and
+grant persistence remain the existing production owners. Separate OS workers and
+operators order actual file/TLS PostgreSQL locks around revocation, configuration
+activation and independently expiring proof/approval intervals. This is additional
+vertical evidence, not a shared shipping abstraction or new lifecycle framework.
+
+Disclosure extends those eight order/time schedules through its own typed source,
+recipient authentication and native ACK. The source is an actual bounded HTTP
+acquisition. Zero content/ACK on final-grant refusal and exact authenticated
+delivery on success are measured independently of provider responses. Same-UID
+gateway/recipient tasks run within a worker process distinct from the operator.
+This adds test-only evidence for the existing disclosure owner, while retaining
+the distinct installed-language, selected-dependency and hostile-process gates.
+
+Selected observation and threshold controls now reuse the acquisition testkit's
+native signed authoring and before/after binding assertions for mutation and
+disclosure. Only test-module visibility changes; no shipping abstraction or
+semantics are extracted. Full native verification, exact-effect comparison and
+the existing final entry fence establish parity when a 1-of-2 proof remains
+Authorized but its selected evidence changes. Native mutation request echoes
+remain bound to the newly signed proof. All earlier order/expiry schedules are
+preserved, including independent wire-byte capture for refused disclosure.
+
+The native mutation/disclosure transaction-prefix fixtures reuse the existing
+owned-schema PostgreSQL trigger/sequence controls and protocol-specific authoring.
+All attempt and unit-specific charge rows roll back together. Disclosure preserves
+its pre-existing source and charges and requires independent recipient authority
+after a fresh native retry; mutation has no disclosure lifetime allocation. These
+are test-only vertical histories using the existing transaction and journal
+owners. No provider branch enters runtime, no new abstraction is extracted and no
+reference implementation becomes a managed decision system.
+
+Real COMMIT-reply loss now uses the acquisition harness's opaque TLS proxy and
+server observer for mutation and disclosure too. The shared test observer selects
+only typed record paths/operation keys and preserves kind-plus-key snapshots;
+it has no decision or permit authority. Mutation response, disclosure ACK and
+native Unknown persistence keep their distinct facts. Existing production journal,
+grant and historical ACK owners are unchanged. Late recipient evidence may
+refine Unknown, while lost grants and terminal replay cannot authorize I/O or new
+charges. The additional six mutation/eight disclosure histories introduce no
+shipping abstraction, provider branch or application-owned decision system.
+
+Native file-error fixtures reuse the acquisition adapter's exact call-site fault
+control with typed mutation/disclosure stage selection. They add fourteen
+mutation and eighteen disclosure histories, including separate ACK and Unknown
+persistence. The only recipient fixture extension deliberately withholds a native
+ACK after verified receipt; it introduces no new approval or grant authority.
+Opaque wire capture, raw file/charge comparisons, native reauthorization and
+historical ACK refinement exercise the existing owners. This is test-only parity
+under simulated syscall returns, not extraction of a shipping lifecycle mechanism
+or qualification of kernel/device/power-loss behavior.
+
+### Final event-bound audit
+
+The declared 64-event limit is enforced as operation revisions 1 through 64 by
+`store::next_operation_revision`, shared by grant publication, typed protocol
+advancement and recovery refinement. Current journal decoding and linked read
+revision validation reject values above that bound. A terminal record at revision
+64 remains readable and replayable without any new capability; advancing it
+fails before persistence. Configuration revisions are a distinct shared-control
+sequence and are not operation events.
+
+The native corrupt-journal corpus exercises exact64 terminal replay and65 refusal
+for mutation, acquisition, acknowledged disclosure and lost-ACK disclosure on
+both adapters. The shared arithmetic boundary test covers every legal increment,
+zero,64,65,JSON maximum and u64 maximum. Native validation is stricter than the
+translated generic JSON-safe transition predicate; every concrete replacement
+still calls full native record validation. No kernel theorem or durability
+assumption is weakened. No bounded-upfront consent mode is exposed: all maintained
+managed operations require native exact approval, while the retained AP070
+reference's different consent semantics remain independent.
diff --git a/docs/specs/0060-evidence-conditioned-authority.md b/docs/specs/0060-evidence-conditioned-authority.md
index 67213535a..c8610a413 100644
--- a/docs/specs/0060-evidence-conditioned-authority.md
+++ b/docs/specs/0060-evidence-conditioned-authority.md
@@ -79,7 +79,7 @@ and a new manifest, or a product-layer evaluator under 0025.
ObservationStatement = {
version: 1,
observer: PrincipalId, ; signer
- schema: ObservationSchemaId, ; e.g. "auths.gateway-outcome/2"
+ schema: ObservationSchemaId, ; e.g. "auths.gateway-outcome/3"
subject: ResourceId, ; what was observed
observed_at: Timestamp,
facts: { 1*16 FactName => FactValue },
@@ -242,7 +242,7 @@ current value equals the action's `expected` value, observed within the last
```text
requirement:
observer_anchor: gateway-observer
- schema: auths.gateway-outcome/2
+ schema: auths.gateway-outcome/3
subject: action_fact("depends_on") ; step N-1's commitment URI
max_age_seconds: 3600
conditions: [ member("stage", ["observed-by-provider"]) ]
diff --git a/docs/specs/0063-generalized-gateway.md b/docs/specs/0063-generalized-gateway.md
index da6442da5..6f9e94c1f 100644
--- a/docs/specs/0063-generalized-gateway.md
+++ b/docs/specs/0063-generalized-gateway.md
@@ -96,7 +96,7 @@ All five share one recipe revision (§3) and one store contract (§9).
- For every claimed submission it records the stage, the gateway evaluation
time, the refusal code when it refused, and the HTTP status and response
digest when a complete response exists. When an observer key is
- provisioned, it signs that record on request as `auths.gateway-outcome/2`.
+ provisioned, it signs a redacted public projection as `auths.gateway-outcome/3`; complete audit facts are operator-private `auths.gateway-private-outcome/1`.
A production gateway has no observer key until AP-SPEC-038 Epic 4 supplies a
custody client (§16).
- Within one namespace and one store, a bounded grant's count bounds every
@@ -1339,13 +1339,16 @@ custody unavailability can never block the kill switch.
## 8. Evidence and assurance
-### 8.1 `auths.gateway-outcome/2`
+### 8.1 Public outcome `/3` and private outcome `/1`
-The outcome keeps its subject
-(`auths-gateway:///operations/`) and observer. Its
-fact set changes meaning, so the schema becomes `/2` and `/1` is removed
-(§13, reading 1). The kernel treats schema identifiers and fact names as
-opaque, so the core CDDL, codec, registry, and corpus stay (060 §3.1).
+AP074 directly replaces the public `/2` fact set with
+`auths.gateway-outcome/3`. Public facts are hiding effect/recipe/record references,
+conservative stage, evaluation time and bounded status/refusal. Full facts below
+are signed only as operator-private `auths.gateway-private-outcome/1`; the
+application cannot request that schema or receive protected pre-entry bytes.
+The existing kernel presence predicate and offline audit checks remain unchanged.
+
+The private subject is `auths-gateway:///operations/`.
| Fact | Type | Present when |
| --- | --- | --- |
@@ -1371,38 +1374,25 @@ apart. At most 13 of an observation's 16 facts are used.
a later window no longer moves its entry into that window. A chained step's
`member("stage", ["observed-by-provider"])` (060 §5) works unchanged.
-**Consumers**, all changed in one cutover:
-
-- in the gateway: `observer.rs` (`OUTCOME_SCHEMA`, `outcome_facts`,
- `verify_outcome`, `ObserverAnchorTemplate`), `engine.rs`, `audit.rs`,
- `harness.rs`, and `observed_tests.rs`;
-- the custody test in `product/integrations/auths-custody/src/lib.rs`;
-- the Python and TypeScript gateway clients, their tests, and their public-API
- inventories;
-- the north-star example and every trusted context's observer anchors;
-- `compliance.toml`, 060, the claim ledger, and the runbook.
-
-**Fixtures.** `bindings/fixtures/gateway/outcome-v2.json` holds, for each stage
-and fact combination, the canonical outcome signed by a fixed test observer
-and its expected facts. Rust verifies every entry. The Python (native) and
-TypeScript (WASM) SDKs accept every entry and, through local verification,
-reach the gateway's verdict on a grant requiring `member("stage",
-["observed-by-provider"])`. Go consumes no gateway schema, so no Go vector is
-added.
-
-### 8.2 Observe request `/2`
-
-`auths.gateway-observe/2` keeps `read-back`, which is refused for a path with
-a `response-field` segment, and `outcome`. It adds `pre-entry
-{operation_id}`, which returns one operation's stored signed pre-entry
-observations as `{"outcome": "pre-entry", "operation_id", "observations_b64":
-[...]}`. It signs nothing new and contacts no provider. The clients add
-`observe_pre_entry` and `observePreEntry`.
-
-### 8.3 Audit bundle and report `/2`
-
-`auths.gateway-audit-bundle/2` entries add an optional `pre_entry_b64`.
-`auths.gateway-audit-report/2` adds the recipe's `recovery` object to its
+Public conformance vectors are `bindings/fixtures/gateway/outcome-v3.json`;
+full private fact/presence vectors are `private-outcome-v1.json`. Native, Python
+and TypeScript consume the public corpus. The native private corpus preserves
+all accepted and refused full-fact cases. Superseded public schemas are rejected.
+
+### 8.2 Observe request `/4`
+
+`auths.gateway-observe/4` permits local outcome and pre-entry history only.
+Pre-entry returns operation ID, count and hiding record commitment. It performs
+no provider I/O and returns no protected observation bytes. Independently
+approved recovery uses `auths.gateway-recover/1`. Operator-only `audit-export`
+collects private signed outcomes and protected pre-entry bytes from one native
+snapshot, and completes private audit bundles in new owner-private files.
+See AP074 for the exact hiding, framing, deadline and non-bypassability contract.
+
+### 8.3 Private audit bundle `/3` and report `/3`
+
+`auths.gateway-audit-bundle/3` entries add an optional `pre_entry_b64`.
+`auths.gateway-audit-report/3` adds the recipe's `recovery` object to its
header and a `provider_result` to each entry:
```json
@@ -1521,7 +1511,7 @@ entry, with its test evidence, in the change that introduces it:
| Recipe | `auths.gateway-recipe-source/2`, `auths.gateway-compiled-recipe/2`, `auths.gateway-recipe-review/2`, `auths.gateway-recovery-capability/1` |
| Store | `auths.gateway-attempt/3`, `auths.gateway-bounded-count/2`, `auths.gateway-bounded-sum/1`, `auths.gateway-connection/1`, `auths.lifecycle.postgresql/6` |
| Policy | `auths.gateway.argument-ceiling-window-count/2`, `auths.gateway.argument-ceiling-policy/2`, `auths.gateway-counter-set/1` |
-| Evidence | `auths.gateway-pre-entry/1`, `auths.gateway-outcome/2`, `auths.gateway-observe/2`, `auths.gateway-audit-bundle/2`, `auths.gateway-audit-report/2`, `auths.gateway-echo-verification/1`, `auths.gateway-codes/1` |
+| Evidence | `auths.gateway-pre-entry/1`, `auths.gateway-outcome/2`, `auths.gateway-observe/2`, `auths.gateway-audit-bundle/3`, `auths.gateway-audit-report/3`, `auths.gateway-echo-verification/1`, `auths.gateway-codes/1` |
| Operator | `auths.gateway-operator-attestation/1`, `auths.gateway-installation/3`, `auths.gateway-admin-request/1`, `auths.gateway-admin-response/1` |
## 9. Store contract
@@ -1876,7 +1866,7 @@ decision (option A, 2026-09-27) and is not listed.
| 32 | An admin state change that loses a compare-and-swap to another process | (a) reload and reapply up to 8 rounds, then `gateway.admin.generation-conflict`; (b) fail at once | (a): a kill switch should not fail on an unrelated concurrent change; the bound keeps it within the admin deadline |
| 33 | The pool `serve`'s descriptor check counts | (a) the `PostgreSQL` pool's maximum connections, and zero for the file store; (b) a fixed number | (a) |
| 34 | Where §8.1's presence rule lives | (a) a leaf of the translated kernel (`outcome_facts_present`), checked exhaustively by Kani, with no refinement theorem; (b) a Lean model and theorem as well | (a): the rule decides what a signed record may say, not whether a credential is leased or a write is sent, so §11 does not require a theorem; the leaf sits in the kernel crate, outside the translated slice, so adding it to the slice and a theorem later needs no move |
-| 35 | How the SDKs "reach the gateway's verdict" on `outcome-v2.json` | (a) the fixture adds verdict cases: each accepted vector's exact bytes attached to an action whose grant requires `member("stage", ["observed-by-provider"])` of it, under a context compiled as the SDKs compile it and pinned to the packaged configuration, with the native decision and code; (b) the SDKs re-derive the facts themselves | (a): the SDKs verify through their own native and WASM verifiers, and no SDK gains a second reading of the facts |
+| 35 | How the SDKs "reach the gateway's verdict" on `outcome-v3.json` | (a) the fixture adds verdict cases: each accepted vector's exact bytes attached to an action whose grant requires `member("stage", ["observed-by-provider"])` of it, under a context compiled as the SDKs compile it and pinned to the packaged configuration, with the native decision and code; (b) the SDKs re-derive the facts themselves | (a): the SDKs verify through their own native and WASM verifiers, and no SDK gains a second reading of the facts |
| 36 | What an audit bundle entry's `pre_entry_b64` holds | (a) an array of at most four unpadded base64url observations, in pointer order, as observe `/2`'s `pre-entry` answer returns them; (b) one concatenated blob | (a) |
| 37 | An entry's `provider_result` when no verified outcome exists | (a) `null`; (b) an object of nulls | (a): nothing signed stands behind it |
| 38 | The clock at which the audit evaluates pre-entry requirements | (a) the latest pre-entry `observed_at`, each already required to lie in `[evaluated-at, evaluated-at + 60]`; (b) `evaluated-at` | (a): the gateway evaluated after its read, so (b) could refuse an observation the gateway saw as fresh |
@@ -1936,12 +1926,12 @@ on the exact revision, or a commit whose diff holds the evidence.
| Epic | Work | Done |
| --- | --- | --- |
-| 1. Case file and fixtures | ADR 0013 and the case 0007 comparisons (§2), for all ten recipe capabilities and the sum budget; hostile recipe cases for every new construct, including basis points 0 and 10 001, a `POST` or `field`-segment denied read, a refused status of 429, five denied reads, `Stripe-Account` in `provider_headers`, an unregistered account-scope header, and an unbound partition; attempt scenarios `/3`, including `pre-entry-replaced-after-observation`, `relative-ceiling-above`, `relative-ceiling-exact-boundary`, `relative-basis-unavailable`, `relative-binding-mismatch`, `account-substituted-at-lease`, `denied-read-answered`, `account-scope-outside-grant`, `account-scope-unbound`, and `account-scope-only-on-action-requests`; §6.8's count and sum cases in `bounds-aggregate.json`; `outcome-v2.json`, `key-identity.json`, and `codes.json` | The vectors exist and fail against current code; the ADR and case file are reviewed |
+| 1. Case file and fixtures | ADR 0013 and the case 0007 comparisons (§2), for all ten recipe capabilities and the sum budget; hostile recipe cases for every new construct, including basis points 0 and 10 001, a `POST` or `field`-segment denied read, a refused status of 429, five denied reads, `Stripe-Account` in `provider_headers`, an unregistered account-scope header, and an unbound partition; attempt scenarios `/3`, including `pre-entry-replaced-after-observation`, `relative-ceiling-above`, `relative-ceiling-exact-boundary`, `relative-basis-unavailable`, `relative-binding-mismatch`, `account-substituted-at-lease`, `denied-read-answered`, `account-scope-outside-grant`, `account-scope-unbound`, and `account-scope-only-on-action-requests`; §6.8's count and sum cases in `bounds-aggregate.json`; `outcome-v3.json`, `key-identity.json`, and `codes.json` | The vectors exist and fail against current code; the ADR and case file are reviewed |
| 2. Recipe `/2` | The compiler, including §3.2's rows for the credential account and denied reads, `account_scope`, `bounds`, and `relative_ceiling`, and §5.1's registry classes; `recovery_capability`, and review `/2` in both CLIs; account-scope headers in request construction; derivation emits `/2`; the fixtures, north-star recipe, and derivation corpus are regenerated | Every fixture compiles to its documented class; every hostile case fails with its code; both packaged CLIs pass the corpus; §11.3's capability theorems and §11.5, including `account_scope_header_exact` and `credential_reads_fixed`, are proved and registered, with `cargo xtask formal` green |
| 3. Store and engine | §9 (kinds, `insert_all`, sweep, the file batch, schema 5, record `/3`); the §5.5 steps marked 3, including the account read and denied reads at onboarding and every lease (§5.7, §5.8) and the relative ceiling (§5.6, `ratio.rs`); re-observation of `attempting` | Conformance passes on both stores (PostgreSQL in its workflow); the scenario corpus drives the counting provider; every hostile suite has zero unauthorized entries, and every pre-claim refusal zero leases; each relative-ceiling, account, and denied-read scenario records `not-entered` with its code, one lease, and zero writes; §11.2, §11.3's transition theorems, and §11.6 are proved and registered, with `cargo xtask formal` green |
| 4. One spend limit | Formal first (the Lean model, theorems, and the three translated leaves, with `cargo xtask formal` green); then evaluator `/2` with policy `/2` (§6.10), the account-scope binding (§5.9), the count and sum slots of §5.5 step 6, §6.5's sweep, §6.7, and every §6.4 description | `bounds-aggregate.json` passes on both stores, including both races; no count or sum slot leaks; the audit flags the capacity-3 and capacity-1 bundle and the sum bundle and passes every valid one; `account-scope-outside-grant` and `account-scope-unbound` refuse with zero leases |
| 5. Operator plane | §7, including connection record `/2` (§7.3), and the §5.5 steps marked 5 | With the application at full capacity, every admin command answers within its deadline; a disable or revoke through process A stops new entries in process B at B's next reload, on PostgreSQL; a second host joins only with the matching secret, including after a disable and enable, and a cross-process `rotate` followed by a disable and enable still leases on every host; `did:key` and `raw-key-v1` aliasing is refused at install and per proof; an invalid attestation is refused; §11.4 is proved and registered, with `cargo xtask formal` green |
-| 6. Evidence and assurance | Step 1, repairing the scheduled Fuzz job (`.github/workflows/fuzz.yml`) so a scheduled campaign can pass, with a unit test on a captured libFuzzer log, merged in #168; scheduled runs pass from 2026-09-27. Then §8: outcome `/2` and its consumers (including `relative-basis`), observe `/2`, audit `/2` (including the relative-ceiling and sum checks), `echo-verify`, the SDK projections, the fuzz crate, property tests, Kani harnesses, and the code inventory | A scheduled Fuzz run is green with the gateway targets; Rust, Python, and TypeScript agree on `outcome-v2.json`; the north-star audit shows `http_status` for every entered refund, including a rejected one, and flags a bundle whose basis does not admit its argument |
+| 6. Evidence and assurance | Step 1, repairing the scheduled Fuzz job (`.github/workflows/fuzz.yml`) so a scheduled campaign can pass, with a unit test on a captured libFuzzer log, merged in #168; scheduled runs pass from 2026-09-27. Then §8: outcome `/2` and its consumers (including `relative-basis`), observe `/2`, audit `/2` (including the relative-ceiling and sum checks), `echo-verify`, the SDK projections, the fuzz crate, property tests, Kani harnesses, and the code inventory | A scheduled Fuzz run is green with the gateway targets; Rust, Python, and TypeScript agree on `outcome-v3.json`; the north-star audit shows `http_status` for every entered refund, including a rejected one, and flags a bundle whose basis does not admit its argument |
| 7. North-star recipe | The Stripe recipe moves to §3.1, declaring all five Stripe checks, with an approval window of at most 86 340 seconds; its grants carry policy `/2` with a sum limit partitioned by currency and a scope listing the test connected account; the counting double returns refunds with metadata, serves PaymentIntents, the account, and 403 refusals, requires `Stripe-Version`, honors `Stripe-Account`, and honors the key; `journey.py` checks `observed-by-provider`, the guard refusing a non-test key, and one hostile case per check: a refund above the ratio of `/amount_received`, a currency that does not match the PaymentIntent, the double reporting another account at the lease, the double answering a denied read with 200, a `connect_account` outside the grant, and a refund that exceeds the currency's remaining sum; the README non-claims and the claim ledger are updated; the test-mode command stays the developer's own step (board §0 step 4) | `stripe-refund-journey` is green from the packed wheel, and each hostile case records its code with zero writes; the ledger entry uses §1's claim and non-claim wording |
| 8. Consolidation | §12 A's removals and rewrites in one pull request | The profiles and routes are gone; the specs, ADRs, and plan are amended, and `AGENTS.md`'s summary of the boundary plan is checked against the rewrite; the board is updated; CI is green |
diff --git a/docs/specs/0074-bound-effects-and-recovery-state-machine.md b/docs/specs/0074-bound-effects-and-recovery-state-machine.md
new file mode 100644
index 000000000..1490d0b36
--- /dev/null
+++ b/docs/specs/0074-bound-effects-and-recovery-state-machine.md
@@ -0,0 +1,1388 @@
+# AP-SPEC-074: Bound effects and one recovery state model
+
+- **Status:** Implementation specification; no completion or qualification claim.
+- **Date:** 28 October 2026.
+- **Baseline:** `origin/main` at `e2d10a74fa0be5e6bfe2b9f698632c835cd7b11b`.
+- **Branch:** `codex/effect-recovery-foundation`.
+- **Priority:** Complete this foundation before expanding multi-agent delegation.
+- **Handoff:** [Implementation prompt](../plans/0074-implementation-handoff.md).
+- **Normative terms:** MUST and MUST NOT are acceptance requirements. Proposed type names below do not imply that those types already exist.
+
+## 1. Objective and scope
+
+Implement two architectural changes together:
+
+1. **One bound effect description.** Authorization, request compilation, approval, admission, execution, recovery and reporting refer to the same immutable effect identity and its authoritative Rust representation.
+2. **One recovery state owner.** Reservation, provider entry, disclosure, acknowledgement and uncertain outcomes are governed by one explicit transition model, with invariants preserved across concurrency, failure, retry and restart.
+
+The first complete demonstration is one human authorizing one agent to acquire a bounded resource, receive an authorized disclosure, and perform one exactly approved mutation followed by independent observation. A deterministic provider fixture must exercise actual transport and restart boundaries. Expanding participant enrollment, delegation trees, federation, identity recovery or native resource enforcement is outside this specification.
+
+### Managed boundary and retained AP070 reference
+
+AP074 establishes one authorization, admission, execution and recovery authority
+for each **managed gateway operation**. AP070 remains an independently
+implemented, application-owned Google read demonstration with its existing
+Python/SQLite decisions, credentials, routes and explicitly narrower claims.
+Its migration is deferred until managed Google Drive access is intentionally
+productized; it is not an AP074 completion requirement. Preserve the broker and
+its independent reference value. Sharing offline native verification does not
+make its application admission or receipts managed gateway facts.
+
+One managed operation MUST NOT have two independently authoritative decision
+systems. Consequently the gateway MUST NOT delegate a managed grant to AP070,
+load its provider adapter as a runtime callback, give it gateway-managed
+credentials/permits, or accept its SQLite decisions/receipts as native entry or
+recovery evidence. Inventories and deployment policy must distinguish credentials,
+channels, operation namespaces and receipt claims. Independent application-held
+credentials or external writers remain outside managed containment; their
+existence must not be hidden by a managed guarantee. If an application-owned
+reference is reachable by a managed agent, the deployment must still prove that
+it cannot exercise managed credentials or authorize the same managed effect.
+
+Retain independent reference/differential tests and document semantic differences;
+do not force an application-owned reference to use the implementation it tests.
+A future managed Google integration must use native approved capabilities and one
+native journal, with provider-specific metadata/custody evidence reviewed then.
+No Google metadata capability, OAuth refresh integration or AP070 Python-owner
+removal is required to establish the generic gateway foundation. Native bounded
+acquisition/disclosure and all A01–A22 evidence remain mandatory.
+
+This is a prerequisite foundation for those later features. Do not import the AP073 branch wholesale, copy its unfinished task protocol, or make this work depend on its merge. Its findings may inform adversarial cases after independent verification against this baseline.
+
+The work must replace competing interpretations, not add another wrapper around them. Completion requires maintained consumers to use the new owner and superseded managed execution paths to be removed. It does not require replacing sound cryptography, the native verifier, established recipes or the whole repository.
+
+## 2. Existing foundations and the actual gap
+
+This inventory is based on source at the baseline, not a claim of exhaustive review.
+
+| Existing surface | What is already present | Required disposition |
+| --- | --- | --- |
+| [`VerifiedAction`](../../core/crates/auths-verifier/src/lib.rs) | Sealed native action, selected branches and evidence satisfactions. | Retain native verification as the authority source. A signature, parsed proof or application principal must not replace it. |
+| [`VerifiedCommand` and `verify_detailed`](../../product/runtime/auths-gateway/src/engine.rs) | Native verification, branch projection and closed request construction. | Consolidate downstream effect binding here; eliminate independently reconstructed security facts. |
+| [`CompiledRecipe` and `ClosedProviderRequest`](../../product/runtime/auths-gateway/src/recipe.rs) | Bounded data-only request compiler, account/credential requirements, observation plan. | Retain reviewed compilation. Bind its exact result, recipe identity and installation to the effect. |
+| [`submit.rs`](../../product/runtime/auths-gateway/src/submit.rs) and [`order.rs`](../../product/runtime/auths-gateway-kernel/src/order.rs) | A pure step machine already directs write admission and I/O ordering. | Extend or refactor this maintained owner. Do not create an unrelated executor alongside it. |
+| [`transition.rs`](../../product/runtime/auths-gateway-kernel/src/transition.rs), [`recovery.rs`](../../product/runtime/auths-gateway-kernel/src/recovery.rs), [`outcome.rs`](../../product/runtime/auths-gateway-kernel/src/outcome.rs) | Attempt transition rules, recovery capability and truthful outcome predicates. | Preserve their invariants and formal links while making reservation, entry and disclosure explicit. |
+| [`GatewayAttempts`](../../product/runtime/auths-gateway/src/store.rs) | Attempt records, atomic bounded claims, counters, file and PostgreSQL adapters. | Use one logical effect journal and one transaction contract. Do not add an independently authoritative task ledger. |
+| [`auths-lifecycle`](../../product/runtime/auths-lifecycle/src/lib.rs), especially [`operation.rs`](../../product/runtime/auths-lifecycle/src/operation.rs) and [`sealed.rs`](../../product/runtime/auths-lifecycle/src/sealed.rs) | Versioned identities, preparation commitments, lifecycle relations and store acknowledgements used by reference consumers. | Compare and reuse exact mechanisms. Document why retained reference models are not a second production owner. Do not invent a third lifecycle framework. |
+| AP070 [`admission.py`](../../demos/google-employee-delegation/broker/src/employee_read_demo/admission.py), [`read_executor.py`](../../demos/google-employee-delegation/broker/src/employee_read_demo/read_executor.py) and [`receipts.py`](../../demos/google-employee-delegation/broker/src/employee_read_demo/receipts.py) | Demo-owned SQLite read accounting, release decisions and receipt orchestration. | Retain as an independently application-owned reference with narrower claims. Compare shared contracts and document differences; do not use its decisions for managed gateway operations. Migration is deferred. |
+| [`DeploymentClock`](../../product/runtime/auths-gateway/src/qualification.rs) | An existing deployment time/trust boundary. | Strengthen this contract where necessary. Do not introduce clocks in individual handlers or reinterpret signed event time. |
+
+Main contains AP070, not the AP071/AP073 native task read/write implementation. This specification must be implementable without files that exist only on those branches.
+
+Baseline managed and independent reference paths:
+
+```mermaid
+flowchart TB
+ N[Native verified action] --> W[Write command and recipe compilation]
+ W --> O[Gateway submission order]
+ O --> J[Gateway attempts and counters]
+ N --> D[Demo read projection]
+ D --> S[Python read reservation and release state]
+ J --> WR[Write outcome projection]
+ S --> RR[Read receipt projection]
+ L[Lifecycle reference contracts] -. separate ownership mapping .-> J
+ S -. retained independent reference; no managed authority .-> J
+```
+
+Target:
+
+```mermaid
+flowchart TB
+ N[Native verified action and installed policy] --> C[Deterministic effect compiler]
+ C --> B[Immutable bound effect]
+ B --> H[Human review and approval of effect identity]
+ H --> A[Fresh admission checks]
+ B --> A
+ A --> R[One pure operation transition model]
+ R <--> J[One transactional effect journal and capacity owner]
+ J --> P[Durably minted one-use entry capability]
+ P --> X[Closed provider or recipient transport]
+ X --> E[Typed response, acknowledgement or uncertainty evidence]
+ E --> R
+ J --> V[Native receipt and activity projections]
+ V --> UI[Thin managed bindings and UI]
+ REF[Independent application-owned AP070 reference] -. differential comparison only .-> R
+```
+
+“One” means one semantic owner and one atomic admission authority per partition. It does not mean one global process, one physical database table, one mutable object passed everywhere or a single linear lifecycle for all effect kinds.
+
+## 3. Architectural decisions
+
+### 3.1 Ownership
+
+- The offline core continues to own canonical actions, proof validity, attenuation, assurance and evidence verification. It acquires no I/O, clock reads, live budgets or recovery store.
+- The existing `auths-gateway-kernel` is the default owner of pure managed-operation transition and outcome predicates. Preserve its bounded explicit-input and translation constraints. Consolidate its existing order/transition/recovery relationships; do not keep two predicates that can independently authorize the same transition.
+- `auths-gateway` owns effect compilation, sealed stage types, live admission, custody, transport and journal orchestration. Store adapters implement atomic persistence, not independent lifecycle rules.
+- `auths-stores` remains the durable adapter boundary. Existing `auths-lifecycle` identities and store acknowledgement mechanisms must be assessed for reuse before adding analogous types.
+- Bindings expose Rust construction, verification, transition and projection APIs. They do not independently hash an effect, decide a recovery transition or infer success.
+
+A different package placement requires a written ownership comparison and ADR before changing dependencies. Moving files is not an acceptance criterion. No new crate is required by this spec.
+
+### 3.2 Shared model, concrete protocols
+
+Use a closed protocol sum with typed data for **provider acquisition**, **recipient disclosure**, and **provider mutation**. These share effect identity, revision checks, accounting, durable entry and immutable evidence rules. Their specific observations and terminal claims remain distinct.
+
+A provider name, arbitrary operation string, callback, runtime plugin or untyped map MUST NOT select hidden authorization or recovery behavior. Provider paths, mappings and supported observations remain reviewed, digest-bound recipe data. A new gateway capability must satisfy the repository abstraction process.
+
+The implementation must write an abstraction case comparing at least the existing mutation path, the independently application-owned AP070 acquisition/release reference and lifecycle reference model, including divergent semantics and distinct claim boundaries. Retaining that independence is required; migration is not the comparison gate. Compare two maintained mutation recipes with different recovery capabilities. If the normal extraction threshold is not met, an ADR must justify the narrowly identical mechanism. This is an implementation deliverable, not permission to skip evidence gates.
+
+### 3.3 Direct cutover
+
+This is prelaunch. Replace affected formats and call sites directly, reject obsolete disposable state, and remove superseded production logic. Do not add migration readers, dual journals, compatibility flags or a runtime switch between old and new semantics. Retain independent test oracles and useful historical evidence with their original scope.
+
+### 3.4 Managed deployment and non-bypassability
+
+Every managed effect MUST pass through bound-effect admission and durable entry.
+Inventory application sockets, SDKs, CLI/admin interfaces, credential-bearing
+ancillary reads, transport adapters and retained reference paths. No interface
+available to the agent may bypass these gates or substitute caller-controlled
+transport bytes. An operator commissioning path must satisfy the same effect
+binding, approval and entry protocol in addition to its commissioning authority.
+
+The application protocol `auths.gateway-observe/4` accepts only local outcome
+and pre-entry history queries. Obsolete observation schemas and the unapproved
+`read-back` request are rejected. `auths.gateway-recover/1` supplies a source
+operation and the bounded `auths.gateway-submit/2` proof/action/approval envelope
+for an independently authorized recovery read. It invokes the same native
+engine and returns only a redacted recovery report or native refusal. Neither
+route authorizes disclosure of acquired provider content. The former raw read
+helper is a test-only reference on the private synthetic harness control plane.
+
+State the deployment threat model: an untrusted agent process is outside the
+credential, private review, journal and operator boundaries. The trusted host,
+gateway operator, custody service and configured transport enforce those
+boundaries. Enumerate credential access, filesystem/process permissions,
+authenticated channels and network restrictions needed for the claim. Possession
+of unrelated provider credentials, a malicious host/operator, and independent
+external writers are not contained by a Rust capability type. Application-owned
+reference paths must not be reachable with managed credentials or presented as
+managed execution.
+
+Installed acceptance evidence must run a hostile agent in a separate restricted
+process and exercise every inventoried application-facing route, including
+malformed/obsolete frames and direct credential/provider access attempts. Record
+the actual isolation and egress controls exercised; a negative test alone does
+not prove non-bypassability under unspecified deployment conditions.
+
+## 4. The bound effect contract
+
+### 4.1 Immutable description and identity
+
+Introduce one versioned canonical `EffectBinding` with an immutable `EffectId`. The concrete name may reuse an existing type after the ownership comparison. Each semantic field has exactly one authority and one encoding.
+
+| Dimension | Bound facts and trusted source |
+| --- | --- |
+| Operation | Operator/admission namespace, partition epoch and logical operation ID. These determine the replay/conflict lookup key. |
+| Authority | Native action commitment, selected actor/branch identities, authorization context/configuration identity, authority scope and the relevant grant/evidence dependencies. |
+| Scope | Engagement/task scope when present, exact resource identity, acting recipient/executor audience and holder/channel restrictions. No task-free value may silently match a task-bound one. |
+| Compilation | Profile identity, recipe digest, compiler semantic version and exact noncredential transport description. |
+| Installation | Connection identity, account/subject binding, credential generation, installation/configuration version and required qualification policy. |
+| Information flow | For acquisition: source/projection and response limits. For disclosure: exact recipient, channel binding, source-effect relationship, permitted projection and byte ceiling. |
+| Accounting | Typed reservation set: partition, counter key, unit, amount, window and any conservation relation. Read calls, bytes and write count are not interchangeable units. |
+| Evidence and recovery | Approval requirements, freshness dependencies, observation/reconciliation contract and its identity. A provider idempotency declaration includes scope and retention assumptions. |
+
+Use tagged, versioned alternatives for protocol-specific facts. Do not represent validity with a large set of optional fields whose illegal combinations are interpreted differently by callers.
+
+The concrete native encoding is `auths.bound-effect/3`, with effect compiler
+`auths.gateway-effect-compiler/8`. Reservations use a closed window alternative:
+fixed epoch-aligned windows or the retained source effect's lifetime. Mutation
+count/sum reservations retain fixed windows; recovery acquisitions declare
+separate acquisition-count and response-byte charges against one source. A
+source-lifetime bound must survive operation-ID changes, replay and restart.
+An encoded bound declares an obligation; only the journal's atomic admission
+and entry transitions can discharge it. Decoding never grants authority.
+
+The replay key MUST NOT include the action/effect digest as its only differentiator: the same logical operation ID with changed binding is a conflict, not a fresh operation. Exact replays return existing state; they never mint another capability. A newly chosen operation ID cannot evade a signed approval's binding or the shared budget.
+
+Managed recovery policies use the same selected-branch tightening and fixed-window
+count/sum arithmetic as mutation. Their argument is `maximum_response_bytes`;
+read calls and full response ceilings have separate counter identities from
+mutation counts and amounts. The mutation recipe's spend-sum requirement does
+not apply to independently authorized read actions. Admission reserves these
+slots atomically with the permanent source allowance, and final entry checks
+all membership and window facts under the writer fence. Unknown outcomes retain
+both charges. Prerequisites do not charge the parent policy a second time.
+The existing sixteen-link and 34-reservation limits remain unchanged.
+
+### 4.2 Compilation and approval without circular commitments
+
+1. Verify the caller's canonical action under installed trust and policy.
+2. Compile only that verified action and pinned installation/recipe into a candidate effect. All security-relevant provider bytes and recipient policy are now fixed.
+3. Construct the canonical binding and `EffectId`, including the compiler output commitment.
+4. Render review data from that same bound value. Obtain required approval over its exact identity and the approved scope/deadline.
+5. Validate approval separately and construct `ApprovedEffect`. Do not include the approval signature in the value it signs.
+6. Admit and execute only this approved value after fresh checks. Recompilation under changed recipe, account, configuration, recipient or payload yields a different binding and requires new authorization/approval as applicable.
+
+The recipe compiler cannot literally consume its own completed output. “Same description” means one authoritative semantic input and a sealed deterministic compilation result, both bound by one identity. It does not require cyclic hashes or trusting a caller-supplied compiled request.
+
+Broad delegated authority and exact human approval remain different facts. Approval can narrow an effect; it cannot create missing resource authority. A bounded-upfront approval is a separate typed mode with explicit scope and expiry.
+For acquisition and disclosure it can approve the source/projection, recipient and
+limits before the content is fetched; the exact disclosure binding is finalized
+after acquisition and checked against that consent. It cannot be used to approve
+a mutation that requires exact human confirmation. Its mere presence must not satisfy an exact-approval requirement.
+
+Each approvable operation MUST declare a versioned review contract, bound through
+the approved recipe/compiler semantics and effect identity. It names all
+security-critical fields, their types/units and required presentation, including
+the destination, action consequences, relevant limits and uncertainty. Domain
+profiles define domain-specific fields: a financial profile may require amount,
+currency, merchant identity, recurrence and settlement consequences. The shared
+gateway must not infer such semantics from provider names or arbitrary prose.
+
+Native code projects those fields from the same sealed effect. Unsupported
+review versions, absent fields, truncation that hides meaning, or an incapable
+renderer MUST prevent approval. Untrusted labels/content cannot replace trusted
+field names or obscure destinations, units or consequences. The approval channel
+must authenticate the approver and associate the displayed projection with the
+exact signed effect; an agent's assertion that review occurred is insufficient.
+Test rendering failures, substituted labels and changed review semantics. This
+establishes faithful, complete presentation under the declared contract, not a
+proof that the human understood it.
+
+The concrete recipe declaration is `auths.effect-review-contract/1` in recipe
+source `/3`: required consequences and one typed presentation for every
+non-binding profile argument. Integer presentations require explicit units.
+The operator supplies domain meaning; OpenAPI derivation produces an incomplete
+draft and cannot authorize these facts. Native review always includes the three
+binding arguments, destination, exact request, installation, reservation and
+recovery facts independently of recipe labels. Private review `/2` supports the
+complete literal renderer `auths.literal-complete-review/1`; control and bidi
+characters are escaped data, and output is never truncated. An installed frontend
+must preserve that contract and associate its displayed result with the exact
+signing request. Native fixture parity alone does not close the installed
+approval-channel or human-journey acceptance gate.
+
+The managed Airtable demo now has an optional simulation-only owner interface
+on a separate loopback origin. In this mode the application receives no exact
+effect approvals at startup. A private operator login file authenticates the
+owner to the installed approver boundary. One opaque pending-review identity
+links the complete native literal projection and signing request to the owner's
+approve/decline decision; unsupported, truncated or mismatched review envelopes
+cannot enter that channel. Only the operator then signs the retained native
+request. The application cannot supply review text, signing bytes or a consent
+decision. Native proof, binding, approval, admission and recovery checks remain
+the sole execution authority.
+
+The owner page uses explicit same-origin bearer headers held only in page
+memory. It uses no cookies or browser storage: cookies would cross localhost
+ports and reach the application's origin. Host/origin, session and CSRF checks,
+one pending review, eight handlers, 1,024-byte commands, the native 131,072-byte
+review envelope, five-second HTTP I/O, fifteen-minute session and ten-minute
+consent bounds are explicit. These UI limits never refresh native signed times
+or evidence. Existing source-only review RPCs retain their 1,024-byte bound;
+new proof/action proposals use a separate bounded operator socket.
+
+Hosted tests are authored for authentication/refusal, complete unmodified
+presentation, withheld approvals, declined effects with no native entry, a
+substituted review, one exact write, separately approved recovery and replay
+without new execution. The source-free installed browser journey simulates owner
+actions and retains screenshots/provenance. It does not establish human
+attention, live-provider qualification, the complete content acquisition/
+disclosure UI or hostile-host containment. The default simulator and live
+operator channel keep their existing narrower claims. A22 remains open until
+the complete installed journey and required exact-head evidence pass.
+
+### 4.3 Closed transport and custody
+
+The private bound value contains the exact method, origin/path/query, noncredential headers, media type, body and closed observation plan. Transport accepts a stage-sealed effect capability, not a second set of URL/body/recipient parameters. Credential injection may fill only the approved credential slot for the bound connection/account/generation. Credentials themselves never enter effect serialization, public commitments, receipts, bindings or agent memory.
+
+Redirects, automatic retries, endpoint substitution and middleware body/header mutation cannot silently change the bound send. Deny unsupported behavior. Idempotency and correlation values are generated by the native compiler and included in the binding. Byte equality means the application request after deterministic compilation, excluding explicitly specified transport framing and secret credential insertion; it is not equality of TLS packets.
+
+Credential-bearing preflight, account checks, provider observations and recovery
+reads must also be bound, bounded ancillary acquisition operations derived from
+the approved recipe. Account for their entry and limits explicitly; they must not
+be invisible I/O outside the model. Acquiring those bytes does not authorize
+disclosing them to the agent.
+
+Mutation compiler `/8` retains an ordered prerequisite acquisition plan beside
+its verified command and commits the complete plan in the transport identity.
+The plan names at most seven requests: one account check, up to four denied
+capability checks, one pre-entry read and one relative-ceiling read. Each pins
+its compiler-owned GET/HEAD, headers, response-byte ceiling (at most 65,536),
+evaluation purpose and resolved values, one call, 15 transport seconds, at most
+30 seconds of evidence age, and no disclosure. Tighter native evidence freshness
+requirements still apply. Onboarding probes and later observations are separate
+operations and cannot borrow this allowance. Private approval review renders
+these declarations. The declaration alone is neither a capacity reservation nor
+an entry permit; runtime must durably account and grant every constituent read
+through the shared journal before contacting the provider.
+
+Candidate onboarding/rotation compilation uses the reserved native tool
+`auths_gateway_candidate_check_v1` and private contract
+`auths.gateway-candidate-check/1`. It pins the complete proposed connection and
+exact predecessor or first-install absence. Rotation preserves descriptor,
+account, authorization and administrative state; changing those is not a
+credential-only rotation. Native read authority and exact approval are required
+separately from the operator installation attestation. Checks run in recipe order:
+probe, account, then denied capabilities. The final declared read is the main
+acquisition and earlier reads are its prerequisites; no synthetic aggregate
+provider response is invented. A recipe without checks has no candidate network
+operation. Each credential identity (connection ID, credential generation and
+reference commitment) has a 16-attempt lifetime ceiling independent of operation
+IDs and timestamps, with the full primary response ceiling and explicit child
+allowances. Selected native managed limits also apply. This compilation alone
+cannot establish predecessor freshness, enter transport or publish a credential;
+the runtime must fence predecessor/absence and retain the checks through safe
+publication in the same authoritative journal contract.
+
+The private operator proposal format `auths.gateway-credential-candidate/1`
+retains canonical connection/predecessor bytes and the exact recipe digest,
+bounded to 65,536 bytes. Decode rejects duplicate, unknown, noncanonical,
+obsolete-schema, wrong-recipe and invalid-relationship data and reconstructs
+checks only from the installed recipe. The proposal contains no provider secret
+and grants no native authority, custody lease, transport or publication. It is
+private operator data suitable for preserving an exact target across approval
+and restart; the journal and fresh fences remain authoritative.
+
+First installation retains a private canonical connection draft before custody,
+with fixed connection ID, account, descriptor and timestamps. For a read-bearing
+recipe, `install` stages the secret and private proposal but leaves the shared
+connection absent. Cleanup notes precede custody. The explicit operator staging
+action binds the host generation floor to the exact unpublished generation-one
+candidate; this does not authorize ordinary entry. Candidate compilation,
+acquisition and publication check that floor without initializing it. Runtime
+startup cannot repair a missing floor. Explicit installation staging can resume
+from the retained draft and same secret after an ambiguous custody result; it
+must not replace a conflicting staged secret or a substituted floor. Joining an
+existing connection records cleanup obligations before custody confirmation.
+
+Rotation staging stores a successor only after validating the live predecessor,
+recipe, secret prefix, custody and generation and recording its durable cleanup
+obligation. It performs no provider read or lease, never deletes an existing
+successor, and returns the private proposal for native authorization. The CLI
+exposes staging, native call authoring, candidate review, acquisition and operator
+publication as separate steps. Read-bearing recipes cannot use commitment-only
+rotation; recipes with no reads retain their offline operator path. A join of
+an already-published connection confirms its exact credential binding without
+re-running provider checks. Local file custody reloads state under a stable
+cross-process lock for each operation; contention is bounded by immediate
+refusal, and stale handles cannot overwrite a successor or serve a generation
+already removed by a completed custody mutation.
+
+Candidate journal admission uses the expected connection row as a source
+rather than an invented source mutation. Both admission and every child/primary
+entry must recheck predecessor/absence, exact native authority, configuration,
+approval and qualification under the same fence. Root source and native checks
+share one snapshot when they address the same row. Entry audit
+`auths.effect-entry/3` records checked absence explicitly, separately from a
+present record's digest. A candidate's source-lifetime byte ceiling cannot
+change once its credential identity has an allocation. Candidate acquisition
+returns retained response/unknown facts; publication remains a separate operator
+action requiring the final purpose and complete fresh retained evidence.
+
+The trusted operator publication API fences the exact candidate root, its
+permanent allocation, every child response/allocation, native configuration
+and expected predecessor (or absence) with the connection replacement/insertion.
+It re-evaluates every actual check purpose, including the primary response,
+and measures evidence age from the retained grant time. Missing, unknown,
+failed or stale evidence cannot publish. Publication creates no transport permit,
+changes no acquisition fact and resets no charge. `Published` means the store
+acknowledged this transition. `AlreadyCurrent` means only that exact candidate
+bytes were observed as current; it does not attribute causation or refresh
+checks. This distinction permits explicit observation after acknowledgement
+loss. The operator channel supplies publication authority separately from native
+read proof/approval. Custody is checked for the exact staged reference before
+publication; independently stored custody and connection state are not claimed
+to be one atomic transaction.
+
+The concrete journal encoding is `auths.gateway-attempt/16`. Each prerequisite
+has a component replay key derived from the parent's logical operation key and
+its compiler ordinal, a permanent one-call/maximum-response-byte allocation,
+and a typed acquisition record retaining the parent's exact approval. A bounded
+canonical parent transport witness proves membership; it cannot construct a
+transport request. Granting a read requires a still-reserved parent under the
+same fence as its allocation and current native dependencies. Final mutation
+entry fences all required prerequisite records and allocations and requires
+recorded, fresh responses. Receipt time is immutable, and evidence age uses the
+earlier grant time conservatively. These requirements do not waive tighter
+native evidence freshness or independently authorized later observation.
+
+Independent recovery uses the same account/denied-capability subset, with at
+most five child reads. Its complete request commitment authenticates these
+declarations, and the native review renders their resolved requests and limits.
+The installation-only probe remains separate. A prerequisite source may be a
+reserved mutation or reserved independent recovery acquisition, never another
+prerequisite. Recovery final entry fences all child records and allocations and
+evaluates their account/capability predicates and freshness. The same consumed
+permit and retained-response runtime helper serves both parent types. No child
+recharges its parent's managed policy. The maximum recovery entry set is 48
+records within the existing 64-record bound.
+
+Final mutation entry must also evaluate each prerequisite's actual purpose against
+those retained responses. Signed pre-entry observations must have valid canonical
+signatures from the pinned observer and exactly match the retained subject,
+projected facts and original conservative time. Reevaluate the selected native
+requirements and anchor validity at the fenced entry time. Relative-ceiling
+checkpoints must preserve the response's basis, digest and time, and satisfy the
+original bound values and ratio. Extraneous checkpoint evidence cannot substitute
+for a missing prerequisite. The selected gateway observer identity is part of the
+installed configuration commitment. All JSON prerequisite interpretations reject
+duplicate members through the shared bounded response decoder.
+
+Independent recovery compilation uses the reserved native MCP tool
+`auths_gateway_recovery_read_v1`; a mutation recipe cannot claim that name.
+Its `auths.gateway-recovery-read/1` contract binds the retained source effect
+and operation, exact GET and response limit, observation projection/expected
+value and echo, count/time/freshness limits, and the absence of content
+release. The read requires its own current native proof and exact approval;
+the source writer's proof and historical approval are provenance only. The
+same native binding, review renderer and approval machinery serve both types.
+This compilation contract does not itself implement read admission, an entry
+permit, observation evidence or disclosure.
+
+A standalone read-only recipe uses source `/3` with no `write` declaration,
+the reserved `auths_gateway_content_read_v1` tool and a required content capability.
+It cannot declare mutation observation, echo, pre-entry, precondition,
+relative-ceiling or spend-sum configuration. Its profile lock has only the
+existing binding fields, exact resource field and optional account scope.
+The native compiler lowers this to `ContentOnly`, which contains no mutation
+plan. Mutation request construction fails before custody or transport. Candidate
+credential reads, content acquisition and disclosure continue through the existing
+native owners. Operator review has null mutation method/recovery, empty write path
+and zero write-body allowance; its content GET declaration is rendered separately.
+Rust review method/recovery APIs return explicit absence for this case. Existing
+mutation source/review bytes remain unchanged. No GET write method, fake write
+endpoint or separate read executor is introduced.
+
+The native read-only corpus freezes source/lock/review and rejects mutation
+adjuncts, missing content, custom tools and write injection under the reserved
+read tool. A real CLI/daemon fixture installs it under read/release-only native
+trust, refuses mutation submission/review and mutation recovery from acquired
+content, and performs exactly one content GET plus authenticated disclosure.
+This supports standalone native reads. Authenticated UI and installed language
+evidence remain required. Google metadata, credential refresh and AP070 migration
+are deferred provider-integration work outside AP074 completion.
+
+Explicit content acquisition is a versioned opt-in recipe capability,
+`content.schema = auths.gateway-content-read/1`, within recipe source `/3`.
+It is included in the recipe digest and operator review; absence permits no
+content acquisition. The declaration selects one string/enum resource field,
+a fixed/verified-resource path (existing 16-segment bound), up to eight sorted
+unique literal query pairs (64-byte ASCII names, 512-byte values), and either
+exact UTF-8 content or a JSON-string pointer (128 bytes/eight tokens with valid
+RFC 6901 escapes). The source origin, credential and registered version/account
+headers remain the existing recipe owners. Dynamic URLs, response-derived
+locators, operation-ID path parameters and arbitrary callbacks are refused.
+
+`auths_gateway_content_read_v1` is a reserved native read tool. Its private
+`auths.gateway-content-acquisition/1` contract binds the exact installed
+connection, recipe, resource/account arguments, GET, projection, response and
+content limits. Both byte limits are at most 65,536; the content ceiling cannot
+exceed the response ceiling. A source allowance has sixteen attempts and the
+full response ceiling per attempt, keyed by recipe, connection ID, credential
+generation/reference and resource/account arguments. Changing logical operation
+IDs or administrative timestamps/state cannot reset that allowance. Selected
+native count/byte policy bounds also apply. Account/denied-capability
+prerequisites use the existing compiled plan. Native exact approval and full
+recompilation are separate from the resource proof. Credential-check and recovery
+admission cannot be borrowed by a content effect.
+
+Projection requires status 200 and the declared `text/plain` or
+`application/json` media type, optionally with UTF-8 charset. It refuses lossy
+UTF-8, duplicate-member JSON, missing/non-string selections and over-limit
+responses/content instead of truncating. It returns private zeroizing bytes;
+projection alone establishes no transport provenance, custody or release
+permission. Stable native response errors distinguish status, media type, raw
+response limit, malformed input, missing projection and projected-content limit.
+Content acquisition now uses the shared journal admission, permanent source
+allocations, prerequisite sequence and final native entry fence. The sealed
+content permit selects a separate metadata-preserving HTTPS response path. Only
+that compiler-owned GET may carry the reviewed fixed query; mutation, credential
+and recovery transports continue to reject query overrides. The client requests
+identity encoding, disables automatic decompression and retains bounded raw
+bytes with status, required-version agreement, receipt time and typed media-type/
+encoding facts. Each header is missing, one value of at most 128 bytes, or invalid
+(including duplicate/oversized/non-text values). These facts cannot be stripped
+by decoding a content record as ordinary prerequisite/recovery evidence.
+
+The typed `content-response-recorded` phase projects to the existing acquisition
+response stage under the sole reducer. Attempt schema `/15` atomically retains
+one additional closed result: protected content or a typed projection refusal.
+Obsolete `/14` records are rejected; no compatibility decoder or migration exists.
+Required-version agreement, identity encoding, one valid media type and the exact
+compiled projection must all hold before protected content is constructed.
+Encoding may be absent or a single case-insensitive `identity` value; compressed
+or ambiguous encodings refuse. Empty content is valid and distinct from absence.
+The raw response and refusal remain durable even when projection fails; charges
+are not refunded. Two additional stable codes distinguish response-version and
+response-encoding failure.
+
+The private `auths.gateway-protected-content/1` record contains a freshly sampled
+32-byte key, a 32-byte HMAC-SHA256 commitment and the exact projected byte count.
+The MAC message is the existing native domain commitment under that schema over
+the canonical tuple `(source EffectId, acquisition projection commitment, byte
+count, private content digest)`. The private content digest uses
+`auths.gateway-content-bytes/1` over the exact projected bytes. The content digest
+and key are never public identifiers; the key stays private with the journal.
+This reuses the workspace HMAC primitive and native canonical/domain framing.
+Every load reprojects the retained response and verifies the byte count and MAC
+with constant-time verification. A refusal is also recomputed rather than trusted
+as an arbitrary stored code. The projected bytes need not be duplicated in the
+journal: immutable raw evidence and the bound projection reconstruct them exactly
+without provider I/O. Key fields and decoded/projected buffers are zeroized on
+drop where supported. Existing protected-journal access policy is required; this
+mechanism is not encryption or defense against an attacker controlling that
+journal and its keys.
+
+Restart and post-grant replay return recorded facts without a credential lease
+or new provider permit. Unknown/oversized responses retain their full charges.
+The public acquisition projection reports only `ready { bytes }` or
+`rejected { code }`, never the private key, content digest, hiding commitment or
+plaintext. A ready result grants no disclosure authority. Standalone read-only
+installation and installed managed CLI/binding journeys remain required. AP070
+metadata and broker integration are separately deferred.
+
+The journal's sealed `RetainedContent` is private provenance, with no public
+plaintext accessor or deserializer. Loading requires the exact namespace,
+operation and partition and revalidates the content response and MAC. It refuses
+unknown, absent, projection-rejected, foreign-protocol or corrupt source state.
+This historical load does not reuse the source proof or approval as release
+authority, and never refetches content.
+
+The reserved native `auths_gateway_disclosure_v1` action compiles a separate
+`auths.gateway-disclosure/1` contract over that provenance. It binds the source,
+protected commitment, exact projection/length, raw-key Ed25519 recipient,
+channel and delivery instance. The byte ceiling is the exact projected length;
+there is no caller plaintext or truncation override. Recipient/channel values
+are proposed data until independently authenticated at entry. The full original
+connection, recipe, partition and epoch must still match; credential replacement
+is not implicitly accepted. Compilation verifies fresh native release authority,
+selected dependencies and observer separation; native review, exact approval and
+full entry-time recompilation are the existing shared owners. Disclosure has no
+provider credential prerequisites and cannot enter acquisition/mutation APIs.
+
+Disclosure adds the typed `disclosure-count` unit alongside `disclosure-bytes`.
+The fixed source allowance is sixteen deliveries, shared across operation IDs,
+recipients, channels and instances for the retained acquisition. Each declaration
+charges one count and the exact content length; the byte ceiling is sixteen times
+that length (one is used only for the nonzero ceiling representation of empty
+content). Empty delivery charges zero bytes and one count. Selected native policy
+bounds use `maximum_disclosure_bytes`, disjoint disclosure count/byte counter
+domains and the existing bounded tightening/window arithmetic. The thirty-four
+reservation bound is unchanged. These obligations are enforced by journal admission below; no admission
+reservation itself authorizes a delivery. The separate final grant must recheck
+the reservations before issuing any one-use delivery capability.
+
+Disclosure admission now uses a third closed journal protocol under attempt
+schema `/15`. Its immutable plan retains source-binding/projection provenance and
+recipient/channel/instance data; it cannot reconstruct an executable capability.
+The existing writer fence validates the source response/MAC, full current native
+authority and configuration, exact approval and fixed-window predecessors before
+atomically inserting the disclosure record, permanent source slot and selected
+policy slots. Lifetime and policy exhaustion create a retained not-entered record
+without a new allocation. Concurrent same-operation claims return one reservation;
+changed binding conflicts. Admission itself authenticates no recipient and emits
+no plaintext or delivery permit.
+
+Public disclosure snapshots distinguish reserved, not-entered, unknown and
+acknowledged, with explicit retained count/byte allocations and policy-counter
+counts. No content, recipient/channel correlation or protected commitment is
+public. The shared operation listing and support projection retain the disclosure
+protocol distinction. Existing acquisition and mutation decoders refuse disclosure
+state rather than interpreting its facts as provider evidence.
+
+The journal ACK recorder authenticates against its retained binding and exact
+content length, requires the sole reducer's original grant, and persists immutable
+canonical ACK bytes. It accepts historical evidence after authority changes;
+current permission is not required to record past facts. Exact replay returns the
+same revision. A concurrent CAS loser reloads at most once to recognize identical
+stored evidence, never to redeliver. The native engine exposes separate proposal,
+compile and admission operations, all without credential lease or byte delivery.
+The native Unix delivery API now authenticates an owned stream before the final
+fenced grant. `auths.gateway-unix-disclosure-channel/1` binds gateway/recipient
+UIDs and a private channel identifier into the approved channel value. Both peers
+check their effective UID and the kernel peer UID. Under
+`auths.gateway-disclosure-auth/1`, the recipient signs a fresh 32-byte challenge
+bound to the complete native delivery statement. This proves key possession for
+that handshake; it is distinct from an ACK and never asserts receipt.
+
+The final writer fence rechecks the source/MAC, exact current native authority,
+approval, configuration, allocation and managed counter memberships, plus the
+live channel and its deadline. It persists the immutable authentication transcript
+with the unique entry audit before returning a noncloneable delivery permit.
+Only consuming that permit can send the exact reconstructed bytes. Attempt `/15`
+requires authentication evidence whenever a disclosure has an entry grant;
+obsolete `/14` is rejected. Decoding historical evidence cannot reconstruct a
+socket or permit. ACK recording verifies the retained transcript and exact
+original delivery even after current authority is lost.
+
+The local protocol uses unsigned big-endian 32-bit frame lengths, bounded before
+allocation: authentication and ACK frames at most 4,096 bytes, content at most
+65,536 bytes and exactly the approved projected length, including zero. Native
+canonical decoding rejects duplicate/unknown/noncanonical authentication fields.
+Authentication has a five-second deadline; the entire authenticated session has a
+fifteen-second monotonic deadline. Those transport deadlines do not replace the
+native deployment clock or final fenced authority evaluation. A lost permit,
+partial send, lost ACK or ambiguous durable acknowledgement stays charged and
+cannot produce a replacement permit. The recipient explicitly ACKs only after
+complete receipt; retaining that native ACK enables later history recording.
+
+The recipient must obtain its expected effect and policy through its trusted
+private authoring channel. This transport relies on host-enforced UID and socket
+isolation; processes trusted under the gateway UID and a malicious host are
+outside the containment guarantee. The existing synthetic-grant journal corpus
+remains historical unit evidence. A separate real socket-pair/file fixture covers
+actual send/ACK loss, reopening, one-winner concurrent grants, grant-before-send
+loss and changes between authentication and grant. It is not independent-process,
+PostgreSQL, hostile-agent or full installed A07/A08/A22 evidence. The application/CLI controls below do not close installed binding/UI integration
+or those complete journeys.
+
+The application now has closed `auths.gateway-content/1`,
+`auths.gateway-disclose/1`, `auths.gateway-content-history/1` and
+`auths.gateway-disclosure-acknowledge/1` frames. Acquisition carries only the
+native proof/action/approval envelope. Its resource arguments are projected from
+the verified action through the same acquisition compiler; no second caller map,
+URL or transport override is accepted. The existing eight-MiB application frame,
+four-MiB proof, 64-KiB action and native approval bounds remain in force.
+
+Disclosure carries a closed source-operation/recipient-key/channel/instance
+proposal and its independent native envelope. The application resolves protected
+source provenance and invokes native admission. Terminal replay returns redacted
+history; a reservation returns `ready` only to switch the owned socket to native
+recipient authentication. Ready is not a grant. The engine then repeats current
+compilation/admission and makes the existing final fenced decision. Changes after
+Ready fail closed. Transport inherits the application session deadline as well as
+its own tighter limits. No external send is cancelled into a false non-entry
+claim, and the session retains capacity until the bounded native work ends.
+
+The native recipient client checks gateway UID, expected binding/channel and its
+key before sending the private request. It accepts at most 4,096 bytes of control
+response, switches only on the typed Ready response, and exposes plaintext only
+through the existing authenticated receipt value. The CLI saves content and
+canonical late-ACK evidence to new owner-private files before sending its ACK;
+`recipient-ack-sent` does not assert that the gateway persisted it. History and
+late ACK use separate control calls and never request a fresh delivery. History
+is restricted to the installed recipe, namespace, partition and epoch and uses
+native redacted acquisition/disclosure projections.
+
+`effect-review --content` and `--disclosure ` use native complete
+review and approval requests. Optional `--binding-out` preserves the exact native
+binding for the trusted recipient channel. Python/TypeScript content acquisition,
+history and late-ACK methods carry native bytes unchanged; they implement no
+content projection, commitment, signing, accounting or retry decision. Recipient
+protocol execution stays in Rust. Native frame/CLI artifacts and socket-client
+cases are narrower than installed authority/UI/binding/hostile-process evidence;
+authenticated UI/installed managed binding integration remains required. AP070
+remains an independent application-owned reference; its cutover is deferred.
+
+Native `effect-content-call` accepts a private resource-argument file and uses
+`ContentReadTarget` against the installed connection. It reuses the bounded
+65,536-byte duplicate-rejecting JSON leaf; extra fields, nonobjects and invalid
+resource values fail closed. `effect-disclosure-call` resolves the retained native
+source from its private recipient proposal and uses `DisclosureTarget`. Both save
+canonical native calls to new absolute 0600 files; neither creates authority,
+approval, reservation, a custody lease or external entry.
+
+`content-process-journey.json` now executes these commands, private native review,
+exact approval, the real daemon dispatcher and recipient CLI. A counting local
+HTTP double observes one content GET, one approved PATCH and one independently
+authorized observation GET. Restart preserves acquisition/disclosure history and
+refuses redelivery. In a second disclosure a separate native recipient process
+saves content/ACK but sends no ACK; killing both peers leaves unknown charged
+history after reopen, and exact late ACK refines it without another delivery.
+A separate operator process disables the connection before a newly approved
+release; the recipient receives no content. Public outputs exclude fixture
+content/credential sentinels. These are local file/process observations with
+public test keys simulating the reviewer, actor and recipient. They do not close
+human approval UI, installed language packages, hostile UID/egress containment,
+PostgreSQL, every crash boundary or live-provider qualification gates.
+
+The AP070 concrete Drive metadata reference is frozen separately from native
+acquisition in `google-metadata-reference.json`. It exercises exact file identity,
+required download capability, document/plain-text selection, optional size and
+trashed semantics, malformed input and the existing 16 KiB bound. The reference
+reader uses one typed domain validator; malformed MIME/decimal inputs deny.
+Historical duplicate, alternate-encoding, nonboolean-trash and Unicode-size
+acceptance are explicitly marked for stricter native refusal. This corpus is an
+optional input to future managed Google capability/differential extraction. The
+existing broker and corpus are preserved; native Google metadata, token refresh
+and Python journal migration are deferred rather than AP074 exit gates.
+
+`content-process-capacity.json` adds a real two-daemon file-store history with
+separate joined local custody. After fifteen source acquisitions, two new exact
+approved reads contend for the final lifetime allocation. One grant/request wins;
+the other native record is not-entered for exhaustion with no new charge. Sixteen
+grant audits, allocations, provider calls and full response ceilings are retained.
+Restart/replay changes neither journal bytes nor provider count. This is finite
+source-lifetime count/byte evidence, not fixed-window count/sum or PostgreSQL
+conformance; those acceptance requirements remain open.
+
+The common `native-process-capacity-{file,postgres}.json` cases additionally use
+full native engine entry validation in separate OS processes over both durable
+adapters, with exact proofs/approvals, fixed test deployment time and separate
+custody. They record one final winner and an exhausted non-entry while an actual
+HTTP double counts sixteen total grants/GETs, and reopening cannot remint access.
+These cases use no final-entry authorization oracle. They establish the stated
+source-lifetime count/byte behavior; fixed-window count/sum, complete fault and
+race schedules and deployed hostile-process containment remain separate gates.
+
+`native-process-crash-{file,postgres}.json` records seven acquisition SIGKILL
+schedules per adapter, covering reservation, entry, partial HTTP response and
+response persistence. Reopening retains original charges and allocation; a lost
+grant cannot be recreated, and a lost response cannot authorize another fetch.
+These transaction-boundary process deaths do not establish actual sync/rename
+failure, partial-batch failure, ambiguous database commit delivery or the complete
+mutation/disclosure crash matrix.
+
+`native-dependency-order-{file,postgres}.json` adds twelve independent worker/
+operator histories per adapter. Connection revocation and configuration activation
+race the actual grant fence in both orders. Separate positive/expired controls
+advance the maintained test clock while entry waits behind a physically distinct
+connection record's writer fence, independently checking human approval and
+native resource authority. Four additional histories per adapter independently
+age a signed observation and change the selected evidence in a real 1-of-2 proof.
+At the inclusive five-second freshness boundary, both positive controls enter.
+At six seconds, single-branch observation evidence is stale; the threshold proof
+remains natively authorized through its longer-lived branch, but only its bound
+native authority projection changes. The original exact approval remains current
+and cannot approve that changed effect. Actual final fenced re-verification
+refuses entry and retains the original reservation and charges. Signed proof,
+action, observation and approval bytes and admission time remain unchanged;
+valid entry records the time after waiting. The schedule occurs after initially
+authorized admission and a custody lease; the final refusal prevents any HTTP
+entry and does not claim to undo that prior lease. These cases establish native
+acquisition ordering under the stated test deployment.
+
+`native-mutation-entry-order-{file,postgres}.json` and
+`native-disclosure-entry-order-{file,postgres}.json` extend the same schedules to
+the typed native mutation and disclosure owners, with twelve histories per
+protocol/adapter. Both preserve original admission and signed input bytes across
+real independent worker/operator lock waits. Changed selected authority cannot
+reuse the old exact approval even while the threshold proof remains Authorized.
+Mutation captures the exact request compiled for its own signed proof, including
+its declared echo. Disclosure separately authenticates a native recipient and
+records an exact ACK only after a grant. An opaque socket relay inspects every
+gateway byte after the authentication challenge: final-grant refusal produces
+zero subsequent wire bytes, including incomplete frame prefixes. Gateway and
+recipient are same-UID tasks within the worker, not hostile OS peers. The signed
+observation fixtures use public test observer keys and do not qualify live-provider
+evidence. Complete dependency/revocation scopes and installed language/human UI,
+adverse durability and containment evidence remain separate requirements.
+
+`native-independent-recovery-{file,postgres}.json` exercises four outcomes per
+adapter using separate native writer and recovery-reader OS processes. A real
+operator configuration activation withdraws the writer root while retaining an
+independent reader root/actor. Previously valid, exactly approved new write and
+recovery-read submissions from the writer then fail native `untrusted-root`
+verification without admission, custody or provider entry. The original proof
+and approval remain within their signed validity windows; this is authority
+withdrawal rather than expiry. The current reader obtains its own native read
+proof and exact approval bound to the original source effect. Source approval
+cannot stand in for recovery approval.
+
+Actual local PATCH/GET exchanges distinguish a recorded response, a lost response,
+matching state and recipe-declared echo linkage. Matching state without that
+link leaves a lost-response source Unknown and the recovery Inconclusive. A
+supported echo link records provider-linked evidence with the original response
+status still absent; it does not invent a write response. Recovery preserves
+source identity, actor attribution, signed evidence, entry audit and charges.
+Reopened replay preserves both journals and performs no new lease or HTTP call;
+retained writer history remains readable after current authority is withdrawn.
+These are native file/TLS PostgreSQL observations with public test keys, a
+maintained clock and trusted host/operator/stores. Signed principal/grant status
+service revocation, installed human UI/language paths, hostile-process isolation
+and live-provider qualification remain separate evidence requirements.
+
+`native-window-capacity-{file,postgres}.json` separately exercises native
+bounded resource grants and exact approvals across independent worker processes.
+After a prefix acquisition, two new operations contend for the last count slot,
+byte-sum allowance or both within the same fixed window. The loser retains an
+uncharged not-entered record; the winner retains the original full response
+ceiling. Native slot witnesses and reopened history preserve window, cumulative
+charges and exact record bytes without another lease or HTTP request. These
+acquisition cases do not replace mutation/disclosure fixed-window evidence,
+cross-window expiry, partial-batch faults or the other A04/A06 requirements.
+
+The `native-file-io-faults.json` generator exercises fifteen injected file-adapter
+failure schedules in independent native worker processes. A test-only call-site
+hook returns an I/O failure at data sync, journal/target installation, record
+replacement or directory acknowledgement after the preceding real operations.
+The four-record bounded admission batch must recover all-or-none; a retained
+reservation can resume only through fresh native entry checks; a retained grant
+after a lost acknowledgement cannot mint another permit. Reopened terminal
+history must preserve all charges and journal bytes without a new lease or HTTP
+request. The failed I/O return is simulated, and production builds contain no
+selector. Actual device faults/power loss, PostgreSQL commit delivery and the
+complete mutation/disclosure fault matrix remain separate requirements.
+
+`native-mutation-file-io-faults.json` adds fourteen native mutation schedules for
+the three-row admission batch, grant and response persistence.
+`native-disclosure-file-io-faults.json` adds eighteen release schedules for its
+four-row admission batch, grant, exact ACK and Unknown persistence. Both use the
+same actual file writer fence and syscall-error selector as acquisition, with
+typed stage paths and operation keys. They cover batch data sync/publication,
+each target install and all three admission directory-sync points, plus data
+sync/replace/directory sync at the modeled grant/outcome boundaries. Error returns
+are simulated; preceding writes, rollback/reopen, native decisions, transport
+and recipient signatures execute unchanged. Existing source and charge bytes
+remain intact. A lost grant acknowledgement sends nothing and never remints a
+permit; absent/reserved work continues only through native reauthorization.
+Separate native recipients and opaque Unix wire capture verify exact delivery
+or zero content, and genuine historical ACKs refine retained uncertainty without
+redelivery or new charging. These histories do not qualify actual kernel/device
+EIO, power-loss/whole-store rollback, hostile OS containment or live providers.
+
+`native-postgres-partial-batch.json` separately rejects each insert of a
+four-record native admission transaction using a server trigger confined to an
+owned test schema. A nontransactional sequence witnesses the reached row prefix,
+including an explicit `is_called` check so an unused sequence's initial value
+cannot masquerade as a rejected first insert;
+actual gateway-table counts must show complete rollback before any credential
+lease or HTTP entry. Removing the test fault permits explicit native
+reauthorization of the unchanged operation, followed by no-I/O terminal replay.
+This is real PostgreSQL transaction-abort evidence, not commit-acknowledgement
+ambiguity, power loss or the complete mutation/disclosure fault matrix.
+
+`native-mutation-postgres-partial-batch.json` extends row-prefix rejection to all
+three rows of a real native Stripe mutation admission: attempt, count and amount
+sum. `native-disclosure-postgres-partial-batch.json` covers all four release rows:
+attempt, lifetime allocation, delivery count and byte sum, while preserving the
+already acquired source and its charges byte-for-byte. The failure workers make
+no custody or provider calls; failed release admission sends no authentication or
+content bytes. After the trigger is removed, explicit full native reauthorization
+of the unchanged signed inputs sends exactly one mutation or authenticates a
+separate native recipient OS process and records its exact ACK. Replay preserves
+all inspected rows and sends nothing. These are transaction-abort and ordinary
+reopen histories over the maintained TLS adapter, not ambiguous commit delivery,
+power loss, hostile OS containment or the complete A06 durability matrix.
+
+`native-postgres-commit-reply-loss.json` separately exercises real lost COMMIT
+replies for admission, entry and response persistence. A deferred fixture trigger
+pauses the actual COMMIT; a direct TLS observer verifies the selected backend's
+TLS session and distinguishes committed state from an aborted transaction. An
+opaque TCP proxy withholds replies and disconnects the native client without
+terminating TLS or manufacturing an adapter error. Reopening must inspect the
+retained native history, continue only absent/pre-grant work through fresh native
+checks and never recreate a lost grant. Charges and exact terminal rows remain
+unchanged through replay. These acquisition histories do not replace the full
+mutation/disclosure commit-fault matrix or establish power-loss/rollback safety.
+
+`native-mutation-commit-reply-loss.json` extends those committed/aborted COMMIT
+histories to native mutation admission, final entry and response persistence.
+`native-disclosure-commit-reply-loss.json` covers release admission, entry, ACK
+and Unknown persistence with gateway and recipient in separate OS processes.
+Both use the same opaque end-to-end TLS proxy and independent server observer;
+only typed native record paths/operation keys select the fixture barrier. The
+observer addresses records by kind and key and snapshots every record family.
+A grant whose transaction committed but whose reply was lost remains charged
+Unknown even if no provider request or content frame was sent. Absent/pre-grant work continues only
+through explicit native reauthorization. Recorded responses and recipient ACKs
+remain distinct retained facts, and terminal replay emits no new frames or I/O.
+An aborted ACK write can be followed by native Unknown persistence; an exact
+historical ACK from the recipient may refine it later without any new delivery
+or charge. These cases add fourteen transaction-outcome histories. Remaining
+file/adverse schedules, server/power-loss and supported rollback assumptions and
+hostile OS/provider qualification remain separate requirements.
+
+A provider-derived locator is not authority. If a later action depends on a response-derived resource, establish its provenance under the declared observation contract and compile a new bound effect linked to that evidence. Do not mutate an already approved effect. This requirement does not expand scope into a full create/populate product workflow.
+
+Public mutation replay evidence now uses `auths.gateway-public-evidence/1`:
+channel, a hiding `evidence_commitment`, and the original bounded receipt time.
+This result projection contains neither raw provider evidence SHA-256 nor the
+action-derived echo.
+The journal owner computes HMAC-SHA256 under its existing private immutable
+128-bit attempt nonce, with domain `auths.gateway-public-commitment/1`, a NUL,
+u32 big-endian length plus canonical binding bytes, u32 big-endian label length
+plus the `provider-evidence` label, and the private 32-byte evidence digest.
+This scopes the hiding commitment to every immutable binding dimension and the
+private record. Replay/reopen retain the same value; different records hide
+equality of identical private response bytes. Raw evidence and its digest remain
+protected journal/audit facts. This projection grants no execution authority.
+
+### 4.4 Stage types and serialization
+
+The intended progression is:
+
+```text
+VerifiedAction -> BoundEffect -> ApprovedEffect -> ReservedEffect
+ -> ProviderEntryPermit or DisclosureEntryPermit -> Evidence
+```
+
+Private constructors enforce the local boundary. Entry permits are neither cloneable nor deserializable. Persisted records are facts, not executable capabilities; restoring a record cannot reconstruct a lost send permit. A portable binding may be decoded as data but must pass native validation before obtaining any sealed type.
+
+Remote permits and multi-executor dispatch are outside this milestone. Do not imply that a Rust type alone prevents a malicious process from bypassing a deployment boundary.
+
+Public submission-result objects have a closed field inventory for every variant,
+including `unknown`. Refusal codes contain 1–128 UTF-8 bytes; journal revisions
+are integers in 1–2^53−1; HTTP statuses are integers in 100–599. A provider-linked
+result always includes `status`, using explicit null if no write response was
+recorded. Missing fields, duplicate or escaped-alias members, unknown fields,
+invalid Unicode, boolean/fractional/exponent numeric spellings and limit-plus-one
+values are refused consistently by Rust and the affected public bindings.
+
+### Signed public history and operator-private audit
+
+Application outcome queries sign `auths.gateway-outcome/3`. Its closed facts are
+`effect-reference`, `recipe-reference`, `record-commitment`, `stage` and
+`evaluated-at`, with bounded `http-status` or `refusal` when applicable. Each
+reference is lowercase HMAC-SHA256 under the record's existing private immutable
+128-bit nonce, the full exact canonical binding and a distinct closed label
+(`action`, `recipe` or `journal-record`) under
+`auths.gateway-public-commitment/1`. The journal-record seal covers the complete
+canonical record including entry/fence/authority/accounting metadata, with the
+nonce zeroized in the sealed message. Raw action, recipe, response, evidence,
+counter, basis and pre-entry digests or values MUST NOT appear in public history.
+A recipe can contain fixed private values, so its public reference is hiding too.
+
+Public pre-entry history returns only the operation ID, observation count (0–4)
+and a hiding record commitment. Protected observation bytes are operator-only.
+The app protocol is `/4`; old public outcome and observe formats are rejected.
+The signature authenticates the gateway's historical statement; no public
+reference, observation or receipt recreates an execution capability.
+
+An authenticated operator requests `audit-export` on the existing peer-UID-gated
+operator socket. One loaded native journal snapshot produces
+`auths.gateway-private-audit/1`: operation ID, signed
+`auths.gateway-private-outcome/1` and protected pre-entry observation bytes. The
+private signed fact set retains the original complete outcome presence predicate
+and every existing audit commitment, counter, basis and pre-entry check. It is
+not an app route and is not an alternative admission or journal owner.
+
+Native CLI `audit-export --operation-id ... --out ...` saves one protected export;
+`--bundle ... --out ...` completes a current submission inventory using those
+same native operator exports. The modes are mutually exclusive. The CLI prints
+only success metadata and writes a new 0600 file in an owner-private directory;
+it never overwrites a file. Bundle assembly grants no audit verdict. Absent
+records remain unverified; failures do not manufacture non-entry, refund or retry.
+The offline auditor accepts only the private schema in audit bundle `/3`, under
+independently pinned trust and observer identity, preserving its existing checks.
+Detailed private bundles/reports MUST NOT return to an app, UI or public artifact.
+
+The existing limits remain: 4,096 observations bytes each, four protected pre-entry
+observations per entry, 4,096 audit entries, 65,536 approval responses and 64 MiB
+aggregate audit bytes. Bounded sequence decoders reject excess items before
+allocating their payloads. Public times fit JavaScript's exact integer range;
+all affected bindings reject duplicate members, noninteger spellings, BOM,
+noncanonical base64 and malformed/obsolete schemas. One CLI bundle export is
+bounded by the existing 60-second operator session deadline and grants no I/O
+capability or provider traffic.
+
+### 4.5 Canonicalization and privacy
+
+Reuse repository canonical codecs and domain-separated commitments. Freeze the selected schema, ordering, optional/absent distinction, integer range, duplicate handling and semantic identities in native fixtures before cutting over callers. No Python/TypeScript hash implementation may become an alternate authority.
+
+The complete effect contains private data. Its ordinary digest and correlation identifiers are not automatically safe for public receipts: low-entropy payloads can be guessed. Define separate private audit and redacted public projections. Use the existing protected-commitment mechanism where suitable; public content commitments must be hiding and scoped. Redact Debug/error output, zeroize retained sensitive buffers where supported, and put any persisted private payload/evidence behind the protected journal's access and retention policy.
+
+## 5. One operation and recovery model
+
+### 5.1 One transition owner
+
+Specify a total bounded reducer conceptually equivalent to:
+
+```text
+transition(current_state, typed_command, verified_facts, transaction_time)
+ -> refusal | proposed_state_and_accounting_delta
+```
+
+The reducer has no I/O, ambient clock, provider callbacks or arbitrary JSON interpretation. The store derives authoritative state and capacity inside its transaction. It persists the reducer result, revisions, accounting delta and required audit event atomically before acknowledging the transition. Only the runtime's validated durable acknowledgement can mint an external-entry capability.
+
+`verified_facts` are stage-specific verified values, not caller assertions. Evidence acquisition can happen before a transaction; freshness and all relevant dependencies must still be checked at its linearization point. The operation's fixed identity and binding never change.
+
+Use protocol-specific typed states beneath a shared operation header. An acquisition followed by disclosure consists of linked operations with distinct authority and charges. A provider GET is not content delivery. A provider response is not observed mutation success.
+
+### 5.2 States and transitions
+
+The following names define semantics, not mandatory wire spellings:
+
+| Transition | Preconditions | Durable result and permission |
+| --- | --- | --- |
+| Admit | Exact approved effect; current authority/configuration/epoch; capacity; unused replay key. | Store binding, reservation and replay claim atomically. No external entry yet. |
+| Refuse before entry | No entry grant exists; refusal is valid under the protocol. | Terminal not-entered fact. Preserve replay identity. Release only explicitly releasable capacity; spent invocation units may remain spent. |
+| Grant provider entry | Reserved acquisition/mutation; fresh native authority and approval; unchanged installation; current revocation/version; exact transport binding. | Persist unique entry grant and required charge before returning one provider permit. |
+| Record provider response | A matching provider entry grant exists; bounded authenticated transport evidence. | Add response fact. Never infer effect solely from HTTP status. |
+| Record uncertain provider outcome | Entry grant exists but completion cannot be established. | Retain charges and unknown outcome. No automatic send retry. |
+| Admit disclosure | Authorized source acquisition and protected content exist; current recipient/scope/channel and projection; disclosure capacity. | Reserve exact release bound to source, protected content commitment and recipient. No delivery yet. |
+| Grant disclosure entry | Fresh release authority and authenticated recipient; exact bytes, channel and commitment; active reservation. | Persist irreversible release charge and unique grant before returning one delivery permit. |
+| Record acknowledgement | Authenticated ACK matches effect, delivery instance, recipient/channel and exact protected bytes. | Add acknowledged fact; never create new delivery authority. Exact repeated ACK is idempotent. |
+| Recover without ACK | Release entry may have occurred. | Retain delivery-unknown and all relevant charges. No automatic redelivery or refund. |
+| Observe/reconcile | Separate current permission for bounded recovery read; exact operation/recipe/evidence binding. | Add supported evidence or leave unknown. Never issue a mutation permit. |
+| Revoke | Authenticated revocation under the same ordered admission authority. | Prevent later grants; retain earlier grants, charges, identity and historical evidence. |
+
+Entry-granted records mean permission was durably handed out and external entry may have happened. They do not prove that a socket send occurred. If the process dies after the grant but before send, the conservative result is still unknown. A missing response cannot be converted to not-entered.
+
+```mermaid
+stateDiagram-v2
+ [*] --> Reserved: atomic admission and reservation
+ Reserved --> NotEntered: close before any entry grant
+ Reserved --> EntryGranted: persist one grant
+ EntryGranted --> Unknown: crash or no conclusive result
+ EntryGranted --> ResponseRecorded: provider transport evidence
+ EntryGranted --> Acknowledged: disclosure ACK
+ ResponseRecorded --> Observed: separate supported observation
+ Unknown --> Observed: sufficient reconciliation evidence
+ Unknown --> ResponseRecorded: late exact source response
+ Unknown --> Acknowledged: late exact disclosure ACK
+```
+
+This diagram is a projection of a typed state model. `ResponseRecorded` belongs to acquisition/mutation, `Acknowledged` to disclosure. `Observed` carries a typed claim such as matching state or linked provider evidence; it is not a universal success state. A matching observation may coexist with unresolved causal attribution; it must
+not erase uncertainty, release charges or imply permission to retry. Invalid
+cross-protocol transitions must be unrepresentable or refused by the sole reducer.
+
+The native recipient evidence contract is `auths.gateway-disclosure-ack/1`,
+bounded to 4,096 canonical bytes before decoding. It signs the exact effect,
+source effect, protected content commitment, raw-key recipient and Ed25519 key,
+channel, delivery instance, projection, byte ceiling and actual acknowledged
+byte count. The key must derive the bound recipient through the existing
+`raw-key-v1` principal mechanism; a key supplied by the envelope is not a new
+trust anchor. The actual byte count is compared with protected content metadata
+from the journal/content owner. Unknown fields, duplicates, alternate encodings,
+obsolete schemas and noncanonical signatures are refused. All correlations stay
+on the private channel; the ACK is not a public receipt.
+
+ACK verification deliberately has no current permission, approval expiry or
+clock input. Its authenticated historical assertion must still match a retained
+delivery grant and pass the sole reducer. Only disclosure may carry immutable
+ACK bytes and reach `Acknowledged`; provider protocols cannot reinterpret them
+as response or observation evidence. Exact repeated canonical ACKs return the
+existing revision without a state transition. The signature establishes the
+recipient's assertion under the pinned identity, not human attention or proof
+of the transport implementation. This evidence contract and pure predicate do
+not themselves implement content acquisition, disclosure admission or delivery.
+
+### 5.3 Retry and recovery contract
+
+For this milestone, provider-entry and disclosure-entry permits are one-use, and replay never reissues them. The model distinguishes re-running pure verification, retrying an uncommitted transaction, retrieving existing state, repeating an authorized observation, and repeating an external effect.
+
+- A transaction proven not committed may be attempted again under the same replay identity.
+- An ambiguous commit requires loading the authoritative journal before doing anything else. Unavailable state fails closed.
+- A durable reservation with no entry grant may progress only through fresh authorization and an atomic single-winner grant; reopening it does not bypass approval or capacity checks.
+- Once an entry grant exists, restart must not reissue it, including when no response or ACK exists.
+- The native submission projection names a retained pre-grant reservation as
+ `reserved` with its exact JSON-safe journal revision. After an unacknowledged
+ grant, loading that state does not issue a capability. A retained grant with
+ no conclusive evidence projects `unknown`; unavailable or mismatched journal
+ state projects `indeterminate` with `gateway.attempt.state-unavailable`.
+ Identical admitted/terminal replay returns stored facts with no credential
+ lease, provider read or new observation credit. Reservation continuation is
+ the separate fresh-authorization rule above.
+- Historical submission matching MUST use the retained native proof and action
+ commitments plus a valid signature for the retained binding under its retained
+ approval policy. Operation IDs or action bodies alone are insufficient. The
+ private journal retains the bounded canonical native authority projection
+ needed to validate those commitments; it is immutable under every protocol
+ transition and grants no execution authority. Current authority, approval
+ expiry, connection revocation or configuration replacement MUST NOT turn an
+ already entered fact into `not-entered`. Extra API resource/source/recipient
+ inputs remain bound to the historical signed action. The same retrieval rule
+ applies to mutation, acquisition and disclosure. A fresh authorization failure
+ for an unchanged pre-grant reservation returns the retained reservation without
+ continuing it, refunding charges or modifying the journal.
+- Provider idempotency may improve provider guarantees but does not override the default prohibition on Auths re-entry. Any future retry mode requires an explicit versioned protocol and evidence for key scope, retention, fencing and late requests.
+- Late valid response/ACK evidence can refine an unknown outcome. Revocation or expiry must not erase legitimate history or prevent recording a previously authorized delivery's exact ACK. Recording history confers no current execution authority.
+- The combined recovery entry point MUST use the same exact historical lookup
+ as acquisition before attempting current reader compilation. Expired or
+ withdrawn reader authority and a disabled connection prevent new entry; they
+ do not erase completed, unknown or pre-grant reserved read facts. Historical
+ recovery obtains its read operation identity from that validated journal match,
+ while new recovery obtains it from the native compiled binding. A retained
+ reservation can resume only through fresh authorization and the final grant;
+ returning its history cannot resume it. Source/read refinement still requires
+ a usable deployment clock and available journal, and never refreshes receipt
+ time or fabricates a zero-usage report on failure.
+- Read-only recovery requires its own bounded authority and accounting. It cannot borrow a revoked writer's proof, use an expired approval as current authority, acquire broader credentials or change the original actor attribution.
+
+Unresolved effects also have an operational contract. Declare finite observation
+attempt/byte/time limits, evidence freshness, and escalation conditions. Expose
+the uncertainty, retained charges, last evidence and permitted next actions in
+native operator projections. Exhaustion or unavailable evidence leaves the
+effect unresolved and charged; it does not schedule an implicit retry, redelivery
+or refund. Recovery may remain unresolved indefinitely.
+
+The native `auths.gateway-recovery/2` projection reports source-lifetime retained
+observation allocations and response-byte reservations, the per-observation and
+aggregate transport deadlines, the evidence-age limit, and the selected read's
+complete-response receipt time when available. Usage is read from all sixteen
+permanent allocation rows under the same journal fence as source refinement;
+changing read operation IDs, replaying, restarting, or refusing a seventeenth
+read cannot reset it. The aggregate transport deadline is the sum of the sixteen
+observation transport deadlines, not a wall-clock completion bound or a bound on
+separately authorized prerequisite transports. Retained byte reservations are
+not actual received-byte measurements and are never refunded for failed reads.
+
+Unusable, incomplete, stale or inconclusive evidence produces native escalation
+guidance; unresolved exhaustion takes precedence. Proposed next actions are
+inspection, requesting a separately authorized observation while capacity
+remains, and escalation. These projections grant no capability and contain no
+retry, redelivery, non-entry assertion, evidence override or refund command.
+Unavailable or corrupt journal/clock inputs return typed unavailability rather
+than fabricating zero usage or an actionable report.
+
+Administrative annotations must remain distinguishable from verified provider
+evidence. Any administrative resolution that changes evidence or accounting must
+be an explicit, separately authorized command accepted by the sole reducer under
+a declared evidence rule. An operator assertion cannot establish non-entry after
+a grant or override the conservation/no-retry rules. Measure bounded recovery
+work and the cases actually resolvable, as well as refusal safety.
+
+### 5.4 Accounting
+
+At every committed revision, for each capacity key and unit:
+
+```text
+available + held + consumed = allocated
+held >= 0; consumed >= 0; allocated <= authorized ceiling
+```
+
+If capacity windows or allocation semantics use a different existing algebra, provide the exact equivalent conservation relation and proof. Never mix units or double-count one operation in multiple categories. Unknown effects retain their held/consumed allocation. Read acquisition count, provider attempts, response-byte reservation, disclosure bytes and mutation units have explicit charging points.
+
+The reducer owns reserve/consume/release transitions. Store adapters cannot silently release capacity on timeout, lease expiry, cancellation, garbage collection or process death. Only an explicit protocol transition with evidence of no entry may release a releasable reservation. Conservative over-accounting is permitted and must be visible; accidental refunds are not.
+
+### 5.5 Revocation, time and configuration ordering
+
+Entry grants and revocation must have one documented linearization order. A revocation committed first blocks the grant; a grant committed first may still be exercised. Do not promise recall of already released data or an unfenced remote request. Disclosure has a later, independent gate even after provider acquisition was granted.
+
+Use the existing deployment clock boundary and sample transaction time after acquiring the relevant lock/transaction. Bound lock duration and native verification work. Distinguish signed event time, evidence observation time, verifier evaluation time and monotonic elapsed time. Re-evaluation does not refresh old evidence.
+
+Re-run the full native verifier, or use a mechanically justified equivalent dependency check, at the final authorization boundary. Checking only action/grant expiry is insufficient when identity/status/assurance/approval/observation evidence can age independently. Bind the selected branches and evidence to the projected effect: another valid threshold subset cannot silently license previously derived facts.
+
+Connection, trust, recipe, policy, qualification and admission epoch changes must be ordered with entry or fenced under a demonstrated version protocol. A mutable value checked before waiting for the lock is insufficient. Exact human approval must remain eligible at entry. If durable consent has a distinct validity/revocation contract, make that a separate mode with explicit evidence; never discard approval dependencies accidentally.
+
+Each operator namespace/partition has one active configuration control across
+epochs. Its key must not create a second authority when the epoch changes.
+Explicit operator activation advances the epoch under the same writer fence;
+a stale host or a fresh review token for a lower epoch cannot reactivate it.
+Historical operation keys remain epoch-scoped, and retained facts can be read
+without creating new entry authority. Advancing epochs must preserve permanent
+revocation memory and source lifetime capacity. This monotonic control assumes
+the declared trusted durable store; it is not an independent rollback anchor.
+
+Initial admission also orders fresh native authority, approval, configuration,
+connection and qualification with publication of the claim and its charges.
+The file and PostgreSQL adapters use the common writer fence for one insert
+batch plus bounded dependency reads. The decision can replace only the first
+record's bytes; allocation keys, kinds, amounts and expiries remain fixed.
+Persisted admission time is sampled inside that fence. A replay returns retained
+history without a new admission decision or charge and confers no entry permit.
+
+Count/sum frontiers may be selected outside the fence only as candidates. The
+transaction checks every target's vacancy and validates immutable sum
+predecessors against the proposed cumulative charges. Native recompilation must
+refuse a changed window before publication. Acquisition admission additionally
+checks its eligible source under the same fence. The combined insertion and
+read set is at most 64 records (at most 55 for mutation admission); the existing
+two-second lock-wait and decision bounds apply. Store failure or ambiguous
+acknowledgement cannot publish an executable capability.
+
+At the entry grant's serialization point, the atomic decision requires current
+native authority, eligible approval, current configuration and revocation state,
+an intact sufficient reservation with conserved accounting, and no previous
+grant for that operation's external-entry boundary. Already reserved capacity
+must not be charged again as a new allocation. The full check and durable grant
+must observe one consistent authoritative state, or an explicitly justified
+equivalent fencing protocol. One durable grant does not prove exactly-once
+provider execution, settlement or exclusive causation.
+
+## 6. Journal and persistence contract
+
+There must be one logical journal owner for operation identity, reservation, entry grants, protocol outcomes and replay tombstones. Multiple physical records are acceptable only when their safety-critical changes commit atomically. An effect must not acquire two independent mutable truths in a task ledger and an attempt ledger.
+
+For both file and PostgreSQL adapters, specify and exercise:
+
+- lock/CAS scope, expected revision, epoch and conflict behavior;
+- atomic operation creation plus all counter updates;
+- atomic entry grant plus its audit event and consumed/held accounting;
+- file/data/directory synchronization or database commit acknowledgement required before capability publication;
+- torn-write, partial-batch, connection-loss and ambiguous-commit recovery;
+- decoding bounds, illegal-state rejection and schema mismatch refusal;
+- replay/tombstone retention through the full authority and late-evidence horizon;
+- bounded garbage collection that cannot resurrect an expired claim or erase unresolved accounting.
+
+Before P2's behavioral cutover, the ownership ADR must name each adapter's exact
+lock/transaction scope, acquisition order, authoritative dependency versions,
+configuration/revocation writers, isolation/CAS behavior, clock sampling point,
+finite work/deadline bound and acknowledgement boundary. It must explain why a
+writer in another process cannot change a checked dependency before the grant
+without detection or ordered exclusion. Independently stored configuration or
+credentials require an explicit fencing/version contract; an attempt-only CAS
+or process-local in-flight counter is insufficient. Unsupported cross-store
+atomicity must fail closed, not become an undocumented deployment assumption.
+
+Existing file and PostgreSQL behavior must not regress. In-memory adapters are test tools and do not establish durability. File conformance establishes the stated local-filesystem assumptions, not distributed fencing or power-loss behavior on every platform.
+
+Whole-store rollback is outside ordinary crash recovery. Keep existing commissioning/generation-floor defenses, bind operations to the admitted epoch, and fail closed when an installed anti-rollback check fails. Where an independent monotonic anchor is absent, explicitly state that restoring all protected state is outside the guarantee. Do not claim that a revision stored in the same restored database detects its own rollback. Building a new consensus service is outside this specification.
+
+Persist sensitive reconstruction data only if necessary and protected. If private content is intentionally not retained, recovery may remain unresolved; it must not refetch or redeliver to manufacture certainty.
+
+Native journal replay accepts only the current native writer encoding. Reject
+obsolete schemas, duplicate/unknown members, alternate encodings, operation
+revisions outside `1..=64`, non-JSON-safe event times and inconsistent
+binding/protocol/evidence before history can confer any retained execution fact.
+Admission is durable revision 1;
+all operation advancement, including entry and recovery refinement, shares the
+64-revision cap. At the exact limit terminal history remains replayable without
+I/O; further advancement fails before persistence or capability publication.
+Corrupt state is indeterminate; it cannot establish that an operation was not entered. Adapter byte limits apply
+before decoding. Runtime must not repair, migrate or re-sign corrupt state.
+
+## 7. Central implementation surfaces
+
+The implementation should expose a small set of Rust-owned operations, with exact names chosen during the source mapping:
+
+| Surface | Contract |
+| --- | --- |
+| Compile/review | Produce canonical bound effect and human review projection from verified action and pinned installation. No reservation or I/O. |
+| Approve/validate approval | Bind exact effect or explicit bounded consent. No entry permission. |
+| Admit | Atomic binding/replay/reservation transaction; return existing state on exact replay. |
+| Grant entry | Fresh ordered authorization and one durable provider/disclosure grant; return one-use capability only after acknowledgement validation. |
+| Record evidence | Validate exact response, ACK or observation binding and apply the sole transition relation. |
+| Recover/inspect | Return truthful current facts and permitted recovery actions. Never implicitly execute a write or deliver content. |
+| Project receipt/activity | Native redacted projection of journal facts with evidence provenance and limitations. |
+
+Centralize clocks, canonicalization, effect comparisons, charge transitions and receipt claims at these owners. Do not scatter fixes across Python presenters, HTTP routes, CLI handlers or recipe branches. UI and SDKs must consume the same native review and state projection, including unknown and not-entered distinctions.
+
+Managed gateway consumers must call the native shared owner directly; remove
+competing decisions for the same managed operation in that cutover. AP070 is a
+retained independently application-owned reference, not such a managed consumer.
+Keep its broker and narrower claims intact and use its independent behavior for
+comparison where contracts overlap. Inventory retained application-owned paths,
+credentials and claims, and prove they cannot bypass the managed deployment's
+credential/channel boundary. Do not load a reference adapter into the gateway or
+promote its decisions/receipts to managed authority. Future managed Google access
+requires its own explicit integration decision and one native admission owner.
+
+## 8. Hard limits and stable failures
+
+Before the first behavioral cutover, commit a limit/encoding table naming the existing constant or new justified finite bound for every public collection and byte buffer. Preserve current tighter limits, including native proof/action bounds, 16,384-byte compiled write bodies, 34 counter entries and the 64-step gateway driver bound unless an explicitly reviewed requirement changes them.
+
+Cover: effect metadata; payload/proof/context bytes; header/path sizes; resource and recipient counts; reservation keys; evidence dependencies; per-operation journal events; ACKs; observation attempts; lock/transaction deadlines; recovery scans; private retained bytes; and public receipt size. Decode and reject over-limit inputs before unbounded allocation. Do not solve exhaustion by silently deleting replay or unknown-outcome records.
+
+Errors distinguish malformed/noncanonical input, unsupported schema/protocol, unauthorized effect, expired/stale evidence, changed binding, replay, conflicting operation, exhausted capacity, invalid transition, unavailable/corrupt store and unresolved outcome. Map them consistently across Rust and bindings without exposing proof bodies, provider content or credentials. Preserve existing codes where semantics are unchanged; version intentional changes.
+
+Operator command objects use `MAX_ADMIN_REQUEST_BYTES = 4,096`, enforced on the
+advertised frame length before allocation and again before decoding. Native proof,
+approval and rotation-secret second frames keep their separate existing limits.
+Operator control members must be unique string fields; escaped aliases, duplicate
+schema/command/argument members and arguments for another command are invalid.
+The command enum and application-frame enum each generate their own native
+interface catalog. The CLI inventory is generated from the built Clap command
+metadata, including generated help, rather than a second runtime dispatcher.
+
+## 9. Required invariants
+
+| ID | Invariant |
+| --- | --- |
+| E01 | Every external request and disclosure derives exclusively from one native-verified, compiled and approved bound effect. |
+| E02 | Changing any security-relevant binding dimension is rejected or creates a newly authorized effect; no stage reconstructs a weaker interpretation. |
+| E03 | Same operation key plus same binding is replay; same key plus different binding is conflict. Neither creates another entry. |
+| E04 | Human approval, native resource authority and live admission are separate required facts. |
+| E05 | Bindings and UI consume native canonical identities and projections; public data does not disclose private content or reusable authority. |
+| R01 | One reducer owns allowed state transitions and accounting; every maintained managed path calls it. |
+| R02 | Durable reservation and the entry grant precede their corresponding external boundaries; failed/ambiguous persistence returns no capability. |
+| R03 | At most one entry capability is issued per effect boundary; permit loss, replay and restart cannot mint another. |
+| R04 | Accounting is conserved under concurrency, errors, revocation, restart and unknown outcomes. |
+| R05 | Acquisition, disclosure, provider response, acknowledgement and observation remain distinct facts. |
+| R06 | Unknown cannot become not-entered, refunded or retryable from elapsed time or missing evidence. |
+| R07 | Revocation, configuration and entry follow the stated transaction/fencing order; time is evaluated at that boundary. |
+| R08 | Evidence is immutable once recorded; late evidence may refine uncertainty only for the exact original operation. |
+| R09 | Recovery has explicit current read-only authority and preserves original effect attribution. |
+| R10 | Restored state and malformed/obsolete records cannot bypass the declared epoch, retention or durability assumptions. |
+
+The implementation must map these invariants to six system properties:
+
+| Property | Conditional guarantee |
+| --- | --- |
+| Authority preservation | Every external entry follows a grant with valid current native authority and eligible approval at its ordered grant point. |
+| Effect preservation | Every request/disclosure is the exact compiled and approved effect, including its review contract and allowed credential insertion. |
+| Single-grant safety | Each logical operation and external-entry boundary has at most one grant; persisted state cannot remint its permit. |
+| Capacity conservation | Concurrent admission, crashes and recovery cannot create authorized capacity or silently refund unknown effects. |
+| Revocation ordering | A committed revocation prevents every later grant within its defined scope; earlier grants retain the documented semantics. |
+| Evidence soundness | Recorded and displayed claims imply no more than their authenticated evidence establishes. |
+
+For each property, name the shipping pure predicates, runtime enforcement,
+adapter assumptions and model-generated/integration histories that support it.
+These are required proof/evidence obligations, not a claim of an already proved
+whole-system theorem. Formal claims must state their assumptions and source link.
+
+## 10. Executable acceptance matrix
+
+All rows require reproducible artifacts, expected journal state, expected external calls, actual result, exact revision and limitations. Source code or a passing compilation alone is not behavioral evidence.
+
+| Case | Required scenario and assertion |
+| --- | --- |
+| A01 | Compile/review/approve/execute the same mutation. Assert exact request bytes and one identity at every boundary. |
+| A02 | Independently mutate action, actor, scope, recipient, resource, recipe, URL/body/header, account, credential generation, verifier configuration, epoch and approval. Refuse before the applicable custody/entry boundary. |
+| A03 | Replay identical operation concurrently and after restart; alter its effect with the same ID; try a new ID with the old exact approval. At most one grant, explicit replay/conflict, conserved counters. |
+| A04 | Race independent operations for the final count/byte/sum capacity. Exactly the permitted total succeeds on both durable adapters. |
+| A05 | Crash before commit, after reservation, after entry persistence but before send, during send, after response/ACK receipt but before persistence, and after persistence but before returning to caller. Reopen and verify state, charges and no duplicate effect. |
+| A06 | Inject file sync/rename failure, partial batch and database ambiguous commit/CAS conflicts. No unacknowledged capability; deterministic bounded recovery. |
+| A07 | Acquire content then revoke before disclosure. Provider entry may be recorded; recipient receives no bytes. Charges reflect the separate boundaries. |
+| A08 | Lose disclosure ACK after actual send; crash/reopen and retry. No automatic redelivery; bytes stay charged; exact late ACK refines history. |
+| A09 | Replay/substitute ACK for another operation, channel, recipient, content, instance or source effect. Refuse without changing history/accounting. |
+| A10 | Provider timeout, response loss, denied response, successful HTTP response and matching read-back produce different truthful claims. A match alone is not causal proof. |
+| A11 | Reconciliation finds zero, multiple, unrelated or incomplete matches. Outcome remains unresolved. A supported exact link adds evidence without a new write. |
+| A12 | Race revocation and connection/configuration changes with entry and disclosure. Histories match the documented linearization rule. |
+| A13 | Advance time while waiting for the lock; independently expire human evidence, observer evidence and resource authority. Re-evaluation does not refresh signed event time. |
+| A14 | A threshold proof remains authorized but its selected branches/evidence change. Earlier projected authority cannot be silently reused. |
+| A15 | Recovery actor differs from original actor; original permission is expired/revoked. Only separately authorized bounded observation is possible; attribution remains unchanged. |
+| A16 | Cross-task, cross-recipe and cross-partition substitutions use otherwise valid proofs and identical resource names. No mixed authority or shared-state aliasing. |
+| A17 | Capture logs, errors, public receipts and binding output with sentinel secrets/private content. No sensitive bytes or guessable public content commitments escape. |
+| A18 | Canonical valid, malformed, duplicate, noncanonical and limit/limit-plus-one cases agree across Rust and every affected public binding. |
+| A19 | Differential corpus covers two mutation recipes with distinct recovery capabilities, plus acquisition/disclosure, against independent retained references. Preserve reference independence and document contract differences; AP070 may remain an application-owned reference without migration. No provider-specific branch enters the shared owner. |
+| A20 | Model-generated event sequences and concurrent histories exercise every transition and illegal edge. Formal/refinement artifacts identify the actual shipping pure predicates and assumptions. |
+| A21 | Replay retention, terminal tombstones, stale cache/configuration and supported rollback-floor failures cannot resurrect entry authority. Unsupported whole-store rollback is explicitly excluded. |
+| A22 | Installed packages and native runtime execute the complete managed single-human/single-agent read, disclosure, exact write and observation journey; UI distinguishes grant, response, ACK, observation and unknown. Use an approved native recipe and real local transport; Google/AP070 integration is not required and an application-owned broker cannot substitute for managed evidence. |
+
+The same rows additionally require the following hardening cases; these are
+part of their exit gates, not optional follow-up work:
+
+| Cases | Additional adversarial evidence |
+| --- | --- |
+| A01, A02, A18, A22 | Unsupported review contract, omitted/truncated mandatory field, misleading untrusted label, changed units/destination/consequences and substituted review/signing channel. No approval of an incomplete or mismatched review. |
+| A04, A06, A12, A13 | Independent processes race dependency changes against the final grant on both adapters, including state in distinct physical records. Histories satisfy the named fencing mechanism; existing reservations are not charged twice. |
+| A11, A15, A22 | Exhaust reconciliation attempts/bytes/time; age its evidence; make authority/store unavailable. Retained charges and escalation remain visible; administrative annotations cannot manufacture non-entry or retry authority. |
+| A16, A20 | Scope/participant and charge inputs cannot create a second admission authority. Exercise the current unscoped/single-agent boundary and reject unsupported delegation representations; no AP073 implementation is required. |
+| A17, A21, A22 | Hostile isolated agent exercises the complete managed interface inventory, obsolete endpoints and alternate credential/transport routes. Assert no unauthorized managed provider calls and document the actual isolation assumptions. Retained application-owned references must not expose managed credentials, permits or an alternate authority for the same managed effect; verify the applicable deployment separation. |
+
+Use external call counters and process failpoints, not only mocked return values or tests asserting the implementation's own intermediate flags. Exercise real local sockets, process restart and the maintained PostgreSQL fixture. Provider-specific live qualification remains a separate claim; no cloud credential should be necessary for the local foundation journey.
+
+## 11. Formal and performance evidence
+
+Preserve current linked gateway order/transition/recovery evidence. Update translated/generated Rust-to-Lean obligations and Kani/property coverage when their source or semantics change. Required properties include exact-binding preservation, single grant, conservation, no unknown-to-retry transition, immutable evidence and revocation ordering. Do not label a separately handwritten model as a proof of shipping Rust.
+
+Publish a bounded mapping from the pure transition proof to the runtime/store adapter assumptions. Filesystem durability, database isolation, authenticated recipient behavior and provider observations remain explicit assumptions tested through conformance/fault injection.
+
+Measure native verification, compilation, transaction/lock duration, memory bounds and recovery scan costs for each representative path. Optimization must retain the same semantics and differential evidence. A second verification under a lock must have a finite work/deadline bound; do not replace full dependency validation with a partial expiry shortcut merely to improve latency.
+
+## 12. Implementation order and exit gates
+
+| Phase | Deliverable | Exit gate |
+| --- | --- | --- |
+| P0: map and decide | Baseline source/consumer inventory; ownership ADR and abstraction case; canonical field, limit and state tables; retain/replace list. | Every existing invariant and production/reference owner accounted for. No implementation built on an unresolved competing owner. |
+| P1: bind one mutation | Native immutable effect, review/approval, deterministic request compilation and thin binding entry points. | E01–E05 on one maintained mutation path; adversarial binding corpus authored; direct cutover removes duplicate reconstruction. |
+| P2: central durable mutation | Sole reducer/journal for reservations, entry, response, unknown and observation, with file/PostgreSQL adapters. | R01–R10 for mutation; restart/concurrency/fault evidence implemented; no dual task/attempt authority. |
+| P3: acquisition and disclosure | Typed managed read/release protocol in the same native owner, explicit source/recipient binding and ACK handling; thin managed bindings. | Independent read/disclosure charges and permissions; ACK loss/revocation/replay cases implemented; no managed consumer independently decides shared transitions. AP070 remains an independent reference. |
+| P4: close integration and evidence | Second recipe differential cases, installed bindings, full local journey, native projections, formal links and final hosted checks. | A01–A22 addressed with exact-revision evidence, documented assumptions and no unsupported claims. |
+
+Commit coherent changes along the way. Keep a durable acceptance ledger separating implemented, source-tested, executed, externally qualified and blocked. Do not mark a phase complete merely because its types exist or the happy path works.
+
+Follow the repository's GitHub-first policy: construct necessary generated artifacts locally and avoid duplicate broad local suites by default. The user has explicitly authorized early CI failure triage; this does not replace final verification. Once implementation, fixtures and documentation are complete, push the finished branch, inspect every required hosted workflow on that exact revision, repair failures, and repeat until the required evidence is resolved. Report external blockers rather than weakening checks or pretending a skipped job passed.
+
+The foundation is not complete until required behavioral and hosted evidence passes. Delaying CI observation is not a waiver of verification. No merge or deployment is authorized by this spec.
+
+## 13. Completion and handoff boundary
+
+Done means both central contracts own the maintained managed paths, obsolete managed interpretations are removed, interfaces and bindings are updated, source-linked formal and operational evidence are current, and the acceptance ledger contains actual results. The result must be a reviewable implementation with a concise explanation of changed behavior, tests and remaining deployment assumptions.
+
+Only after this foundation is closed should multi-agent delegation add participant trees and ancestor budget allocation on top. Those features must consume the established effect and recovery contracts instead of creating their own parallel journals, approval identities or retry rules.
+
+Future AP073 composition must bind the exact participant/executor identity,
+task partition/epoch, native authority chain and selected dependencies, scope and
+typed ancestor/task charges into the effect. These are inputs to AP074's single
+admission authority, never a second independent task approval or mutable effect
+journal. Any cross-domain charge/admission update must commit atomically or use
+an explicitly justified fencing protocol. Reject unsupported delegated scopes
+until that integration is implemented. This contract neither imports AP073 code
+nor authorizes work on its paused checkout.
+
+## References
+
+- [Repository contract](../../AGENTS.md).
+- [Profile and domain abstraction boundary](../target-state/PROFILE_AND_DOMAIN_ABSTRACTION_BOUNDARY_PLAN.md).
+- [AP-SPEC-026: reservation and execution semantics](0026-reservation-and-execution-state-semantics.md).
+- [AP-SPEC-057: exact-action evidence program](0057-evidence-program-for-the-exact-action-boundary.md).
+- [AP-SPEC-063: generalized gateway](0063-generalized-gateway.md).
+- [AP-SPEC-066: gateway qualification](0066-production-gateway-polish-and-recipe-qualification.md).
+- [AP-SPEC-070: employee read delegation](0070-google-employee-agent-delegation-demo.md).
+- [ADR 0012: credential-isolated gateway](../adr/0012-declarative-credential-isolated-gateway-boundary.md).
+- [ADR 0013: recipe capabilities and budgets](../adr/0013-recipe-capabilities-and-sum-budget.md).
+- [Existing gateway abstraction case](../research/domains/abstraction-cases/0007-declarative-credential-isolated-request.md).
diff --git a/docs/target-state/PROFILE_AND_DOMAIN_ABSTRACTION_BOUNDARY_PLAN.md b/docs/target-state/PROFILE_AND_DOMAIN_ABSTRACTION_BOUNDARY_PLAN.md
index 450e5481e..51f39b1d3 100644
--- a/docs/target-state/PROFILE_AND_DOMAIN_ABSTRACTION_BOUNDARY_PLAN.md
+++ b/docs/target-state/PROFILE_AND_DOMAIN_ABSTRACTION_BOUNDARY_PLAN.md
@@ -125,7 +125,8 @@ makes the data-only gateway of
[AP-SPEC-053](../specs/0053-declarative-credential-isolated-gateway.md), as
generalized by [AP-SPEC-063](../specs/0063-generalized-gateway.md), the
production mechanism for credential-isolated provider writes. Its application
-channel supplies proof and canonical action only; an operator separately
+channel supplies proof, canonical action and native exact-effect approval;
+an operator separately
approves the typed recipe digest, trust, connection, and credential
generation. The gateway owns the shared mechanics once: bounded request
construction, the atomic claim, the spend limit, attempt records, recovery,
diff --git a/examples/stripe-refund-approval/README.md b/examples/stripe-refund-approval/README.md
index f2be31022..b17eb1f6c 100644
--- a/examples/stripe-refund-approval/README.md
+++ b/examples/stripe-refund-approval/README.md
@@ -8,8 +8,8 @@ day. The Stripe key lives only in the Auths gateway, which also checks the
key and the payment with Stripe before each refund. An auditor checks every
refund afterwards, offline.
-**10 steps. The unattended run of all of them (step 10) took 8.1 s on an
-Apple-silicon laptop once the wheel and gateway were built.** Building the
+**10 steps. Step 10 runs the local double with a finite scripted operator reviewer;
+its report records actual timing and does not establish an independent human trial.** Building the
gateway the first time takes a few minutes.
| Who | Holds | Can do |
@@ -131,32 +131,47 @@ python mock_stripe.py --ledger "$WORK/ledger.jsonl" \
# it prints its port, for example 54321
```
-**5. Install the gateway with the Stripe key on stdin, and start it.**
+**5. Stage the credential, independently approve its reads, and start the gateway.**
+
+The separate development operator owns a read-only actor and an exact-effect
+reviewer. For this local double, `--simulation` selects a finite scripted
+reviewer; it is not evidence of human review. Omit that flag for interactive
+operation: `literal-review.txt` preserves the complete native renderer text and
+`signing-request.json` preserves its native request in the private operator
+directory; the operator reads both and answers through its owner-only consent FIFO.
+Neither mode gives these signing keys to the refund application.
```sh
+python native_operator.py --directory "$WORK/operator" setup \
+ --facts "$WORK/state/setup.json" --key-directory "$WORK/state/keys" \
+ --gateway "$(command -v auths-gateway)" --simulation
mkdir -m 700 "$WORK/gateway"
printf '%s\n' "$STRIPE_KEY" | auths-gateway install --state-dir "$WORK/gateway" \
--recipe recipe.json --profile-lock profile.lock.json \
--trusted-context "$WORK/state/trust/gateway.context.cbor" \
+ --effect-policy "$WORK/operator/effect-policy.json" \
--approve-digest \
- --provider stripe --alias refunds --account-label acct_1AuthsPlatform0 --credential-stdin \
- --loopback-provider 54321
+ --provider stripe --alias refunds --account-label acct_1AuthsPlatform0 --credential-stdin
unset STRIPE_KEY
+python native_operator.py --directory "$WORK/operator" publish \
+ --state "$WORK/gateway" --loopback-provider 54321
auths-gateway observer-init --state-dir "$WORK/gateway"
+python native_operator.py --directory "$WORK/operator" activate-configuration --state "$WORK/gateway"
auths-gateway serve --state-dir "$WORK/gateway" --app-socket "$WORK/app.sock" \
--loopback-provider 54321 &
+python native_operator.py --directory "$WORK/operator" serve \
+ --state "$WORK/gateway" --socket "$WORK/operator.sock" &
```
-If your state directory has to be long, `serve --admin-socket PATH` moves
-the admin socket into another owner-only directory (mode 0700, owned by you,
-not reached through a symbolic link); pass the same `--admin-socket` to
-`disable`, `revoke`, `rotate`, and `doctor`.
+`install` stages the credential and performs no provider read. A key without
+`rk_test_` is refused before staging. `publish` obtains a separately authorized
+native candidate-check proof, complete exact approval, and retained provider
+checks before publishing. `observer-init` changes the reviewed configuration;
+activation explicitly compares that exact revision. Keep the printed observer
+principal and setup's `trusted_context_sha256` for the auditor.
-`install` makes the guard's four reads before it stores the key, and stores
-nothing if one fails: a key without `rk_test_` is refused
-(`gateway.install.credential-guard`) before any read. `observer-init` prints
-`observer_anchor.principal`: the key the gateway signs its outcomes with.
-Give it and `trusted_context_sha256` to your auditor.
+If state paths are long, `serve --admin-socket PATH` moves the admin socket to
+another owner-only directory; pass the same option to operator commands.
**6. The agent asks for a refund; managers A and B approve; manager C never answers; the gateway submits.**
@@ -194,18 +209,19 @@ agent collects whatever responses exist, signs its exact refund, and submits:
```sh
python refunds.py submit --state "$WORK/state" --socket "$WORK/app.sock" \
- --operation-id refund-1 --responses "$WORK/refund-1"
+ --operation-id refund-1 --responses "$WORK/refund-1" \
+ --operator-socket "$WORK/operator.sock" --recover
```
-`submit` sends a refund to the gateway once two managers approved it, and
-decides nothing itself: only the gateway enforces the approval threshold,
-the ceiling, the per-window count and sum, and at most one run per operation
-ID. It prints the gateway's result with `"decided_by": "gateway"`, here
-`{"outcome": "observed-by-provider", "status": 200, "decided_by": "gateway",
-"bundle": "appended", ...}`: the gateway checked the key and the
-PaymentIntent (60.00 received, so at most 30.00 refundable here), sent the
-refund, and found its echo token in the refund it read back. The ledger has
-one write. The record lists the managers who approved (`"approved"`), those
+`submit` asks the separate operator for the native exact-effect approval after
+collecting the managers' signed requests, then forwards the unchanged proof,
+action and approval. The gateway alone enforces the threshold, ceiling, window
+budgets and replay rules. Its write result is `response-recorded` with status
+200. The explicit `--recover` requests a second, independently approved native
+read. The unchanged `recovery` report must show a provider-linked source at
+`observed-by-provider`; an HTTP response alone does not establish that claim.
+The audit observes the retained source after recovery. The ledger has one write.
+The record lists the managers who approved (`"approved"`), those
who have not answered (`"pending"`, here manager C), and any who declined
(`"declined"`). One decline beside two approvals does not stop
the refund. When too many managers declined for two approvals to remain
@@ -226,7 +242,7 @@ gateway code.
| `--connect-account acct_1AuthsOtherAcct`, which the grant does not list | gateway | `not-entered` `gateway.policy.scope-denied` | 0 |
| 50.00 when 45.00 of the day's 60.00 USD is left | gateway | `not-entered` `gateway.policy.sum-exhausted` | 0 |
| A third refund on the same UTC day | gateway | `not-entered` `gateway.policy.window-exhausted` | 0 |
-| `refund-1` requested again, with fresh approvals | gateway | `not-entered` `gateway.attempt.replay` | 0 |
+| `refund-1` requested again, with fresh approvals | gateway | `not-entered` `gateway.effect.operation-conflict` | 0 |
| `refund-1`'s proof sent for another refund | gateway | `denied` `action-body-mismatch` | 0 |
| The first lowered-threshold refund requested again at the installed threshold, after the window is full | gateway | `not-entered` `gateway.policy.window-exhausted`: a verification denial claims nothing, so this is not a replay | 0 |
| Managers B and C run `auths approve … --decline` | managers B and C | `not-submitted` (`approvers-declined`): manager A's approval alone cannot make a quorum, so nothing is sent | 0 |
@@ -266,8 +282,16 @@ its own grant (`python refunds.py grant --agent agent-checks --max-count 4`):
```sh
python refunds.py export --state "$WORK/state" --out audit-bundle.json
+# Trusted operator shell; gateway must still be running. $WORK is owner-private.
+auths-gateway audit-export --state-dir "$WORK/gateway" \
+ --bundle audit-bundle.json --out "$WORK/private-audit-bundle.json"
```
+The automated journeys capture this private snapshot before their deliberate
+old-backup restoration. Their later offline audit checks that retained signed
+snapshot. A separate export after restoration must contain only the operation
+actually retained or resubmitted; missing journal evidence is never recreated.
+
It holds one entry per operation ID: the first submission, or a later one
that the gateway recorded in place of a first one it did not. Resubmissions
of an operation ID are not in the bundle; the gateway's attempt store is the
@@ -284,10 +308,16 @@ lowered-threshold refund's entry is the signed one of its later retry. The bundl
response the agent collected (declines included), and the installed recipe
and trusted context.
+The application export contains redacted public outcomes. The operator command
+replaces these with the complete private signed audit facts and protected
+pre-entry observations from native retained history. The private bundle is a
+new 0600 file in a private directory. Do not return it or its detailed audit
+report to the application or publish it as public evidence.
+
**9. Audit offline.** With the gateway stopped and the network off:
```sh
-auths-gateway audit --bundle audit-bundle.json \
+auths-gateway audit --bundle "$WORK/private-audit-bundle.json" \
--trusted-context-sha256 --observer \
--allow-unverified-refusals
```
@@ -363,7 +393,7 @@ that check. It checks every result, including exactly two Stripe writes
Stripe call at all for the refusals before the lease, the `Idempotency-Key`
each write carried, `Stripe-Version` on every request, `Stripe-Account` on
every request about the refund and on none that tests the key, refund 1's
-read-back, and the audit's `http_status` of 200 and 400.
+separately authorized read-back, and the audit's `http_status` of 200 and 400.
It prints timings. It then restores the gateway's store from a backup taken
before the first refund, which keeps the shared connection record but no
claim, and resubmits refund 1's approved proof. With the claim gone the
@@ -389,7 +419,9 @@ The same journey runs from the npm package, `@auths-dev/sdk`, against the
same recipe, gateway, and Stripe double. `typescript/refunds.ts` and
`typescript/journey.ts` are the TypeScript `refunds.py` and `journey.py`, and
`generated.ts` is the command contract the package's `auths generate`
-wrote from the same `profile.toml`. You need Node 20.6+ and Rust. From
+wrote from the same `profile.toml`. The application needs Node 20.6+ and Rust. Its separate operator helper also
+uses the installed Python SDK from step 1; this is not TypeScript read-proof
+authoring evidence. From
`typescript/`:
```sh
@@ -401,11 +433,10 @@ npm --prefix ../../../bindings/typescript run build
npm pack ../../../bindings/typescript && mv auths-dev-sdk-*.tgz auths-dev-sdk.tgz
npm install && npm run build
cargo install --locked --path ../../../product/runtime/auths-gateway --features loopback-provider
-node build/journey.js --gateway "$(command -v auths-gateway)"
+node build/journey.js --gateway "$(command -v auths-gateway)" --python "$(command -v python)"
```
-The last command took 7.0 s on an Apple-silicon laptop once the package and
-gateway were built. Each `python refunds.py …` step above is
+The command runs the finite scripted reviewer only against the local double. Each `python refunds.py …` step above is
`node build/refunds.js …` with the same flags (`--precheck` included) and
prints the same JSON records, each manager runs `npx auths approve …` with
the same flags, and `journey.js` takes the same test-mode options as
diff --git a/examples/stripe-refund-approval/journey.py b/examples/stripe-refund-approval/journey.py
index 53baaeca4..08bb4a4c4 100644
--- a/examples/stripe-refund-approval/journey.py
+++ b/examples/stripe-refund-approval/journey.py
@@ -235,6 +235,8 @@ def submit(self, operation: str, folder: Path) -> Dict[str, Any]:
operation,
"--responses",
str(folder),
+ "--operator-socket", str(self.work / "operator.sock"),
+ "--recover",
]
return json.loads(self.run(*command).stdout)
@@ -292,7 +294,7 @@ def expect(condition: bool, message: str) -> None:
raise SystemExit(f"journey check failed: {message}")
-SUBMIT_SCHEMA = "auths.gateway-submit/1"
+SUBMIT_SCHEMA = "auths.gateway-submit/2"
# Words only the gateway's submit result may carry.
GATEWAY_OUTCOMES = {
"denied",
@@ -415,11 +417,18 @@ def resubmit(journey: Journey, operation: str) -> Dict[str, Any]:
async def submit() -> Any:
gateway = GatewayClient(GatewayEndpoint(journey.socket.resolve()))
- return await gateway.submit(
- proof=unb64(entry["proof_b64"]), action=unb64(entry["action_b64"])
- )
+ from native_operator import request
+ reviewed = request(journey.work / "operator.sock", {"kind": "approve",
+ "proof": entry["proof_b64"], "action": entry["action_b64"]})
+ result = dataclasses.asdict(await gateway.submit(proof=unb64(entry["proof_b64"]),
+ action=unb64(entry["action_b64"]), approval=unb64(reviewed["approval"])))
+ if result.get("outcome") == "response-recorded" and result.get("status") == 200:
+ read = request(journey.work / "operator.sock", {"kind": "recover", "source": operation})
+ result["recovery"] = json.loads(await gateway.recover(operation,
+ proof=unb64(read["proof"]), action=unb64(read["action"]), approval=unb64(read["approval"])))
+ return result
- return dataclasses.asdict(asyncio.run(submit()))
+ return asyncio.run(submit())
def tamper(
@@ -547,6 +556,13 @@ def main() -> int:
"4",
),
)
+ from native_operator import Operator, provision
+ operator_directory = journey.work / "operator"
+ asyncio.run(provision(operator_directory, facts,
+ {name: journey.state / "keys" / f"{name}.seed" for name in ("root", "reader", *MANAGERS)},
+ args.gateway, simulation=not live))
+ effect_policy = operator_directory / "effect-policy.json"
+
# The double checks the bearer token by digest; the key itself stays
# only in the gateway's credential store.
mock_token_sha256 = hashlib.sha256(secret.encode()).hexdigest()
@@ -577,6 +593,8 @@ def install(state_dir: Path, key: str) -> subprocess.CompletedProcess[str]:
return journey.run(
args.gateway,
"install",
+ "--effect-policy",
+ str(effect_policy.resolve()),
"--state-dir",
str(state_dir),
"--recipe",
@@ -594,7 +612,6 @@ def install(state_dir: Path, key: str) -> subprocess.CompletedProcess[str]:
"--account-label",
args.platform_account,
"--credential-stdin",
- *loopback,
*extra_install,
stdin=key + "\n",
check=False,
@@ -624,6 +641,7 @@ def non_test_key() -> Dict[str, Any]:
def install_key() -> None:
installed = install(journey.gateway_state, secret)
expect(installed.returncode == 0, f"install failed: {installed.stderr.strip()}")
+ Operator(operator_directory, journey.gateway_state).publish(loopback[-1] if loopback else None)
onboarding_from = len(journey.provider_entries())
journey.step("gateway install with the key on stdin", install_key)
@@ -641,6 +659,19 @@ def install_key() -> None:
)["observer_anchor"]["principal"],
)
+ configuration = json.loads(journey.run(args.gateway, "effect-configuration-review",
+ "--state-dir", str(journey.gateway_state)).stdout)
+ journey.run(args.gateway, "effect-configuration-activate", "--state-dir", str(journey.gateway_state),
+ "--expected-revision", configuration["previous_revision_hex"],
+ "--expected-candidate", configuration["candidate_hex"])
+ # The rollback test restores this actual active configuration, not an
+ # unpublished installation that can never grant entry.
+ shutil.rmtree(attempts_backup)
+ shutil.copytree(journey.gateway_state / "attempts", attempts_backup)
+ reviewer = journey.background(PYTHON, "native_operator.py", "--directory", str(operator_directory),
+ "serve", "--state", str(journey.gateway_state), "--socket", str(journey.work / "operator.sock"))
+ expect(reviewer.stdout.readline().strip() == "ready", "operator reviewer did not start")
+
serve = [
args.gateway,
"serve",
@@ -739,6 +770,7 @@ def reused_approvals() -> Dict[str, Any]:
str(proof),
"--action",
str(action),
+ "--approval", str(operator_directory / "adversarial-substitution.approval"),
)
# The CLI prints the gateway's submit result unchanged.
return {"decided_by": "gateway", **json.loads(sent.stdout)}
@@ -825,6 +857,7 @@ def self_approval() -> Dict[str, Any]:
str(proof_path),
"--action",
str(action_path),
+ "--approval", str(operator_directory / "adversarial-substitution.approval"),
)
return {"decided_by": "gateway", "sdk": sdk, **json.loads(sent.stdout)}
@@ -840,7 +873,7 @@ def self_approval() -> Dict[str, Any]:
"refund-other-account": ("not-entered", "gateway.policy.scope-denied"),
"refund-over-sum": ("not-entered", "gateway.policy.sum-exhausted"),
"refund-5-window": ("not-entered", "gateway.policy.window-exhausted"),
- "refund-1-replay": ("not-entered", "gateway.attempt.replay"),
+ "refund-1-replay": ("not-entered", "gateway.effect.operation-conflict"),
"refund-8-reuse": ("denied", "action-body-mismatch"),
"refund-2-retry": ("not-entered", "gateway.policy.window-exhausted"),
}
@@ -1030,11 +1063,14 @@ def checked(operation: str, amount: int, control: Dict[str, Any], **extra: str)
"refund-1",
observed,
frames["refund-1"],
- "observed-by-provider",
+ "response-recorded",
status=200,
action_b64=requests["refund-1"]["action_b64"],
)
- expect(observed["evidence"]["channel"] == "read-back", f"refund-1: {observed}")
+ expect(observed["recovery"]["report"]["source"]["provider_linked"] is True
+ and observed["recovery"]["report"]["source"]["stage"] == "observed-by-provider"
+ and observed["recovery"]["report"]["acquisition"]["response_status"] == 200,
+ f"refund-1 requires separately authorized linked recovery: {observed}")
expect(observed["bundle"] == "appended", f"refund-1: {observed}")
expect(
observed["approved"] == ["manager-a", "manager-b"]
@@ -1157,6 +1193,26 @@ def checked(operation: str, amount: int, control: Dict[str, Any], **extra: str)
for case in ("refund-1-replay", "refund-8-reuse", "refund-2-retry", "refund-9-precheck"):
expect(results[case]["provider_requests"] == 0, f"{case} reached the provider")
+ # Preserve operator-private signed evidence before the deliberate rollback.
+ # The later offline audit evaluates this historical snapshot, not the restored store.
+ bundle_path = journey.work / "audit-bundle.json"
+ journey.step(
+ "export audit bundle",
+ lambda: journey.run(
+ PYTHON,
+ "refunds.py",
+ "export",
+ "--state",
+ str(journey.state),
+ "--out",
+ str(bundle_path),
+ ),
+ )
+ private_bundle = journey.work / "private-audit-bundle.json"
+ journey.run(journey.gateway, "audit-export", "--state-dir", str(journey.gateway_state),
+ "--bundle", str(bundle_path), "--out", str(private_bundle))
+ bundle_path = private_bundle
+
state_loss: Optional[Dict[str, Any]] = None
if not live:
for operation in before_lease:
@@ -1199,10 +1255,12 @@ def checked(operation: str, amount: int, control: Dict[str, Any], **extra: str)
expect(len(writes) == 2, f"expected exactly 2 provider writes, saw {len(writes)}")
expect(all(entry["well_formed"] for entry in writes), "provider saw a malformed refund")
expect([entry["amount"] for entry in writes] == [1_500, 4_000], f"provider writes {writes}")
- expect(
- writes[0]["echo"] == observed["evidence"]["echo"],
- f"refund-1 echo {writes[0]['echo']} != {observed['evidence']['echo']}",
- )
+ echo_record, echo_action = journey.work / "echo.json", journey.work / "echo.action"
+ echo_record.write_text(json.dumps({"echo": writes[0]["echo"]}))
+ echo_action.write_bytes(unb64(requests["refund-1"]["action_b64"]))
+ journey.run(args.gateway, "echo-verify", "--record", str(echo_record), "--pointer", "/echo",
+ "--operation-id", "refund-1", "--namespace", facts["operator_namespace"],
+ "--action", str(echo_action))
read_back = f"/v1/refunds/{writes[0]['refund']}"
expect(
any(entry["path"] == read_back and entry["status"] == 200 for entry in entries),
@@ -1235,7 +1293,8 @@ def lose_attempts_and_resubmit() -> Dict[str, Any]:
lose_attempts_and_resubmit,
)
expect(
- state_loss["outcome"] == "observed-by-provider" and state_loss["status"] == 200,
+ state_loss["outcome"] == "response-recorded" and state_loss["status"] == 200
+ and state_loss["recovery"]["report"]["source"]["provider_linked"] is True,
f"resubmitted refund-1: {state_loss}",
)
writes = [entry for entry in journey.provider_entries() if entry["kind"] == "write"]
@@ -1248,21 +1307,17 @@ def lose_attempts_and_resubmit() -> Dict[str, Any]:
)
refunds_created = {entry["refund"] for entry in writes if entry["refund"]}
expect(len(refunds_created) == 1, f"expected 1 refund, saw {sorted(refunds_created)}")
+ restored_bundle = journey.work / "restored-private-audit-bundle.json"
+ journey.run(journey.gateway, "audit-export", "--state-dir", str(journey.gateway_state),
+ "--bundle", str(bundle_path), "--out", str(restored_bundle))
+ restored = json.loads(restored_bundle.read_text())
+ retained = {entry["operation_id"] for entry in restored["entries"]
+ if entry.get("outcome_b64") is not None}
+ expect(retained == {"refund-1"}, "restored store manufactured lost audit evidence")
+ state_loss["retained_audit_operations"] = sorted(retained)
+ state_loss["historical_audit_export_preserved"] = True
+
- # README step 8: the audit bundle.
- bundle_path = journey.work / "audit-bundle.json"
- journey.step(
- "export audit bundle",
- lambda: journey.run(
- PYTHON,
- "refunds.py",
- "export",
- "--state",
- str(journey.state),
- "--out",
- str(bundle_path),
- ),
- )
journey.stop()
# README step 9: the offline audit, with the gateway stopped. The
@@ -1526,6 +1581,7 @@ def remove_and_flip(entry: Dict[str, Any]) -> None:
writes = [entry for entry in journey.provider_entries() if entry["kind"] == "write"]
summary = {
"journey": "stripe-refund-approval",
+ "effect_reviewer": "interactive operator consent; development custody" if live else "finite scripted simulation; no human-trial claim",
"provider": "stripe-test-mode" if live else "counting-mock",
"credential_store": extra_install[extra_install.index("--credential-store") + 1]
if "--credential-store" in extra_install
diff --git a/examples/stripe-refund-approval/native_operator.py b/examples/stripe-refund-approval/native_operator.py
new file mode 100644
index 000000000..ce0a52088
--- /dev/null
+++ b/examples/stripe-refund-approval/native_operator.py
@@ -0,0 +1,455 @@
+"""Development operator tooling for the refund examples and generated scaffold.
+
+The native gateway authors and reviews every effect. This helper transports
+SDK-authored read proofs and external signatures; it decides no gateway verdict.
+Simulation is an explicitly finite scripted reviewer. Interactive runs require
+consent to each complete native literal review before the exact signature.
+Root, read-actor, manager and reviewer keys belong in an operator-private directory.
+"""
+from __future__ import annotations
+
+import argparse
+import asyncio
+import base64
+import binascii
+from dataclasses import dataclass
+import json
+import os
+import re
+from pathlib import Path
+import secrets
+import socket
+import struct
+import subprocess
+import sys
+import time
+from typing import Optional
+
+from auths.adapters.custody import PublicControlEvidence
+from auths.authoring import (GrantEvidence, approval_requests, approve, collect_approvals,
+ open_approval_request, propose_mcp_approval, sign_approval_action)
+from auths.self_hosted import (ExactMcpTool, StringField, IntegerField, OptionalField,
+ Permission, Profile, author_root_grant, load_signer_file)
+
+READ_TOOLS = ("auths_gateway_candidate_check_v1", "auths_gateway_recovery_read_v1")
+MAX_FRAME = 8 * 1024 * 1024
+MAX_REVIEWS = 128
+MAX_PROOF_BYTES = 4 * 1024 * 1024
+MAX_ACTION_BYTES = 64 * 1024
+PEER_TIMEOUT = 5
+NATIVE_COMMAND_TIMEOUT = 60
+TOKEN = re.compile(r"[A-Za-z0-9][A-Za-z0-9._-]{0,127}")
+BASE64URL = re.compile(r"[A-Za-z0-9_-]+")
+
+
+class OperatorRefusal(ValueError):
+ """A closed local refusal code, safe to return without peer data or paths."""
+
+
+def strict_object(pairs):
+ result = {}
+ for key, value in pairs:
+ if key in result:
+ raise OperatorRefusal("operator.duplicate-field")
+ result[key] = value
+ return result
+
+
+def reject_constant(_):
+ raise OperatorRefusal("operator.invalid-json")
+
+
+def validate_request(value):
+ if not isinstance(value, dict):
+ raise OperatorRefusal("operator.invalid-request")
+ if value.get("kind") == "approve" and set(value) == {"kind", "proof", "action"}:
+ unb64(value["proof"], MAX_PROOF_BYTES)
+ unb64(value["action"], MAX_ACTION_BYTES)
+ return value
+ if value.get("kind") == "recover" and set(value) == {"kind", "source"}:
+ if not isinstance(value["source"], str) or TOKEN.fullmatch(value["source"]) is None:
+ raise OperatorRefusal("operator.invalid-source")
+ return value
+ raise OperatorRefusal("operator.invalid-request")
+
+
+def b64(value):
+ return base64.urlsafe_b64encode(value).rstrip(b"=").decode()
+
+
+def unb64(value, maximum=MAX_PROOF_BYTES):
+ if (not isinstance(value, str) or not 0 < len(value) <= (maximum * 4 + 2) // 3
+ or BASE64URL.fullmatch(value) is None):
+ raise OperatorRefusal("operator.invalid-base64url")
+ try:
+ decoded = base64.b64decode(value + "=" * (-len(value) % 4), altchars=b"-_", validate=True)
+ except (binascii.Error, ValueError):
+ raise OperatorRefusal("operator.invalid-base64url") from None
+ if not 0 < len(decoded) <= maximum or b64(decoded) != value:
+ raise OperatorRefusal("operator.invalid-base64url")
+ return decoded
+
+
+def write(path, value):
+ path = Path(path)
+ path.parent.mkdir(mode=0o700, parents=True, exist_ok=True)
+ with os.fdopen(os.open(path, os.O_WRONLY | os.O_CREAT | os.O_EXCL, 0o600), "wb") as out:
+ out.write(value)
+
+
+def run(binary, *arguments, check=True):
+ try:
+ result = subprocess.run([str(binary), *map(str, arguments)], capture_output=True,
+ timeout=NATIVE_COMMAND_TIMEOUT)
+ except subprocess.TimeoutExpired:
+ raise OperatorRefusal("operator.native-command-timeout") from None
+ except OSError:
+ raise OperatorRefusal("operator.native-command-unavailable") from None
+ if check and result.returncode:
+ raise OperatorRefusal("operator.native-command-refused")
+ return result
+
+
+@dataclass(frozen=True)
+class CandidateCheck:
+ operation_id: str
+ operator_namespace: str
+ recipe_digest: str
+ candidate_connection: str
+ expected_previous: Optional[str]
+ read_contract: str
+ maximum_response_bytes: int
+
+
+@dataclass(frozen=True)
+class RecoveryRead:
+ operation_id: str
+ operator_namespace: str
+ recipe_digest: str
+ source_effect: str
+ source_operation: str
+ read_contract: str
+ maximum_response_bytes: int
+
+
+def read_contract(call):
+ if call["name"] not in READ_TOOLS:
+ raise OperatorRefusal("operator.unlisted-read-tool")
+ command_type = CandidateCheck if call["name"] == READ_TOOLS[0] else RecoveryRead
+ fields = {key: StringField(min_length=1, max_length=128)
+ for key in command_type.__annotations__}
+ fields["maximum_response_bytes"] = IntegerField(1, 1048576)
+ if command_type is CandidateCheck:
+ fields["expected_previous"] = OptionalField(StringField(min_length=64, max_length=64))
+ contract = ExactMcpTool(service=call["service"], name=call["name"],
+ command_type=command_type, fields=fields)
+ return contract, contract.validate_arguments(call["arguments"])
+
+
+def signer_file(directory, name, seed):
+ path = directory / (name + ".json")
+ write(path, json.dumps({"schema": "auths.approval-signer/1", "custody": "development-ed25519",
+ "seed_file": str(Path(seed).resolve())}).encode())
+ return path
+
+
+async def provision(directory, facts, key_paths, gateway, simulation):
+ """Called by operator setup before install; grants only the two native read tools."""
+ directory.mkdir(mode=0o700, parents=True)
+ signers = {name: signer_file(directory, name, seed) for name, seed in key_paths.items()}
+ root = load_signer_file(signers["root"]).signer
+ reader = load_signer_file(signers["reader"]).signer
+ now = int(time.time())
+ try:
+ grant = await author_root_grant(signer=root, subject=reader.descriptor.principal,
+ profile=Profile("auths.mcp", 2),
+ permissions=[Permission("tools/call", facts["audience"] + "/tools/" + name)
+ for name in READ_TOOLS], audiences=[facts["audience"]],
+ not_before=facts["not_before"], expires_at=facts["expires_at"],
+ remaining_depth=0, assurance_floor="raw-key-baseline", requested_at=now)
+ finally:
+ await root.aclose()
+ await reader.aclose()
+ write(directory / "read-grant.json", json.dumps({"bytes": b64(grant.signed_grant),
+ "evidence": [{"type": item.evidence_type, "media_type": item.media_type,
+ "bytes": b64(item.bytes)} for item in grant.evidence]}).encode())
+ seed = secrets.token_bytes(32)
+ write(directory / "reviewer.seed", seed)
+ from auths import _native
+ reviewer = _native.DevelopmentEd25519Key.from_seed(seed)
+ policy = run(gateway, "effect-policy", "--partition", "refund-development",
+ "--epoch", "1", "--approver-public-key", bytes(reviewer.public_key).hex()).stdout
+ write(directory / "effect-policy.json", policy)
+ write(directory / "configuration.json", json.dumps({"gateway": str(Path(gateway).resolve()),
+ "facts": facts, "signers": {name: str(path.resolve()) for name, path in signers.items()},
+ "simulation": simulation, "maximum_reviews": MAX_REVIEWS}).encode())
+
+
+def consent(directory, name):
+ """Exact bounded response through a private operator FIFO, never application input."""
+ fifo = directory / "consent"
+ os.mkfifo(fifo, 0o600)
+ descriptor = os.open(fifo, os.O_RDONLY | os.O_NONBLOCK | os.O_NOFOLLOW)
+ try:
+ print(f"Read the full text {directory / 'literal-review.txt'} and canonical signing request "
+ f"{directory / 'signing-request.json'} in your trusted operator shell, "
+ f"then write 'approve {name}' followed by a newline to {fifo}", file=sys.stderr, flush=True)
+ deadline, response = time.monotonic() + 600, b""
+ while time.monotonic() < deadline and len(response) < 256:
+ try:
+ response += os.read(descriptor, 256 - len(response))
+ except BlockingIOError:
+ pass
+ if b"\n" in response:
+ if response != ("approve " + name + "\n").encode():
+ raise OperatorRefusal("operator.review-declined")
+ return
+ time.sleep(0.05)
+ raise OperatorRefusal("operator.review-expired")
+ finally:
+ os.close(descriptor)
+ fifo.unlink(missing_ok=True)
+
+
+class Operator:
+ def __init__(self, directory, state):
+ self.directory, self.state = Path(directory), Path(state)
+ self.config = json.loads((self.directory / "configuration.json").read_text())
+ self.facts = self.config["facts"]
+ self.binary = self.config["gateway"]
+ from auths import _native
+ self.reviewer = _native.DevelopmentEd25519Key.from_seed((self.directory / "reviewer.seed").read_bytes())
+ self.sequence = len(list(self.directory.glob("review-*")))
+
+ async def read_packet(self, call):
+ contract, command = read_contract(call)
+ grant = json.loads((self.directory / "read-grant.json").read_text())
+ names = self.facts.get("approvers", self.facts.get("managers"))
+ reader = load_signer_file(self.config["signers"]["reader"]).signer
+ managers = [load_signer_file(self.config["signers"][name]).signer for name in names]
+ try:
+ proposal = propose_mcp_approval(contract=contract, command=command,
+ required=self.facts["approvals_required"],
+ approvers=[self.facts["principals"][name] for name in names],
+ actor=reader.descriptor.principal, actor_grant=unb64(grant["bytes"]),
+ challenge=bytes.fromhex(self.facts["challenge_hex"]),
+ evaluation_time=int(time.time()), validity_seconds=1800)
+ responses = []
+ for signer in managers[:self.facts["approvals_required"]]:
+ request = next(item for item in approval_requests(proposal)
+ if item.approver == signer.descriptor.principal)
+ responses.append((await approve(open_approval_request(request.text,
+ now=int(time.time())), signer)).text)
+ signed = await sign_approval_action(proposal, reader, grants=[GrantEvidence(
+ unb64(grant["bytes"]), tuple(PublicControlEvidence(item["type"],
+ item["media_type"], unb64(item["bytes"])) for item in grant["evidence"]))])
+ return collect_approvals(proposal, responses).assemble(signed), proposal.action
+ finally:
+ await reader.aclose()
+ for signer in managers:
+ await signer.aclose()
+
+ def exact_approval(self, proof, action, *mode):
+ self.sequence += 1
+ if self.sequence > self.config["maximum_reviews"]:
+ raise OperatorRefusal("operator.review-budget-exhausted")
+ name = f"review-{self.sequence}"
+ directory = self.directory / name
+ directory.mkdir(mode=0o700)
+ write(directory / "proof", proof)
+ write(directory / "action", action)
+ now = int(time.time())
+ result = run(self.binary, "effect-review", "--state-dir", self.state,
+ "--proof", directory / "proof", "--action", directory / "action",
+ "--not-before", now - 1, "--expires-at", now + 1800,
+ "--request-out", directory / "request", *mode, check=False)
+ if result.returncode:
+ return None
+ review = json.loads(result.stdout)
+ if (review.get("schema") != "auths.private-effect-signing/2"
+ or review.get("renderer") != "auths.literal-complete-review/1"
+ or not isinstance(review.get("literal_review"), str) or not review["literal_review"]):
+ raise OperatorRefusal("operator.complete-native-review-required")
+ write(directory / "literal-review.json", result.stdout)
+ write(directory / "literal-review.txt", review["literal_review"].encode("utf-8"))
+ write(directory / "signing-request.json", result.stdout)
+ if not self.config["simulation"]:
+ consent(directory, name)
+ signature = self.reviewer.sign(bytes.fromhex(review["signing_message_hex"]))
+ approval = run(self.binary, "effect-approval", "--request", directory / "request",
+ "--signature-hex", bytes(signature).hex()).stdout
+ write(directory / "approval", approval)
+ return directory
+
+ def publish(self, loopback=None):
+ candidate = self.state / "candidate-installation.json"
+ call = self.directory / "candidate-call.json"
+ run(self.binary, "effect-candidate-call", "--state-dir", self.state,
+ "--candidate", candidate, "--operation", "install-check", "--call-out", call)
+ proof, action = asyncio.run(self.read_packet(json.loads(call.read_text())))
+ approved = self.exact_approval(proof, action, "--candidate", candidate)
+ if approved is None:
+ raise OperatorRefusal("operator.candidate-review-refused")
+ arguments = ["--state-dir", self.state, "--candidate", candidate,
+ "--proof", approved / "proof", "--action", approved / "action",
+ "--approval", approved / "approval"]
+ run(self.binary, "candidate-acquire", *arguments,
+ *(["--loopback-provider", loopback] if loopback else []))
+ run(self.binary, "candidate-publish", *arguments)
+ write(self.directory / "adversarial-substitution.approval", (approved / "approval").read_bytes())
+
+ def activate_configuration(self):
+ review = run(self.binary, "effect-configuration-review", "--state-dir", self.state).stdout
+ write(self.directory / "configuration-review.json", review)
+ exact = json.loads(review)
+ run(self.binary, "effect-configuration-activate", "--state-dir", self.state,
+ "--expected-revision", exact["previous_revision_hex"], "--expected-candidate", exact["candidate_hex"])
+
+ def request(self, request):
+ request = validate_request(request)
+ if request.get("kind") == "approve" and set(request) == {"kind", "proof", "action"}:
+ directory = self.exact_approval(unb64(request["proof"]), unb64(request["action"]))
+ if directory is not None:
+ return {"approval": b64((directory / "approval").read_bytes()),
+ "literal_review": json.loads((directory / "literal-review.json").read_text())["literal_review"]}
+ if self.config["simulation"]:
+ # Deliberately mismatched, genuine candidate approval for hostile
+ # proof tests. Never manufactures valid authority for their effect.
+ return {"approval": b64((self.directory / "adversarial-substitution.approval").read_bytes()),
+ "adversarial_substitution": True}
+ raise OperatorRefusal("operator.mutation-review-refused")
+ if request.get("kind") == "recover" and set(request) == {"kind", "source"}:
+ source = request["source"]
+ call = json.loads(run(self.binary, "effect-recovery-call", "--state-dir", self.state,
+ "--source-operation", source, "--recovery-operation", f"read-{self.sequence + 1}").stdout)
+ proof, action = asyncio.run(self.read_packet(call))
+ directory = self.exact_approval(proof, action, "--recovery-of", source)
+ if directory is None:
+ raise OperatorRefusal("operator.recovery-review-refused")
+ return {"proof": b64(proof), "action": b64(action),
+ "approval": b64((directory / "approval").read_bytes())}
+ raise OperatorRefusal("operator.invalid-request")
+
+
+def exact(connection, count, deadline):
+ value = b""
+ while len(value) < count:
+ remaining = deadline - time.monotonic()
+ if remaining <= 0:
+ raise OperatorRefusal("operator.peer-timeout")
+ connection.settimeout(remaining)
+ chunk = connection.recv(count - len(value))
+ if not chunk:
+ raise OperatorRefusal("operator.truncated-frame")
+ value += chunk
+ return value
+
+
+def receive(connection, timeout=PEER_TIMEOUT):
+ deadline = time.monotonic() + timeout
+ length = struct.unpack(">I", exact(connection, 4, deadline))[0]
+ if not 0 < length <= MAX_FRAME:
+ raise OperatorRefusal("operator.frame-bound")
+ try:
+ value = json.loads(exact(connection, length, deadline).decode("utf-8"),
+ object_pairs_hook=strict_object, parse_constant=reject_constant)
+ except (UnicodeError, json.JSONDecodeError, RecursionError):
+ raise OperatorRefusal("operator.invalid-json") from None
+ if not isinstance(value, dict):
+ raise OperatorRefusal("operator.invalid-request")
+ return value
+
+
+def send(connection, value):
+ encoded = json.dumps(value, separators=(",", ":")).encode()
+ if not 0 < len(encoded) <= MAX_FRAME:
+ raise OperatorRefusal("operator.frame-bound")
+ connection.sendall(struct.pack(">I", len(encoded)) + encoded)
+
+
+def request(socket_path, value):
+ with socket.socket(socket.AF_UNIX, socket.SOCK_STREAM) as connection:
+ connection.settimeout(620)
+ connection.connect(str(socket_path))
+ send(connection, value)
+ result = receive(connection, timeout=620)
+ if "error" in result:
+ raise RuntimeError(result["error"])
+ return result
+
+
+def serve_peer(operator, connection):
+ """One bounded exchange; malformed or disconnected peers never kill the listener."""
+ accepted = False
+ with connection:
+ connection.settimeout(PEER_TIMEOUT)
+ try:
+ request = validate_request(receive(connection))
+ accepted = True
+ result = operator.request(request)
+ except OperatorRefusal as error:
+ result = {"error": str(error)}
+ except (socket.timeout, TimeoutError):
+ result = {"error": "operator.peer-timeout"}
+ except OSError:
+ result = {"error": "operator.peer-unavailable"}
+ except (ValueError, TypeError, KeyError, AttributeError, RecursionError):
+ # Native response/signing adapters can also refuse. No diagnostic
+ # text, submitted content or private filesystem path crosses back.
+ result = {"error": "operator.request-refused"}
+ try:
+ send(connection, result)
+ except (OSError, OperatorRefusal):
+ pass # A disconnected peer has no new authority or retry entitlement.
+ return accepted
+
+
+def serve(operator, socket_path):
+ with socket.socket(socket.AF_UNIX, socket.SOCK_STREAM) as server:
+ old = os.umask(0o177)
+ try:
+ server.bind(str(socket_path))
+ finally:
+ os.umask(old)
+ server.listen(8)
+ print("ready", flush=True)
+ accepted = 0
+ while accepted < MAX_REVIEWS:
+ connection, _ = server.accept()
+ # Malformed peers do not spend the finite valid-request allowance.
+ accepted += int(serve_peer(operator, connection))
+
+
+def main():
+ parser = argparse.ArgumentParser()
+ parser.add_argument("--directory", type=Path, required=True)
+ commands = parser.add_subparsers(dest="command", required=True)
+ setup = commands.add_parser("setup")
+ setup.add_argument("--facts", type=Path, required=True)
+ setup.add_argument("--key-directory", type=Path, required=True)
+ setup.add_argument("--gateway", required=True)
+ setup.add_argument("--simulation", action="store_true")
+ publish = commands.add_parser("publish")
+ publish.add_argument("--state", type=Path, required=True)
+ publish.add_argument("--loopback-provider")
+ activate = commands.add_parser("activate-configuration")
+ activate.add_argument("--state", type=Path, required=True)
+ listener = commands.add_parser("serve")
+ listener.add_argument("--state", type=Path, required=True)
+ listener.add_argument("--socket", type=Path, required=True)
+ args = parser.parse_args()
+ if args.command == "setup":
+ facts = json.loads(args.facts.read_text())
+ names = ["root", "reader", *facts.get("approvers", facts.get("managers"))]
+ asyncio.run(provision(args.directory, facts,
+ {name: args.key_directory / (name + ".seed") for name in names}, args.gateway, args.simulation))
+ elif args.command == "publish":
+ Operator(args.directory, args.state).publish(args.loopback_provider)
+ elif args.command == "activate-configuration":
+ Operator(args.directory, args.state).activate_configuration()
+ else:
+ serve(Operator(args.directory, args.state), args.socket)
+
+
+if __name__ == "__main__":
+ main()
diff --git a/examples/stripe-refund-approval/recipe.json b/examples/stripe-refund-approval/recipe.json
index 506260bc3..36e3f65ff 100644
--- a/examples/stripe-refund-approval/recipe.json
+++ b/examples/stripe-refund-approval/recipe.json
@@ -1,5 +1,5 @@
{
- "schema": "auths.gateway-recipe-source/2",
+ "schema": "auths.gateway-recipe-source/3",
"profile_schema_digest": "db3a47820a6f5c3ce3883068bea16542721e049cab8453a2af579c77a46d5c5e",
"service": "stripe-refunds",
"tool": "create_refund_v1",
@@ -7,64 +7,195 @@
"credential": {
"kind": "bearer",
"guard": {
- "prefixes": ["rk_test_"],
+ "prefixes": [
+ "rk_test_"
+ ],
"probe": {
- "path": [{"kind": "fixed", "value": "v1"}, {"kind": "fixed", "value": "balance"}],
+ "path": [
+ {
+ "kind": "fixed",
+ "value": "v1"
+ },
+ {
+ "kind": "fixed",
+ "value": "balance"
+ }
+ ],
"json_pointer": "/livemode",
"equals": false,
"maximum_response_bytes": 16384
},
"account": {
- "path": [{"kind": "fixed", "value": "v1"}, {"kind": "fixed", "value": "account"}],
+ "path": [
+ {
+ "kind": "fixed",
+ "value": "v1"
+ },
+ {
+ "kind": "fixed",
+ "value": "account"
+ }
+ ],
"json_pointer": "/id",
"maximum_response_bytes": 65536
},
"denied_reads": [
- {"method": "GET", "path": [{"kind": "fixed", "value": "v1"}, {"kind": "fixed", "value": "customers"}], "refused_status": [403]},
- {"method": "GET", "path": [{"kind": "fixed", "value": "v1"}, {"kind": "fixed", "value": "payouts"}], "refused_status": [403]}
+ {
+ "method": "GET",
+ "path": [
+ {
+ "kind": "fixed",
+ "value": "v1"
+ },
+ {
+ "kind": "fixed",
+ "value": "customers"
+ }
+ ],
+ "refused_status": [
+ 403
+ ]
+ },
+ {
+ "method": "GET",
+ "path": [
+ {
+ "kind": "fixed",
+ "value": "v1"
+ },
+ {
+ "kind": "fixed",
+ "value": "payouts"
+ }
+ ],
+ "refused_status": [
+ 403
+ ]
+ }
]
}
},
"origin": "https://api.stripe.com",
- "provider_headers": {"Stripe-Version": "2025-03-31.basil"},
- "account_scope": {"header": "Stripe-Account", "field": "connect_account"},
- "bounds": {"sum": {"argument": "amount", "partition": "currency"}},
+ "provider_headers": {
+ "Stripe-Version": "2025-03-31.basil"
+ },
+ "account_scope": {
+ "header": "Stripe-Account",
+ "field": "connect_account"
+ },
+ "bounds": {
+ "sum": {
+ "argument": "amount",
+ "partition": "currency"
+ }
+ },
"relative_ceiling": {
"argument": "amount",
"basis_points": 5000,
"path": [
- {"kind": "fixed", "value": "v1"},
- {"kind": "fixed", "value": "payment_intents"},
- {"kind": "field", "name": "payment_intent"}
+ {
+ "kind": "fixed",
+ "value": "v1"
+ },
+ {
+ "kind": "fixed",
+ "value": "payment_intents"
+ },
+ {
+ "kind": "field",
+ "name": "payment_intent"
+ }
],
"json_pointer": "/amount_received",
- "bind": [{"pointer": "/currency", "field": "currency"}],
+ "bind": [
+ {
+ "pointer": "/currency",
+ "field": "currency"
+ }
+ ],
"maximum_response_bytes": 65536
},
"write": {
"method": "POST",
- "path": [{"kind": "fixed", "value": "v1"}, {"kind": "fixed", "value": "refunds"}],
+ "path": [
+ {
+ "kind": "fixed",
+ "value": "v1"
+ },
+ {
+ "kind": "fixed",
+ "value": "refunds"
+ }
+ ],
"body": {
"kind": "form",
"fields": {
- "payment_intent": {"kind": "field", "name": "payment_intent"},
- "amount": {"kind": "field", "name": "amount"}
+ "payment_intent": {
+ "kind": "field",
+ "name": "payment_intent"
+ },
+ "amount": {
+ "kind": "field",
+ "name": "amount"
+ }
}
},
- "idempotency": {"kind": "derived-header", "retention_seconds": 86400}
+ "idempotency": {
+ "kind": "derived-header",
+ "retention_seconds": 86400
+ }
},
"observation": {
"path": [
- {"kind": "fixed", "value": "v1"},
- {"kind": "fixed", "value": "refunds"},
- {"kind": "response-field", "pointer": "/id", "max_bytes": 255}
+ {
+ "kind": "fixed",
+ "value": "v1"
+ },
+ {
+ "kind": "fixed",
+ "value": "refunds"
+ },
+ {
+ "kind": "response-field",
+ "pointer": "/id",
+ "max_bytes": 255
+ }
],
"json_pointer": "/amount",
"expected_field": "amount",
"maximum_response_bytes": 16384
},
"echo": {
- "write": {"kind": "form-field", "name": "metadata[auths_echo]"},
+ "write": {
+ "kind": "form-field",
+ "name": "metadata[auths_echo]"
+ },
"observe": "/metadata/auths_echo"
+ },
+ "review": {
+ "schema": "auths.effect-review-contract/1",
+ "consequences": [
+ "Issue one refund against the identified payment intent for the displayed amount. This is not a recurring charge.",
+ "The provider determines refund settlement timing. A response or receipt does not itself establish settlement."
+ ],
+ "fields": {
+ "amount": {
+ "kind": "integer",
+ "label": "Amount",
+ "unit": "minor units of the declared currency"
+ },
+ "connect_account": {
+ "kind": "identifier",
+ "label": "Connect account"
+ },
+ "currency": {
+ "kind": "identifier",
+ "label": "Currency"
+ },
+ "payment_intent": {
+ "kind": "identifier",
+ "label": "Payment intent"
+ }
+ }
}
}
diff --git a/examples/stripe-refund-approval/refunds.py b/examples/stripe-refund-approval/refunds.py
index 6c185f0ee..c212f37ac 100644
--- a/examples/stripe-refund-approval/refunds.py
+++ b/examples/stripe-refund-approval/refunds.py
@@ -83,7 +83,7 @@
# The gateway's trusted context requires approvals from this many of the
# three managers, any of them.
APPROVALS_REQUIRED = 2
-ROLES = ("root", "agent") + MANAGERS
+ROLES = ("root", "agent", "reader") + MANAGERS
ASSURANCE = "raw-key-baseline"
DAY = 86_400
# The test connected account the grant's scope lists; the gateway sends it as
@@ -304,7 +304,8 @@ def anchor(name: str, depth: int) -> Any:
_native.Principal(principals[name]),
["raw-key-v1"],
[("auths.mcp", 2)],
- [permission],
+ [permission, *[("tools/call", f"{audience}/tools/{name}") for name in
+ ("auths_gateway_candidate_check_v1", "auths_gateway_recovery_read_v1")]],
[audience],
[audience],
not_before,
@@ -610,9 +611,25 @@ async def _submit(args: argparse.Namespace) -> Dict[str, Any]:
)
proof = collection.assemble(action)
gateway = GatewayClient(GatewayEndpoint(args.socket.resolve()))
- result = await gateway.submit(proof=proof, action=proposal.action)
+ if args.operator_socket:
+ from native_operator import request, unb64
+ reviewed = request(args.operator_socket, {"kind": "approve", "proof": _b64(proof),
+ "action": _b64(proposal.action)})
+ approval = unb64(reviewed["approval"])
+ record["effect_review"] = {key: value for key, value in reviewed.items() if key != "approval"}
+ else:
+ with args.effect_approval.open("rb") as approval_file:
+ approval = approval_file.read(4097)
+ result = await gateway.submit(proof=proof, action=proposal.action, approval=approval)
record.update(dataclasses.asdict(result))
record["decided_by"] = "gateway"
+ if args.recover and record.get("outcome") == "response-recorded" and record.get("status") == 200:
+ if not args.operator_socket:
+ raise ValueError("--recover requires a separately authorizing --operator-socket")
+ from native_operator import request, unb64
+ read = request(args.operator_socket, {"kind": "recover", "source": args.operation_id})
+ record["recovery"] = json.loads(await gateway.recover(args.operation_id,
+ proof=unb64(read["proof"]), action=unb64(read["action"]), approval=unb64(read["approval"])))
observation = await gateway.observe_outcome(args.operation_id)
outcome = observation.observation if isinstance(observation, GatewaySignedObservation) else None
entry = {
@@ -634,7 +651,7 @@ def export(args: argparse.Namespace) -> None:
json.loads(line) for line in approvals.read_text().splitlines() if line
] if approvals.exists() else []
bundle = {
- "schema": "auths.gateway-audit-bundle/2",
+ "schema": "auths.gateway-audit-bundle/3",
"recipe_b64": _b64(RECIPE.read_bytes()),
"profile_lock_b64": _b64(PROFILE_LOCK.read_bytes()),
"trusted_context_b64": _b64((state / "trust" / "gateway.context.cbor").read_bytes()),
@@ -693,6 +710,10 @@ def main(argv: Optional[List[str]] = None) -> int:
submit = commands.add_parser("submit", help="collect the responses and submit through the gateway")
submit.add_argument("--state", type=Path, required=True)
submit.add_argument("--socket", type=Path, required=True)
+ approval = submit.add_mutually_exclusive_group(required=True)
+ approval.add_argument("--effect-approval", type=Path, help="native exact-effect approval")
+ approval.add_argument("--operator-socket", type=Path, help="separate development operator reviewer")
+ submit.add_argument("--recover", action="store_true", help="request a separate authorized recovery read after a successful response")
submit.add_argument("--operation-id", required=True)
submit.add_argument("--responses", type=Path, required=True, help="directory of *.response files")
bundle = commands.add_parser("export", help="write the audit bundle")
diff --git a/examples/stripe-refund-approval/typescript/journey.ts b/examples/stripe-refund-approval/typescript/journey.ts
index 0b5e980b0..79b9b578b 100644
--- a/examples/stripe-refund-approval/typescript/journey.ts
+++ b/examples/stripe-refund-approval/typescript/journey.ts
@@ -52,7 +52,7 @@ import {
writeFileSync,
} from "node:fs";
import { platform } from "node:os";
-import { join } from "node:path";
+import { join, resolve } from "node:path";
import { createInterface } from "node:readline";
import { fileURLToPath } from "node:url";
import { parseArgs } from "node:util";
@@ -61,7 +61,7 @@ import { setTimeout as sleep } from "node:timers/promises";
import { authorMcpProof, compileTrustedContext, proposeMcpApproval } from "@auths-dev/sdk/self-hosted";
import { CONTRACT, type CreateRefund } from "./generated.js";
-import { ASSURANCE, EXAMPLE, anchor, developmentSigner, roleKey, type SetupFacts } from "./refunds.js";
+import { ASSURANCE, EXAMPLE, anchor, developmentSigner, roleKey, requestOperator, type SetupFacts } from "./refunds.js";
const REFUNDS = fileURLToPath(new URL("./refunds.js", import.meta.url));
// The double's accounts and its already-refunded PaymentIntent.
@@ -203,7 +203,7 @@ class Journey {
submit(operation: string, folder: string): Json {
return JSON.parse(this.run(process.execPath, [
REFUNDS, "submit", "--state", this.state, "--socket", this.witnessSocket, "--operation-id", operation,
- "--responses", folder,
+ "--responses", folder, "--operator-socket", join(this.work, "operator.sock"), "--recover",
]).stdout) as Json;
}
@@ -255,7 +255,7 @@ function expect(condition: boolean, message: string): asserts condition {
if (!condition) throw new Error(`journey check failed: ${message}`);
}
-const SUBMIT_SCHEMA = "auths.gateway-submit/1";
+const SUBMIT_SCHEMA = "auths.gateway-submit/2";
// Words only the gateway's submit result may carry.
const GATEWAY_OUTCOMES = new Set([
"denied", "not-entered", "indeterminate", "unknown", "response-recorded", "observed", "observed-by-provider",
@@ -438,6 +438,12 @@ async function main(): Promise {
REFUNDS, "grant", "--state", journey.state, "--gateway", journey.gateway,
"--agent", CHECKS_AGENT, "--max-count", "4",
]));
+ const operatorDirectory = join(work, "operator");
+ const operatorScript = join(EXAMPLE, "native_operator.py");
+ journey.run(journey.python, [operatorScript, "--directory", operatorDirectory, "setup",
+ "--facts", join(journey.state, "setup.json"), "--key-directory", join(journey.state, "keys"),
+ "--gateway", journey.gateway, ...(live ? [] : ["--simulation"])]);
+ const effectPolicy = join(operatorDirectory, "effect-policy.json");
// The double checks the bearer token by digest; the key itself stays
// only in the gateway's credential store.
const mockTokenSha256 = createHash("sha256").update(secret).digest("hex");
@@ -462,11 +468,11 @@ async function main(): Promise {
const install = (stateDir: string, key: string): SpawnSyncReturns => {
mkdirSync(stateDir, { mode: 0o700 });
return journey.run(journey.gateway, [
- "install", "--state-dir", stateDir, "--recipe", "recipe.json",
+ "install", "--effect-policy", resolve(effectPolicy), "--state-dir", stateDir, "--recipe", "recipe.json",
"--profile-lock", "profile.lock.json",
"--trusted-context", join(journey.state, "trust", "gateway.context.cbor"),
"--approve-digest", facts.recipe_digest, "--provider", "stripe", "--alias", "refunds",
- "--account-label", platformAccount, "--credential-stdin", ...loopback,
+ "--account-label", platformAccount, "--credential-stdin",
], `${key}\n`, false);
};
@@ -491,6 +497,8 @@ async function main(): Promise {
await journey.step("gateway install with the key on stdin", () => {
const installed = install(journey.gatewayState, secret);
expect(installed.status === 0, `install failed: ${installed.stderr.trim()}`);
+ journey.run(journey.python, [operatorScript, "--directory", operatorDirectory, "publish",
+ "--state", journey.gatewayState, ...loopback]);
});
const onboarding = journey.providerEntries().slice(onboardingFrom);
// The store as a backup taken now would hold it: the shared connection
@@ -500,6 +508,20 @@ async function main(): Promise {
const observer = await journey.step("gateway observer key", () =>
((JSON.parse(journey.run(journey.gateway, ["observer-init", "--state-dir", journey.gatewayState]).stdout) as
{ observer_anchor: { principal: string } }).observer_anchor.principal));
+ const configuration = JSON.parse(journey.run(journey.gateway,
+ ["effect-configuration-review", "--state-dir", journey.gatewayState]).stdout) as
+ { previous_revision_hex: string; candidate_hex: string };
+ journey.run(journey.gateway, ["effect-configuration-activate", "--state-dir", journey.gatewayState,
+ "--expected-revision", configuration.previous_revision_hex, "--expected-candidate", configuration.candidate_hex]);
+ rmSync(attemptsBackup, { recursive: true });
+ copyTree(join(journey.gatewayState, "attempts"), attemptsBackup);
+ const reviewer = journey.background(journey.python, [operatorScript, "--directory", operatorDirectory,
+ "serve", "--state", journey.gatewayState, "--socket", join(work, "operator.sock")], "pipe");
+ const reviewerReady = await Promise.race([
+ new Promise((done) => createInterface({ input: reviewer.stdout! }).once("line", done)),
+ sleep(10_000, "", { ref: false }),
+ ]);
+ expect(reviewerReady === "ready", "operator reviewer did not start");
let running: ChildProcess | undefined;
const startGateway = async (): Promise => {
// A stopped gateway leaves its socket file behind; wait for a new one.
@@ -569,6 +591,7 @@ async function main(): Promise {
// The witness is its own process, so this blocking call cannot stall it.
const sent = journey.run(journey.gateway, [
"submit", "--app-socket", journey.witnessSocket, "--proof", proof, "--action", action,
+ "--approval", join(operatorDirectory, "adversarial-substitution.approval"),
]);
// The CLI prints the gateway's submit result unchanged.
return { decided_by: "gateway", ...(JSON.parse(sent.stdout) as Json) };
@@ -647,6 +670,7 @@ async function main(): Promise {
writeFileSync(actionPath, authored.action);
const sent = journey.run(journey.gateway, [
"submit", "--app-socket", journey.witnessSocket, "--proof", proofPath, "--action", actionPath,
+ "--approval", join(operatorDirectory, "adversarial-substitution.approval"),
]);
return { decided_by: "gateway", sdk, ...(JSON.parse(sent.stdout) as Json) };
};
@@ -672,7 +696,7 @@ async function main(): Promise {
"refund-other-account": ["not-entered", "gateway.policy.scope-denied"],
"refund-over-sum": ["not-entered", "gateway.policy.sum-exhausted"],
"refund-5-window": ["not-entered", "gateway.policy.window-exhausted"],
- "refund-1-replay": ["not-entered", "gateway.attempt.replay"],
+ "refund-1-replay": ["not-entered", "gateway.effect.operation-conflict"],
"refund-8-reuse": ["denied", "action-body-mismatch"],
"refund-2-retry": ["not-entered", "gateway.policy.window-exhausted"],
};
@@ -781,11 +805,12 @@ async function main(): Promise {
}
const same = (left: unknown, right: unknown): boolean => JSON.stringify(left) === JSON.stringify(right);
- const observed = results["refund-1"]! as Json & { evidence?: { channel: string; echo: string } };
+ const observed = results["refund-1"]! as Json & { recovery: { report: { source: { provider_linked: boolean; stage: string }; acquisition: { response_status: number } } } };
expectGatewayDecided("refund-1", observed, frames["refund-1"]!, {
- outcome: "observed-by-provider", status: 200, actionB64: requests["refund-1"]!.action_b64 as string,
+ outcome: "response-recorded", status: 200, actionB64: requests["refund-1"]!.action_b64 as string,
});
- expect(observed.evidence?.channel === "read-back" && observed.bundle === "appended",
+ expect(observed.recovery.report.source.provider_linked && observed.recovery.report.source.stage === "observed-by-provider" &&
+ observed.recovery.report.acquisition.response_status === 200 && observed.bundle === "appended",
`refund-1: ${JSON.stringify(observed)}`);
expect(same(observed.approved, ["manager-a", "manager-b"]) && same(observed.pending, ["manager-c"]) &&
!("declined" in observed) &&
@@ -862,6 +887,17 @@ async function main(): Promise {
expect(results[case_]!.provider_requests === 0, `${case_} reached the provider`);
}
+ // Preserve private signed evidence before deliberate rollback. The offline
+ // auditor evaluates this historical snapshot, not the restored store.
+ let bundlePath = join(journey.work, "audit-bundle.json");
+ await journey.step("export audit bundle", () => journey.run(process.execPath, [
+ REFUNDS, "export", "--state", journey.state, "--out", bundlePath,
+ ]));
+ const privateBundle = join(journey.work, "private-audit-bundle.json");
+ journey.run(journey.gateway, ["audit-export", "--state-dir", journey.gatewayState,
+ "--bundle", bundlePath, "--out", privateBundle]);
+ bundlePath = privateBundle;
+
let stateLoss: Json | null = null;
if (!live) {
for (const operation of Object.keys(beforeLease)) {
@@ -888,8 +924,11 @@ async function main(): Promise {
expect(writes.length === 2, `expected exactly 2 provider writes, saw ${writes.length}`);
expect(writes.every((entry) => entry.well_formed === true), "provider saw a malformed refund");
expect(same(writes.map((entry) => entry.amount), [1_500, 4_000]), `provider writes ${JSON.stringify(writes)}`);
- expect(writes[0]!.echo === observed.evidence?.echo,
- `refund-1 echo ${String(writes[0]!.echo)} != ${String(observed.evidence?.echo)}`);
+ const echoRecord = join(work, "echo.json"), echoAction = join(work, "echo.action");
+ writeFileSync(echoRecord, JSON.stringify({ echo: writes[0]!.echo }), { mode: 0o600 });
+ writeFileSync(echoAction, Buffer.from(requests["refund-1"]!.action_b64 as string, "base64url"), { mode: 0o600 });
+ journey.run(journey.gateway, ["echo-verify", "--record", echoRecord, "--pointer", "/echo",
+ "--operation-id", "refund-1", "--namespace", facts.operator_namespace, "--action", echoAction]);
const readBack = `/v1/refunds/${String(writes[0]!.refund)}`;
expect(entries.some((entry) => entry.path === readBack && entry.status === 200),
`refund-1 was not read back at ${readBack}`);
@@ -919,11 +958,25 @@ async function main(): Promise {
const action = join(journey.work, "resubmit.action");
writeFileSync(proof, new Uint8Array(Buffer.from(first.proof_b64 as string, "base64url")));
writeFileSync(action, new Uint8Array(Buffer.from(first.action_b64 as string, "base64url")));
- return JSON.parse(journey.run(journey.gateway, [
+ const reviewed = await requestOperator(join(work, "operator.sock"),
+ { kind: "approve", proof: first.proof_b64, action: first.action_b64 });
+ const approval = join(work, "resubmit.approval");
+ writeFileSync(approval, Buffer.from(String(reviewed.approval), "base64url"), { mode: 0o600 });
+ const result = JSON.parse(journey.run(journey.gateway, [
"submit", "--app-socket", journey.socket, "--proof", proof, "--action", action,
+ "--approval", approval,
]).stdout) as Json;
+ const read = await requestOperator(join(work, "operator.sock"), { kind: "recover", source: "refund-1" });
+ for (const field of ["proof", "action", "approval"] as const) {
+ writeFileSync(join(work, `read.${field}`), Buffer.from(String(read[field]), "base64url"), { mode: 0o600 });
+ }
+ result.recovery = JSON.parse(journey.run(journey.gateway, ["app-recover", "--app-socket", journey.socket,
+ "--source-operation", "refund-1", "--proof", join(work, "read.proof"),
+ "--action", join(work, "read.action"), "--approval", join(work, "read.approval")]).stdout) as unknown;
+ return result;
});
- expect(stateLoss.outcome === "observed-by-provider" && stateLoss.status === 200,
+ expect(stateLoss.outcome === "response-recorded" && stateLoss.status === 200 &&
+ (stateLoss.recovery as { report: { source: { provider_linked: boolean } } }).report.source.provider_linked,
`resubmitted refund-1: ${JSON.stringify(stateLoss)}`);
writes = journey.providerEntries().filter((entry) => entry.kind === "write");
expect(writes.length === 3, `expected 3 provider writes, saw ${writes.length}`);
@@ -932,13 +985,18 @@ async function main(): Promise {
`resubmitted refund-1 was not de-duplicated: ${JSON.stringify(writes[2])}`);
const created = new Set(writes.map((entry) => entry.refund).filter((refund) => refund));
expect(created.size === 1, `expected 1 refund, saw ${JSON.stringify([...created])}`);
+ const restoredBundle = join(journey.work, "restored-private-audit-bundle.json");
+ journey.run(journey.gateway, ["audit-export", "--state-dir", journey.gatewayState,
+ "--bundle", bundlePath, "--out", restoredBundle]);
+ const restored = JSON.parse(readFileSync(restoredBundle, "utf8")) as { entries: Json[] };
+ const retained = restored.entries.filter((entry) => entry.outcome_b64 != null)
+ .map((entry) => entry.operation_id as string).sort();
+ expect(same(retained, ["refund-1"]), "restored store manufactured lost audit evidence");
+ stateLoss.retained_audit_operations = retained;
+ stateLoss.historical_audit_export_preserved = true;
+
}
- // README step 8: the audit bundle.
- const bundlePath = join(journey.work, "audit-bundle.json");
- await journey.step("export audit bundle", () => journey.run(process.execPath, [
- REFUNDS, "export", "--state", journey.state, "--out", bundlePath,
- ]));
const providerRequests = journey.providerEntries();
await journey.stop();
@@ -1119,6 +1177,8 @@ async function main(): Promise {
const summary = {
journey: "stripe-refund-approval",
sdk: "@auths-dev/sdk (packed npm)",
+ operator_tooling: "separate Python SDK; native gateway authors and verifies read semantics",
+ effect_reviewer: live ? "interactive operator consent; development custody" : "finite scripted simulation; no human-trial claim",
provider: live ? "stripe-test-mode" : "counting-mock",
wall_seconds: Math.round(performance.now() - journey.started) / 1000,
steps: journey.steps,
diff --git a/examples/stripe-refund-approval/typescript/refunds.ts b/examples/stripe-refund-approval/typescript/refunds.ts
index 4525b4b18..9f119db33 100644
--- a/examples/stripe-refund-approval/typescript/refunds.ts
+++ b/examples/stripe-refund-approval/typescript/refunds.ts
@@ -39,9 +39,10 @@
import { execFileSync } from "node:child_process";
import { createHash, randomBytes } from "node:crypto";
import {
- appendFileSync, closeSync, existsSync, mkdirSync, openSync, readFileSync, readdirSync, renameSync, writeFileSync,
+ appendFileSync, closeSync, existsSync, mkdirSync, openSync, readFileSync, readSync, readdirSync, renameSync, writeFileSync,
} from "node:fs";
import { dirname, join, resolve } from "node:path";
+import { createConnection } from "node:net";
import { fileURLToPath } from "node:url";
import { parseArgs } from "node:util";
@@ -67,7 +68,7 @@ export const MANAGERS = ["manager-a", "manager-b", "manager-c"] as const;
// The gateway's trusted context requires approvals from this many of the
// three managers, any of them.
const APPROVALS_REQUIRED = 2;
-const ROLES = ["root", "agent", ...MANAGERS] as const;
+const ROLES = ["root", "agent", "reader", ...MANAGERS] as const;
const DAY = 86_400n;
// The test connected account the grant's scope lists; the gateway sends it as
// `Stripe-Account` on the refund and on every read of the refund's records.
@@ -339,7 +340,10 @@ async function setup(options: Readonly<{
const challenge = new Uint8Array(randomBytes(32));
// The root is the only trust anchor and may delegate once, to the agent.
// The managers are approver anchors: they approve and hold no authority.
- const anchors = [anchor("root", principals.root!, audience, tool, 1, notBefore, expiresAt)];
+ const rootAnchor = anchor("root", principals.root!, audience, tool, 1, notBefore, expiresAt);
+ const anchors = [{ ...rootAnchor, permissions: [...rootAnchor.permissions,
+ ...["auths_gateway_candidate_check_v1", "auths_gateway_recovery_read_v1"].map((name) =>
+ ({ capability: "tools/call", resource: `${audience}/tools/${name}` }))] }];
const approvers: ApproverAnchor[] = MANAGERS.map((name) => ({
principal: principals[name]!, acceptedMethods: ["raw-key-v1"], notBefore, expiresAt,
}));
@@ -569,8 +573,61 @@ function recordEntry(audit: string, entry: Entry): "appended" | "replaced" | "un
return "unchanged";
}
+/** Private bounded development operator channel; the application never signs an effect approval. */
+export function requestOperator(path: string, request: Record): Promise> {
+ return new Promise((resolveRequest, reject) => {
+ const connection = createConnection(path);
+ const body = Buffer.from(JSON.stringify(request));
+ const maximum = 8 * 1024 * 1024;
+ if (body.length === 0 || body.length > maximum) {
+ connection.destroy(); reject(new Error("operator.frame-bound")); return;
+ }
+ const header = Buffer.alloc(4); header.writeUInt32BE(body.length);
+ let received = Buffer.alloc(0);
+ let completed = false;
+ connection.setTimeout(620_000, () => connection.destroy(new Error("operator.review-timeout")));
+ connection.on("error", reject);
+ connection.on("connect", () => connection.write(Buffer.concat([header, body])));
+ connection.on("data", (chunk: Buffer) => {
+ received = Buffer.concat([received, chunk]);
+ if (received.length < 4) return;
+ const length = received.readUInt32BE(0);
+ if (length === 0 || length > maximum || received.length > length + 4) {
+ connection.destroy(new Error("operator.frame-bound")); return;
+ }
+ if (received.length === length + 4) {
+ try {
+ const result = JSON.parse(received.subarray(4).toString()) as Record;
+ if (typeof result.error === "string") throw new Error(result.error);
+ completed = true; resolveRequest(result); connection.end();
+ } catch (error) { connection.destroy(); reject(error); }
+ }
+ });
+ connection.on("end", () => { if (!completed) reject(new Error("operator.truncated-frame")); });
+ });
+}
+
+function readEffectApproval(path: string): Uint8Array {
+ const file = openSync(path, "r");
+ try {
+ const bytes = new Uint8Array(4097);
+ let length = 0;
+ while (length < bytes.length) {
+ const count = readSync(file, bytes, length, bytes.length - length, null);
+ if (count === 0) break;
+ length += count;
+ }
+ // The native client enforces its public bound; parsing and signature
+ // validation remain in the gateway.
+ return bytes.subarray(0, length);
+ } finally {
+ closeSync(file);
+ }
+}
+
async function submit(options: Readonly<{
- state: string; socket: string; operationId: string; responses: string;
+ state: string; socket: string; operationId: string; responses: string; effectApproval?: string | undefined;
+ operatorSocket?: string | undefined; recover: boolean;
}>): Promise> {
const state = options.state;
const facts = JSON.parse(readFileSync(join(state, "setup.json"), "utf8")) as SetupFacts;
@@ -619,7 +676,24 @@ async function submit(options: Readonly<{
});
const proof = collection.assemble(action);
const gateway = new GatewayClient(new GatewayEndpoint(resolve(options.socket)));
- Object.assign(record, await gateway.submit({ proof, action: built.action }), { decided_by: "gateway" });
+ let approval: Uint8Array;
+ if (options.operatorSocket !== undefined) {
+ const reviewed = await requestOperator(options.operatorSocket,
+ { kind: "approve", proof: b64(proof), action: b64(built.action) });
+ approval = Buffer.from(String(reviewed.approval), "base64url");
+ record.effect_review = Object.fromEntries(Object.entries(reviewed).filter(([name]) => name !== "approval"));
+ } else {
+ approval = readEffectApproval(required(options.effectApproval, "effect-approval or operator-socket"));
+ }
+ Object.assign(record, await gateway.submit({ proof, action: built.action, approval }), { decided_by: "gateway" });
+ if (options.recover && record.outcome === "response-recorded" && record.status === 200) {
+ const read = await requestOperator(required(options.operatorSocket, "operator-socket for separate recovery"),
+ { kind: "recover", source: options.operationId });
+ record.recovery = JSON.parse(Buffer.from(await gateway.recover(options.operationId, {
+ proof: Buffer.from(String(read.proof), "base64url"), action: Buffer.from(String(read.action), "base64url"),
+ approval: Buffer.from(String(read.approval), "base64url"),
+ })).toString()) as unknown;
+ }
const observation = await gateway.observeOutcome(options.operationId);
const outcome = observation.outcome === "signed" ? observation.observation : null;
record.bundle = recordEntry(join(state, "audit"), {
@@ -641,7 +715,7 @@ function exportBundle(options: Readonly<{ state: string; out: string }>): void {
.map((line) => JSON.parse(line) as unknown)
: [];
const bundle = {
- schema: "auths.gateway-audit-bundle/2",
+ schema: "auths.gateway-audit-bundle/3",
recipe_b64: b64(new Uint8Array(readFileSync(RECIPE))),
profile_lock_b64: b64(new Uint8Array(readFileSync(PROFILE_LOCK))),
trusted_context_b64: b64(new Uint8Array(readFileSync(join(options.state, "trust", "gateway.context.cbor")))),
@@ -694,6 +768,9 @@ async function main(): Promise {
amount: { type: "string" },
required: { type: "string" },
responses: { type: "string" },
+ "effect-approval": { type: "string" },
+ "operator-socket": { type: "string" },
+ recover: { type: "boolean", default: false },
out: { type: "string" },
precheck: { type: "boolean", default: false },
help: { type: "boolean", default: false },
@@ -746,6 +823,9 @@ async function main(): Promise {
socket: required(values.socket, "socket"),
operationId: required(values["operation-id"], "operation-id"),
responses: resolve(required(values.responses, "responses")),
+ effectApproval: values["effect-approval"],
+ operatorSocket: values["operator-socket"],
+ recover: values.recover!,
}))}\n`);
break;
case "export":
diff --git a/formal/Auths/Product/Recovery.lean b/formal/Auths/Product/Recovery.lean
index 7d4e55403..357296d29 100644
--- a/formal/Auths/Product/Recovery.lean
+++ b/formal/Auths/Product/Recovery.lean
@@ -1,4 +1,5 @@
import Mathlib.Logic.ExistsUnique
+import Mathlib.Tactic.Tauto
/-!
# Recovery capability of a gateway recipe
@@ -7,7 +8,8 @@ What a recipe can prove after an ambiguous write is a total function of three
declarations (the observation locator, the echo, and the idempotency
mechanism) plus whether a pre-entry re-read is declared. The class names what
a later read-back can establish; `unknownResolution` says whether the gateway
-can ever resolve an `unknown` write, which only a verified-locator link can.
+can ever resolve an `unknown` write by provider observation, which only a verified-locator link can.
+Late source responses and authenticated recipient acknowledgements are separate evidence routes.
No declaration makes the provider write conditional.
The attempt record then moves through a closed stage table. The gateway store
@@ -143,14 +145,23 @@ theorem write_never_conditional (declarations : Declarations) :
/-! ## Attempt-record transitions -/
+/-- The independently authorized external boundary of this journal. -/
+inductive Protocol where
+ | mutation
+ | acquisition
+ | disclosure
+ deriving DecidableEq, Repr
+
/-- The stage of an attempt record. -/
inductive Stage where
- | attempting
+ | reserved
+ | entryGranted
| notEntered
| unknown
| responseRecorded
| observed
| observedByProvider
+ | acknowledged
deriving DecidableEq, Repr
/-- The read-back comparison recorded with an `observed` stage. -/
@@ -171,11 +182,15 @@ inductive Link where
/-- The fields of one stored attempt record the transition rule inspects;
byte strings are empty when absent. -/
structure AttemptView where
+ protocol : Protocol
+ revision : Nat
fixed : List Nat
+ entry : List Nat
stage : Stage
refusal : Bool
preEntry : List Nat
response : List Nat
+ acknowledgement : List Nat
locator : List Nat
reading : Reading
evidence : Bool
@@ -219,17 +234,49 @@ def providerStage : Stage → Bool
/-- The stages with no outgoing transition. -/
def terminalStage : Stage → Bool
- | .notEntered | .observed | .observedByProvider => true
- | .attempting | .unknown | .responseRecorded => false
+ | .notEntered | .observed | .observedByProvider | .acknowledged => true
+ | .reserved | .entryGranted | .unknown | .responseRecorded => false
def hasReading : Reading → Bool
| .none => false
| _ => true
+def requiresEntry : Stage → Bool
+ | .reserved | .notEntered => false
+ | _ => true
+
+def acknowledgementStage : Stage → Bool
+ | .acknowledged => true
+ | _ => false
+
+/-- Protocols have disjoint evidence vocabularies. A response is not a delivery
+acknowledgement; neither is an independently authorized provider observation. -/
+def protocolConsistent (view : AttemptView) : Bool :=
+ match view.protocol with
+ | .mutation => decide (view.stage ≠ .acknowledged) && view.acknowledgement.isEmpty
+ | .acquisition =>
+ (match view.stage with
+ | .reserved | .entryGranted | .notEntered | .unknown | .responseRecorded => true
+ | _ => false) &&
+ view.acknowledgement.isEmpty && view.preEntry.isEmpty && view.locator.isEmpty &&
+ !view.observation && !view.evidence && decide (view.reading = .none) &&
+ decide (view.link = .none)
+ | .disclosure =>
+ (match view.stage with
+ | .reserved | .entryGranted | .notEntered | .unknown | .acknowledged => true
+ | _ => false) &&
+ view.response.isEmpty && view.preEntry.isEmpty && view.locator.isEmpty &&
+ !view.observation && !view.evidence && decide (view.reading = .none) &&
+ decide (view.link = .none)
+
/-- A view's fields agree with its stage. -/
def consistent (view : AttemptView) : Bool :=
let hasResponse := !view.response.isEmpty
- decide (refusalStage view.stage = view.refusal) &&
+ protocolConsistent view && decide (view.revision > 0) &&
+ decide (view.revision ≤ 9007199254740991) &&
+ decide (requiresEntry view.stage ≠ view.entry.isEmpty) &&
+ decide (acknowledgementStage view.stage ≠ view.acknowledgement.isEmpty) &&
+ decide (refusalStage view.stage = view.refusal) &&
(!requiresResponse view.stage || hasResponse) &&
(!hasResponse || permitsResponse view.stage) &&
(view.locator.isEmpty || hasResponse) &&
@@ -242,28 +289,39 @@ def consistent (view : AttemptView) : Bool :=
/-- The closed stage table, given the stored plan's link. -/
def stageTransitionAllowed (source target : Stage) (link : Link) : Bool :=
match source, target with
- | .attempting, .attempting | .attempting, .notEntered | .attempting, .unknown
- | .attempting, .responseRecorded => true
- | .attempting, .observedByProvider => resolvesUnknown link
- | .unknown, .observedByProvider => resolvesUnknown link
+ | .reserved, .reserved | .reserved, .notEntered | .reserved, .entryGranted => true
+ | .entryGranted, .unknown | .entryGranted, .responseRecorded | .entryGranted, .acknowledged
+ | .unknown, .responseRecorded | .unknown, .acknowledged => true
+ | .entryGranted, .observedByProvider | .unknown, .observedByProvider => resolvesUnknown link
| .responseRecorded, .observed | .responseRecorded, .observedByProvider => true
| _, _ => false
def addsPreEntry (source target : Stage) : Bool :=
match source, target with
- | .attempting, .attempting | .attempting, .notEntered => true
+ | .reserved, .reserved | .reserved, .notEntered => true
| _, _ => false
def addsResponse (source target : Stage) : Bool :=
match source, target with
- | .attempting, .responseRecorded => true
+ | .entryGranted, .responseRecorded | .unknown, .responseRecorded => true
| _, _ => false
def checkpointValid (old new : AttemptView) : Bool :=
match old.stage, new.stage with
- | .attempting, .attempting => old.preEntry.isEmpty && !new.preEntry.isEmpty
+ | .reserved, .reserved => old.preEntry.isEmpty && !new.preEntry.isEmpty
| _, _ => true
+def entryKept (old new : AttemptView) : Bool :=
+ match old.stage, new.stage with
+ | .reserved, .entryGranted => old.entry.isEmpty && !new.entry.isEmpty
+ | _, _ => decide (old.entry = new.entry)
+
+def acknowledgementKept (old new : AttemptView) : Bool :=
+ match old.stage, new.stage with
+ | .entryGranted, .acknowledged | .unknown, .acknowledged =>
+ old.acknowledgement.isEmpty && !new.acknowledgement.isEmpty
+ | _, _ => decide (old.acknowledgement = new.acknowledgement)
+
def preEntryKept (old new : AttemptView) : Bool :=
if old.preEntry.isEmpty then new.preEntry.isEmpty || addsPreEntry old.stage new.stage
else decide (old.preEntry = new.preEntry)
@@ -277,9 +335,11 @@ def planKept (old new : AttemptView) : Bool :=
/-- Whether `new` may replace `old`. -/
def validTransition (old new : AttemptView) : Bool :=
- decide (old.fixed = new.fixed) && planKept old new &&
+ decide (old.protocol = new.protocol) && decide (old.revision < 9007199254740991) &&
+ decide (new.revision = old.revision + 1) && consistent old && entryKept old new &&
+ decide (old.fixed = new.fixed) && planKept old new &&
stageTransitionAllowed old.stage new.stage old.link && checkpointValid old new &&
- preEntryKept old new && responseKept old new && consistent new
+ preEntryKept old new && responseKept old new && acknowledgementKept old new && consistent new
/-- Each record replaces the previous one by a valid transition. -/
def validChain : List AttemptView → Prop
@@ -311,27 +371,45 @@ theorem transition_consistent {old new : AttemptView} (valid : validTransition o
simp only [validTransition, Bool.and_eq_true] at valid
exact valid.2
+/-- Every accepted transition retains its protocol, advances exactly one bounded
+revision, and preserves a grant once present. This is a projection invariant;
+the runtime authenticates the grant before constructing the view. -/
+theorem transition_ownership {old new : AttemptView} (valid : validTransition old new = true) :
+ new.protocol = old.protocol ∧ new.revision = old.revision + 1 ∧
+ new.revision ≤ 9007199254740991 ∧
+ (old.entry ≠ [] → new.entry = old.entry) := by
+ have parts := valid
+ simp only [validTransition, Bool.and_eq_true, decide_eq_true_eq] at parts
+ have consistentOld : consistent old = true := by tauto
+ have consistentNew : consistent new = true := by tauto
+ have kept : entryKept old new = true := by tauto
+ refine ⟨by tauto, by tauto, ?_, ?_⟩
+ · simp only [consistent, Bool.and_eq_true, decide_eq_true_eq] at consistentNew
+ tauto
+ · intro present
+ cases oldStage : old.stage <;> cases newStage : new.stage <;>
+ simp_all [entryKept, consistent, requiresEntry, List.isEmpty_iff]
+
theorem transition_keeps {old new : AttemptView} (consistentOld : consistent old = true)
(valid : validTransition old new = true) : Keeps old new := by
simp only [validTransition, Bool.and_eq_true, decide_eq_true_eq] at valid
- obtain ⟨⟨⟨⟨⟨⟨fixedEq, planEq⟩, _⟩, _⟩, preKept⟩, responseKeptEq⟩, _⟩ := valid
+ have fixedEq : old.fixed = new.fixed := by tauto
+ have planEq : planKept old new = true := by tauto
+ have preKept : preEntryKept old new = true := by tauto
+ have responseKeptEq : responseKept old new = true := by tauto
simp only [planKept, Bool.and_eq_true, decide_eq_true_eq] at planEq
refine ⟨fixedEq.symm, planEq.1.symm, planEq.2.symm, fun present => ?_, fun present => ?_⟩
· have nonempty : old.preEntry.isEmpty = false := by
cases equal : old.preEntry with
| nil => exact absurd equal present
| cons _ _ => rfl
- simp only [preEntryKept, nonempty, Bool.false_eq_true, if_false, decide_eq_true_eq]
- at preKept
+ simp only [preEntryKept, nonempty, Bool.false_eq_true, if_false, decide_eq_true_eq] at preKept
exact preKept.symm
- · simp only [responseKept, Bool.or_eq_true, Bool.and_eq_true, decide_eq_true_eq]
- at responseKeptEq
+ · simp only [responseKept, Bool.or_eq_true, Bool.and_eq_true, decide_eq_true_eq] at responseKeptEq
rcases responseKeptEq with adds | ⟨responseEq, locatorEq⟩
· exfalso
- rcases old with ⟨fixed, stage, refusal, preEntry, response, locator, reading, evidence,
- observation, link⟩
- cases stage <;> cases equal : new.stage <;> simp_all [addsResponse] <;>
- cases response <;> simp_all [consistent, permitsResponse]
+ cases oldStage : old.stage <;> cases newStage : new.stage <;>
+ simp_all [addsResponse, consistent, permitsResponse, List.isEmpty_iff]
· exact ⟨responseEq.symm, locatorEq.symm⟩
theorem chain_step {views : List AttemptView} (chain : validChain views) (index : Nat)
@@ -386,75 +464,133 @@ theorem chain_keeps (views : List AttemptView) (chain : validChain views)
exact induction restChain restStart earlier later (by omega)
(by simp at bound; omega)
-/-- A record in `unknown` leaves it only for `observed-by-provider`, only
-with a verified-locator link (the link of exactly the `linked` class), and
-only on a transition that records provider evidence; over every valid
-sequence of records, the same holds at every position. -/
+/-- Only the separately linked provider-evidence route resolves an unknown
+as observed-by-provider. Late source responses and recipient ACKs carry distinct
+facts; they do not establish this observation claim. -/
theorem unknown_resolves_only_linked :
(∀ old new : AttemptView, validTransition old new = true → old.stage = .unknown →
- new.stage = .observedByProvider ∧ old.link = .verified ∧ new.link = .verified ∧
- new.evidence = true) ∧
+ new.stage = .observedByProvider →
+ old.link = .verified ∧ new.link = .verified ∧ new.evidence = true) ∧
(∀ views : List AttemptView, validChain views →
∀ (index : Nat) (bound : index + 1 < views.length),
(views[index]'(by omega)).stage = .unknown →
- views[index + 1].stage = .observedByProvider ∧
- views[index + 1].link = .verified ∧ views[index + 1].evidence = true) ∧
+ views[index + 1].stage = .observedByProvider →
+ views[index + 1].link = .verified ∧ views[index + 1].evidence = true) ∧
(∀ declarations : Declarations,
linkOf (capability declarations).providerLink = .verified ↔
(capability declarations).recoveryClass = .linked) := by
have single : ∀ old new : AttemptView, validTransition old new = true →
- old.stage = .unknown →
- new.stage = .observedByProvider ∧ old.link = .verified ∧ new.link = .verified ∧
- new.evidence = true := by
- intro old new valid unknown
- rcases old with ⟨fixed, stage, refusal, preEntry, response, locator, reading, evidence,
- observation, link⟩
- rcases new with ⟨fixed', stage', refusal', preEntry', response', locator', reading',
- evidence', observation', link'⟩
- subst unknown
- cases stage' <;> cases link <;>
- simp_all [validTransition, planKept, stageTransitionAllowed, resolvesUnknown, consistent,
- providerStage]
+ old.stage = .unknown → new.stage = .observedByProvider →
+ old.link = .verified ∧ new.link = .verified ∧ new.evidence = true := by
+ intro old new valid unknown provider
+ have stage : stageTransitionAllowed old.stage new.stage old.link = true := by
+ simp only [validTransition, Bool.and_eq_true] at valid
+ tauto
+ have plan : planKept old new = true := by
+ simp only [validTransition, Bool.and_eq_true] at valid
+ tauto
+ have consistency := transition_consistent valid
+ cases link : old.link <;>
+ simp_all [stageTransitionAllowed, resolvesUnknown, planKept, consistent, providerStage]
refine ⟨single, ?_, ?_⟩
- · intro views chain index bound unknown
- obtain ⟨stage, _, link, evidence⟩ := single _ _ (chain_step chain index bound) unknown
- exact ⟨stage, link, evidence⟩
+ · intro views chain index bound unknown provider
+ exact (single _ _ (chain_step chain index bound) unknown provider).2
· intro declarations
obtain ⟨observation, echo, idempotency, preEntry⟩ := declarations
- cases observation <;> cases echo <;>
- simp [capability, providerLink, classOf, linkOf]
+ cases observation <;> cases echo <;> simp [capability, providerLink, classOf, linkOf]
+
+/-- The closed unknown exit table includes late source evidence, but forbids
+re-entry, a new reservation, and a refunding not-entered result. -/
+theorem unknown_exits_with_bound_evidence (old new : AttemptView)
+ (valid : validTransition old new = true) (unknown : old.stage = .unknown) :
+ new.stage = .responseRecorded ∨ new.stage = .acknowledged ∨
+ (new.stage = .observedByProvider ∧ old.link = .verified) := by
+ have stage : stageTransitionAllowed old.stage new.stage old.link = true := by
+ simp only [validTransition, Bool.and_eq_true] at valid
+ tauto
+ cases target : new.stage <;> cases link : old.link <;>
+ simp_all [stageTransitionAllowed, resolvesUnknown]
+
+/-- A late response retains the source grant and protocol and advances one
+revision, records a nonempty response, and does not assert provider observation.
+Runtime validation of the response/source binding is outside this byte projection. -/
+theorem late_response_retains_source (old new : AttemptView)
+ (valid : validTransition old new = true) (unknown : old.stage = .unknown)
+ (response : new.stage = .responseRecorded) :
+ new.protocol = old.protocol ∧ new.revision = old.revision + 1 ∧
+ new.entry = old.entry ∧ old.entry ≠ [] ∧ new.response ≠ [] ∧
+ new.evidence = false ∧ new.acknowledgement = [] := by
+ have own := transition_ownership valid
+ have oldCons : consistent old = true := by
+ simp only [validTransition, Bool.and_eq_true] at valid
+ tauto
+ have newCons := transition_consistent valid
+ have entry : old.entry ≠ [] := by
+ simp [consistent, unknown, requiresEntry, List.isEmpty_iff] at oldCons
+ tauto
+ have newFields : new.response ≠ [] ∧ new.evidence = false ∧ new.acknowledgement = [] := by
+ simp [consistent, response, requiresResponse, providerStage, acknowledgementStage,
+ List.isEmpty_iff] at newCons
+ tauto
+ exact ⟨own.1, own.2.1, own.2.2.2 entry, entry, newFields⟩
+
+/-- A late recipient ACK is restricted to the disclosure protocol. It keeps the
+source grant/revision lineage, carries nonempty authenticated evidence, and
+cannot be confused with a response or provider observation. Signature and exact
+recipient/challenge binding are checked before the runtime projects these bytes. -/
+theorem late_acknowledgement_retains_source (old new : AttemptView)
+ (valid : validTransition old new = true) (unknown : old.stage = .unknown)
+ (acknowledged : new.stage = .acknowledged) :
+ old.protocol = .disclosure ∧ new.protocol = .disclosure ∧
+ new.revision = old.revision + 1 ∧ new.entry = old.entry ∧ old.entry ≠ [] ∧
+ new.acknowledgement ≠ [] ∧ new.response = [] ∧ new.evidence = false := by
+ have own := transition_ownership valid
+ have oldCons : consistent old = true := by
+ simp only [validTransition, Bool.and_eq_true] at valid
+ tauto
+ have newCons := transition_consistent valid
+ have entry : old.entry ≠ [] := by
+ simp [consistent, unknown, requiresEntry, List.isEmpty_iff] at oldCons
+ tauto
+ have fields : new.protocol = .disclosure ∧ new.acknowledgement ≠ [] ∧
+ new.response = [] ∧ new.evidence = false := by
+ cases protocol : new.protocol <;>
+ simp [consistent, protocolConsistent, protocol, acknowledged, acknowledgementStage,
+ providerStage, permitsResponse, List.isEmpty_iff] at newCons <;>
+ simp_all only [and_true, true_and] <;> tauto
+ exact ⟨own.1 ▸ fields.1, fields.1, own.2.1, own.2.2.2 entry, entry, fields.2⟩
/-- `observed-by-provider` is reached only through a transition that records
provider evidence, under a declared link of the plan stored at the claim:
-from `response-recorded`, or from `attempting` or `unknown` only with a
-verified-locator link. For a sequence starting from a claimed `attempting`
+from `response-recorded`, or from `entry-granted` or `unknown` only with a
+verified-locator link. For a sequence starting from a claimed `reserved`
record, every `observed-by-provider` record was reached this way. -/
theorem provider_claim_requires_evidence :
(∀ old new : AttemptView, validTransition old new = true →
new.stage = .observedByProvider →
new.evidence = true ∧ hasLink new.link = true ∧ new.link = old.link ∧
(old.stage = .responseRecorded ∨
- ((old.stage = .attempting ∨ old.stage = .unknown) ∧ old.link = .verified))) ∧
- (∀ (first : AttemptView) (rest : List AttemptView), first.stage = .attempting →
+ ((old.stage = .entryGranted ∨ old.stage = .unknown) ∧ old.link = .verified))) ∧
+ (∀ (first : AttemptView) (rest : List AttemptView), first.stage = .reserved →
consistent first = true → validChain (first :: rest) →
∀ (index : Nat) (bound : index < rest.length),
rest[index].stage = .observedByProvider →
rest[index].evidence = true ∧ rest[index].link = first.link ∧
hasLink first.link = true ∧
(((first :: rest)[index]'(by simp; omega)).stage = .responseRecorded ∨
- ((((first :: rest)[index]'(by simp; omega)).stage = .attempting ∨
+ ((((first :: rest)[index]'(by simp; omega)).stage = .entryGranted ∨
((first :: rest)[index]'(by simp; omega)).stage = .unknown) ∧
first.link = .verified))) := by
have single : ∀ old new : AttemptView, validTransition old new = true →
new.stage = .observedByProvider →
new.evidence = true ∧ hasLink new.link = true ∧ new.link = old.link ∧
(old.stage = .responseRecorded ∨
- ((old.stage = .attempting ∨ old.stage = .unknown) ∧ old.link = .verified)) := by
+ ((old.stage = .entryGranted ∨ old.stage = .unknown) ∧ old.link = .verified)) := by
intro old new valid provider
- rcases old with ⟨fixed, stage, refusal, preEntry, response, locator, reading, evidence,
- observation, link⟩
- rcases new with ⟨fixed', stage', refusal', preEntry', response', locator', reading',
- evidence', observation', link'⟩
+ rcases old with ⟨protocol, revision, fixed, entry, stage, refusal, preEntry, response, acknowledgement,
+ locator, reading, evidence, observation, link⟩
+ rcases new with ⟨protocol', revision', fixed', entry', stage', refusal', preEntry', response', acknowledgement',
+ locator', reading', evidence', observation', link'⟩
subst provider
cases stage <;> cases link <;> cases link' <;>
simp_all [validTransition, planKept, stageTransitionAllowed, resolvesUnknown, consistent,
@@ -478,12 +614,12 @@ theorem provider_claim_requires_evidence :
· rw [← previousLink]
exact source
-/-- `not-entered`, `observed`, and `observed-by-provider` are exactly the
+/-- `not-entered`, `observed`, `observed-by-provider`, and `acknowledged` are the
terminal stages; no transition leaves them, so in every valid sequence a
terminal record is the last. -/
theorem terminal_stages_final :
(∀ stage : Stage, terminalStage stage = true ↔
- stage = .notEntered ∨ stage = .observed ∨ stage = .observedByProvider) ∧
+ stage = .notEntered ∨ stage = .observed ∨ stage = .observedByProvider ∨ stage = .acknowledged) ∧
(∀ old new : AttemptView, terminalStage old.stage = true →
validTransition old new = false) ∧
(∀ views : List AttemptView, validChain views →
@@ -492,8 +628,8 @@ theorem terminal_stages_final :
have single : ∀ old new : AttemptView, terminalStage old.stage = true →
validTransition old new = false := by
intro old new terminal
- rcases old with ⟨fixed, stage, refusal, preEntry, response, locator, reading, evidence,
- observation, link⟩
+ rcases old with ⟨protocol, revision, fixed, entry, stage, refusal, preEntry, response, acknowledgement,
+ locator, reading, evidence, observation, link⟩
cases stage <;> simp_all [terminalStage, validTransition, stageTransitionAllowed]
refine ⟨fun stage => by cases stage <;> simp [terminalStage], single, ?_⟩
intro views chain index bound terminal
diff --git a/formal/Auths/Product/Refinement/Gateway.lean b/formal/Auths/Product/Refinement/Gateway.lean
index ee13c66c5..8b466a866 100644
--- a/formal/Auths/Product/Refinement/Gateway.lean
+++ b/formal/Auths/Product/Refinement/Gateway.lean
@@ -1796,11 +1796,6 @@ theorem translated_relative_basis_refines_model (value : Std.U64) (subtrahend :
section Order
open Auths.Product.SubmitOrder
-def modeOf : order.Mode → Mode
- | .Entry => .entry
- | .ReadBack => .readBack
- | .Reobserve => .reobserve
-
def planOf (plan : order.SubmitPlan) : Plan where
accountScope := plan.account_scope
accountRead := plan.account_read
@@ -1817,22 +1812,21 @@ def phaseOf : order.Phase → Phase
| .Scope => .scope
| .Prepare => .prepare
| .Claim => .claim
- | .Resume => .resume
- | .Reload mode => .reload (modeOf mode)
- | .Lease mode => .lease (modeOf mode)
- | .Prefix mode => .prefixGuard (modeOf mode)
- | .Account mode => .account (modeOf mode)
- | .Denied mode index => .denied (modeOf mode) index.val
+ | .Reload => .reload
+ | .Lease => .lease
+ | .Prefix => .prefixGuard
+ | .Account => .account
+ | .Denied index => .denied index.val
| .PreEntry => .preEntry
| .Ceiling => .ceiling
| .Checkpoint => .checkpoint
| .EntryReload => .entryReload
| .Deadline => .deadline
+ | .Grant => .grant
| .Send => .send
| .RecordResponse => .recordResponse
| .RecordUnknown => .recordUnknown
| .RecordNotEntered => .recordNotEntered
- | .ReadBack mode => .readBack (modeOf mode)
| .Done => .done
def stateOf (state : order.SubmitState) : State where
@@ -1845,7 +1839,7 @@ def verificationOf : order.Verification → Verification
| .Authorized argument => .authorized argument.val
def claimOf : order.ClaimResult → ClaimResult
- | .Inserted => .inserted
+ | .Reserved => .reserved
| .Refused => .refused
| .Replay => .replay
| .Unavailable => .unavailable
@@ -1870,13 +1864,12 @@ def ceilingOf : order.CeilingRead → CeilingRead
| .Read basis bindsEqual => .read basis.val bindsEqual
def writeOf : order.WriteResult → WriteResult
- | .NotEntered => .notEntered
| .Unknown => .unknown
| .Response => .response
def responseOf : order.ResponseRecord → ResponseRecord
| .Failed => .failed
- | .Recorded observable => .recorded observable
+ | .Recorded => .recorded
def eventOf : order.SubmitEvent → Event
| .Start => .start
@@ -1886,7 +1879,6 @@ def eventOf : order.SubmitEvent → Event
| .Scope bound => .scope bound
| .Preparation prepared => .preparation prepared
| .Claim result => .claim (claimOf result)
- | .Resume resumable => .resume resumable
| .Reload unchanged => .reload unchanged
| .Lease leased => .lease leased
| .Prefix allowed => .prefixGuard allowed
@@ -1896,6 +1888,7 @@ def eventOf : order.SubmitEvent → Event
| .Ceiling read => .ceiling (ceilingOf read)
| .Recorded stored => .recorded stored
| .Deadline within => .deadline within
+ | .Granted granted => .granted granted
| .Write result => .write (writeOf result)
| .Response record => .response (responseOf record)
@@ -1922,8 +1915,8 @@ def stopOf : order.Stop → Stop
| .NotEntered => .notEntered
| .Unknown => .unknown
| .Response => .response
- | .Observed => .observed
- | .ReplayRefused => .replayRefused
+ | .Stored => .stored
+ | .GrantUnavailable => .grantUnavailable
| .Halted => .halted
def actionOf : order.SubmitAction → Action
@@ -1933,22 +1926,21 @@ def actionOf : order.SubmitAction → Action
| .BindScope => .bindScope
| .Prepare => .prepare
| .Claim => .claim
- | .Resume => .resume
- | .Reload mode => .reload (modeOf mode)
- | .Lease mode => .lease (modeOf mode)
- | .CheckPrefix mode => .checkPrefix (modeOf mode)
- | .ReadAccount mode => .readAccount (modeOf mode)
- | .DeniedRead mode index => .deniedRead (modeOf mode) index.val
+ | .Reload => .reload
+ | .Lease => .lease
+ | .CheckPrefix => .checkPrefix
+ | .ReadAccount => .readAccount
+ | .DeniedRead index => .deniedRead index.val
| .PreEntryRead => .preEntryRead
| .CeilingRead => .ceilingRead
| .RecordCheckpoint => .recordCheckpoint
| .ReloadBeforeEntry => .reloadBeforeEntry
| .CheckDeadline => .checkDeadline
+ | .GrantEntry => .grantEntry
| .Send => .send
| .RecordResponse => .recordResponse
| .RecordUnknown => .recordUnknown
| .RecordNotEntered refusal => .recordNotEntered (refusalCodeOf refusal)
- | .ReadBack mode => .readBack (modeOf mode)
| .Stop reason => .stop (stopOf reason)
def decisionOf (decision : order.SubmitDecision) : Decision where
@@ -1964,9 +1956,9 @@ theorem translated_start_refines_model (plan : order.SubmitPlan) :
/-- The next denied index of a refused read never overflows `u8`: it is at
most the declared count. -/
-theorem after_denied_refused_spec (state : order.SubmitState) (mode : order.Mode) (index : Std.U8) :
- order.after_denied state mode index .Refused ⦃ decision =>
- decisionOf decision = afterDenied (stateOf state) (modeOf mode) index.val .refused ⦄ := by
+theorem after_denied_refused_spec (state : order.SubmitState) (index : Std.U8) :
+ order.after_denied state index .Refused ⦃ decision =>
+ decisionOf decision = afterDenied (stateOf state) index.val .refused ⦄ := by
unfold order.after_denied
simp only [afterDenied, stateOf, planOf]
split
@@ -1987,11 +1979,10 @@ theorem after_denied_refused_spec (state : order.SubmitState) (mode : order.Mode
have done' : ¬ next.val < state.plan.denied_reads.val := by simpa using done
rw [nextVal'] at done'
simp only [done', if_false]
- rcases mode with _ | _ | _
all_goals
- simp [order.after_credentials, order.pre_entry_from, order.ceiling_from,
- order.after_reads, order.go, afterCredentials, preEntryFrom, ceilingFrom, afterReads, go,
- decisionOf, stateOf, planOf, phaseOf, actionOf, modeOf]
+ simp [order.pre_entry_from, order.ceiling_from,
+ order.after_reads, order.go, preEntryFrom, ceilingFrom, afterReads, go,
+ decisionOf, stateOf, planOf, phaseOf, actionOf]
repeat' split
all_goals simp_all
· rename_i above
@@ -2009,18 +2000,17 @@ theorem after_denied_refused_spec (state : order.SubmitState) (mode : order.Mode
@[simp] theorem stateOf_argument (state : order.SubmitState) :
(stateOf state).argument = state.argument.val := rfl
-theorem after_denied_spec (state : order.SubmitState) (mode : order.Mode) (index : Std.U8)
+theorem after_denied_spec (state : order.SubmitState) (index : Std.U8)
(result : order.DeniedResult) :
- order.after_denied state mode index result ⦃ decision =>
+ order.after_denied state index result ⦃ decision =>
decisionOf decision =
- afterDenied (stateOf state) (modeOf mode) index.val (deniedOf result) ⦄ := by
+ afterDenied (stateOf state) index.val (deniedOf result) ⦄ := by
match result with
- | .Refused => exact after_denied_refused_spec state mode index
+ | .Refused => exact after_denied_refused_spec state index
| .Answered | .Unavailable =>
- rcases mode with _ | _ | _ <;>
- simp [order.after_denied, order.lease_failed, order.not_entered, order.stop, order.go,
- afterDenied, leaseFailed, notEntered, stop, go, decisionOf, stateOf, phaseOf, actionOf,
- refusalCodeOf, stopOf, deniedOf, modeOf]
+ simp [order.after_denied, order.not_entered, order.stop, order.go,
+ afterDenied, notEntered, stop, go, decisionOf, stateOf, phaseOf, actionOf,
+ refusalCodeOf, stopOf, deniedOf]
theorem after_ceiling_spec (state : order.SubmitState) (read : order.CeilingRead) :
order.after_ceiling state read ⦃ decision =>
@@ -2046,26 +2036,25 @@ theorem translated_next_step_refines_model (state : order.SubmitState)
order.next_step state event ⦃ decision =>
decisionOf decision = nextStep (stateOf state) (eventOf event) ⦄ := by
rcases state with ⟨plan, phase, argument⟩
- rcases phase with _ | _ | _ | _ | _ | _ | _ | _ | (_ | _ | _) | (_ | _ | _) | (_ | _ | _) |
- (_ | _ | _) | ⟨(_ | _ | _), _⟩ | _ | _ | _ | _ | _ | _ | _ | _ | _ | (_ | _ | _) | _ <;>
- rcases event with _ | _ | (_ | _) | _ | _ | _ | (_ | _ | _ | _) | _ | _ | _ | _ |
- (_ | _ | _) | _ | (_ | _ | _) | _ | _ | _ | (_ | _ | _) | (_ | _) <;>
+ rcases phase with _ | _ | _ | _ | _ | _ | _ | _ | _ | _ | _ | _ | _ | _ | _ | _ | _ | _ | _ | _ | _ | _ | _ <;>
+ rcases event with _ | _ | (_ | _) | _ | _ | _ | (_ | _ | _ | _) | _ | _ | _ |
+ (_ | _ | _) | _ | (_ | _ | _) | _ | _ | _ | _ | (_ | _) | (_ | _) <;>
simp only [order.next_step, order.pre_claim_step, order.lease_step, order.entry_step,
- nextStep, preClaimStep, leaseStep, entryStep, stateOf_phase, phaseOf, eventOf, modeOf] <;>
+ nextStep, preClaimStep, leaseStep, entryStep, stateOf_phase, phaseOf, eventOf] <;>
first
- | exact after_denied_spec _ _ _ _
+ | exact after_denied_spec _ _ _
| exact after_ceiling_spec _ _
| skip
all_goals
- simp [order.go, order.stop, order.halt, order.not_entered, order.lease_failed,
- order.after_reads, order.ceiling_from, order.pre_entry_from, order.after_credentials,
+ simp [order.go, order.stop, order.halt, order.not_entered,
+ order.after_reads, order.ceiling_from, order.pre_entry_from,
order.denied_from, order.after_prefix, order.after_admission, order.pre_claim,
order.after_claim, order.after_account, order.after_pre_entry, order.after_write,
- order.after_response, order.after_read_back, go, stop, halt, notEntered, leaseFailed,
- afterReads, ceilingFrom, preEntryFrom, afterCredentials, deniedFrom, afterPrefix,
+ order.after_response, go, stop, halt, notEntered,
+ afterReads, ceilingFrom, preEntryFrom, deniedFrom, afterPrefix,
afterAdmission, preClaim, afterClaim, afterAccount, afterPreEntry, afterWrite,
- afterResponse, afterReadBack, decisionOf, stateOf, planOf, phaseOf, actionOf, refusalCodeOf,
- stopOf, modeOf, verificationOf, claimOf, accountOf, preEntryOf, writeOf,
+ afterResponse, decisionOf, stateOf, planOf, phaseOf, actionOf, refusalCodeOf,
+ stopOf, verificationOf, claimOf, accountOf, preEntryOf, writeOf,
responseOf] <;>
(repeat' split) <;> simp_all
@@ -2089,14 +2078,22 @@ theorem byteList_inj {left right : List Std.U8} : byteList left = byteList right
· intro equal
rw [equal]
+def protocolOf : transition.Protocol → Auths.Product.Recovery.Protocol
+ | .Mutation => .mutation
+ | .Acquisition => .acquisition
+ | .Disclosure => .disclosure
+
def stageOf : transition.Stage → Auths.Product.Recovery.Stage
- | .Attempting => .attempting
+ | .Reserved => .reserved
+ | .EntryGranted => .entryGranted
| .NotEntered => .notEntered
| .Unknown => .unknown
| .ResponseRecorded => .responseRecorded
| .Observed => .observed
| .ObservedByProvider => .observedByProvider
+ | .Acknowledged => .acknowledged
+
def readingOf : transition.Reading → Auths.Product.Recovery.Reading
| .None => .none
| .Match => .matched
@@ -2109,6 +2106,10 @@ def storedLinkOf : transition.Link → Auths.Product.Recovery.Link
| .AfterResponse => .afterResponse
def viewOf (view : transition.AttemptView) : Auths.Product.Recovery.AttemptView where
+ protocol := protocolOf view.protocol
+ revision := view.revision.val
+ entry := byteList view.entry.val
+ acknowledgement := byteList view.acknowledgement.val
fixed := byteList view.fixed.val
stage := stageOf view.stage
refusal := view.refusal
@@ -2204,46 +2205,166 @@ theorem same_bytes_eq (left right : Slice Std.U8) :
have := congrArg List.length same
simpa [byteList] using this
+namespace TransitionFacts
+
+open Auths.Product.Recovery
+
+theorem requires_entry_eq (value : transition.Stage) :
+ transition.requires_entry value = ok (requiresEntry (stageOf value)) := by
+ cases value <;> rfl
+
+theorem acknowledgement_stage_eq (value : transition.Stage) :
+ transition.acknowledgement_stage value = ok (acknowledgementStage (stageOf value)) := by
+ cases value <;> rfl
+
+theorem refusal_stage_eq (value : transition.Stage) :
+ transition.refusal_stage value = ok (refusalStage (stageOf value)) := by
+ cases value <;> rfl
+
+theorem requires_response_eq (value : transition.Stage) :
+ transition.requires_response value = ok (requiresResponse (stageOf value)) := by
+ cases value <;> rfl
+
+theorem permits_response_eq (value : transition.Stage) :
+ transition.permits_response value = ok (permitsResponse (stageOf value)) := by
+ cases value <;> rfl
+
+theorem observed_stage_eq (value : transition.Stage) :
+ transition.observed_stage value = ok (observedStage (stageOf value)) := by
+ cases value <;> rfl
+
+theorem provider_stage_eq (value : transition.Stage) :
+ transition.provider_stage value = ok (providerStage (stageOf value)) := by
+ cases value <;> rfl
+
+theorem has_reading_eq (value : transition.Reading) :
+ transition.has_reading value = ok (hasReading (readingOf value)) := by
+ cases value <;> rfl
+
+theorem has_link_eq (value : transition.Link) :
+ transition.has_link value = ok (hasLink (storedLinkOf value)) := by
+ cases value <;> rfl
+
+theorem resolves_unknown_eq (value : transition.Link) :
+ transition.resolves_unknown value = ok (resolvesUnknown (storedLinkOf value)) := by
+ cases value <;> rfl
+
+theorem protocol_consistent_eq (view : transition.AttemptView) :
+ transition.protocol_consistent view = ok (protocolConsistent (viewOf view)) := by
+ rcases view with ⟨protocol, revision, fixed, entry, stage, refusal, preEntry, response,
+ acknowledgement, locator, reading, evidence, observation, link⟩
+ cases protocol <;> cases stage <;> cases reading <;> cases link <;>
+ simp [transition.protocol_consistent, transition.acknowledgement_stage, absent_eq,
+ alloc.vec.Vec.deref, protocolConsistent, viewOf, protocolOf, stageOf, readingOf,
+ storedLinkOf] <;> (repeat' split) <;> simp_all
+
theorem consistent_eq (view : transition.AttemptView) :
- transition.consistent view = ok (Auths.Product.Recovery.consistent (viewOf view)) := by
- rcases view with ⟨fixed, stage, refusal, preEntry, response, locator, reading, evidence,
- observation, link⟩
- simp only [transition.consistent, absent_eq, bind_tc_ok, alloc.vec.Vec.deref,
- Auths.Product.Recovery.consistent, viewOf]
- rcases stage with _ | _ | _ | _ | _ | _ <;> rcases reading with _ | _ | _ | _ <;>
- rcases link with _ | _ | _ <;>
- simp [transition.refusal_stage, transition.requires_response, transition.permits_response,
- transition.observed_stage, transition.provider_stage, transition.has_reading,
- transition.has_link, transition.resolves_unknown, stageOf, readingOf, storedLinkOf,
- Auths.Product.Recovery.refusalStage, Auths.Product.Recovery.requiresResponse,
- Auths.Product.Recovery.permitsResponse, Auths.Product.Recovery.observedStage,
- Auths.Product.Recovery.providerStage, Auths.Product.Recovery.hasReading,
- Auths.Product.Recovery.hasLink, Auths.Product.Recovery.resolvesUnknown] <;>
+ transition.consistent view = ok (consistent (viewOf view)) := by
+ rcases view with ⟨protocol, revision, fixed, entry, stage, refusal, preEntry, response,
+ acknowledgement, locator, reading, evidence, observation, link⟩
+ cases stage <;> cases reading <;> cases link <;>
+ simp [transition.consistent, absent_eq, protocol_consistent_eq, requires_entry_eq,
+ acknowledgement_stage_eq, refusal_stage_eq, requires_response_eq, permits_response_eq,
+ observed_stage_eq, provider_stage_eq, has_reading_eq, has_link_eq, resolves_unknown_eq,
+ alloc.vec.Vec.deref, consistent, viewOf, UScalar.lt_equiv, UScalar.le_equiv,
+ stageOf, readingOf, storedLinkOf, requiresEntry, acknowledgementStage, refusalStage,
+ requiresResponse, permitsResponse, observedStage, providerStage, hasReading,
+ hasLink, resolvesUnknown] <;> (repeat' split) <;> simp_all
+
+theorem same_protocol_eq (left right : transition.Protocol) :
+ transition.same_protocol left right = ok (decide (protocolOf left = protocolOf right)) := by
+ cases left <;> cases right <;> rfl
+
+theorem link_equal_eq (left right : transition.Link) :
+ transition.link_equal left right = ok (decide (storedLinkOf left = storedLinkOf right)) := by
+ cases left <;> cases right <;> rfl
+
+theorem stage_transition_allowed_eq (source target : transition.Stage) (link : transition.Link) :
+ transition.stage_transition_allowed source target link =
+ ok (stageTransitionAllowed (stageOf source) (stageOf target) (storedLinkOf link)) := by
+ cases source <;> cases target <;> cases link <;> rfl
+
+theorem adds_pre_entry_eq (source target : transition.Stage) :
+ transition.adds_pre_entry source target = ok (addsPreEntry (stageOf source) (stageOf target)) := by
+ cases source <;> cases target <;> rfl
+
+theorem adds_response_eq (source target : transition.Stage) :
+ transition.adds_response source target = ok (addsResponse (stageOf source) (stageOf target)) := by
+ cases source <;> cases target <;> rfl
+
+theorem checkpoint_valid_eq (old new : transition.AttemptView) :
+ transition.checkpoint_valid old new = ok (checkpointValid (viewOf old) (viewOf new)) := by
+ cases oldStage : old.stage <;> cases newStage : new.stage <;>
+ simp [transition.checkpoint_valid, checkpointValid, viewOf,
+ oldStage, newStage, stageOf, absent_eq, same_bytes_eq, alloc.vec.Vec.deref] <;>
+ (repeat' split) <;> simp_all
+
+theorem entry_kept_eq (old new : transition.AttemptView) :
+ transition.entry_kept old new = ok (entryKept (viewOf old) (viewOf new)) := by
+ cases oldStage : old.stage <;> cases newStage : new.stage <;>
+ simp [transition.entry_kept, entryKept, viewOf,
+ oldStage, newStage, stageOf, absent_eq, same_bytes_eq, alloc.vec.Vec.deref] <;>
+ (repeat' split) <;> simp_all
+
+theorem acknowledgement_kept_eq (old new : transition.AttemptView) :
+ transition.acknowledgement_kept old new = ok (acknowledgementKept (viewOf old) (viewOf new)) := by
+ cases oldStage : old.stage <;> cases newStage : new.stage <;>
+ simp [transition.acknowledgement_kept, transition.acknowledgement_stage, acknowledgementKept, viewOf,
+ oldStage, newStage, stageOf, absent_eq, same_bytes_eq, alloc.vec.Vec.deref] <;>
(repeat' split) <;> simp_all
-/-- The translated transition rule returns exactly the model's decision for
-every pair of records. -/
+theorem pre_entry_kept_eq (old new : transition.AttemptView) :
+ transition.pre_entry_kept old new = ok (preEntryKept (viewOf old) (viewOf new)) := by
+ simp only [transition.pre_entry_kept, preEntryKept, viewOf, absent_eq, same_bytes_eq,
+ alloc.vec.Vec.deref, adds_pre_entry_eq, bind_tc_ok]
+ (repeat' split) <;> simp_all
+
+theorem response_kept_eq (old new : transition.AttemptView) :
+ transition.response_kept old new = ok (responseKept (viewOf old) (viewOf new)) := by
+ simp only [transition.response_kept, responseKept, viewOf, same_bytes_eq,
+ alloc.vec.Vec.deref, adds_response_eq, bind_tc_ok]
+ (repeat' split) <;> simp_all
+
+theorem plan_kept_eq (old new : transition.AttemptView) :
+ transition.plan_kept old new = ok (planKept (viewOf old) (viewOf new)) := by
+ simp only [transition.plan_kept, planKept, viewOf, link_equal_eq]
+ (repeat' split) <;> simp_all
+
+end TransitionFacts
+
+private theorem result_guard (condition : Prop) [Decidable condition] (value : Bool) :
+ (if condition then ok value else ok false) = ok (decide condition && value) := by
+ by_cases condition <;> simp_all
+
+/-- Exact refinement for all three protocols, including revision overflow refusal,
+immutable entry grants, and distinct response, recipient ACK and observation evidence. -/
theorem translated_valid_transition_refines_model (old new : transition.AttemptView) :
transition.valid_transition old new ⦃ valid =>
valid = Auths.Product.Recovery.validTransition (viewOf old) (viewOf new) ⦄ := by
- rcases old with ⟨fixed, stage, refusal, preEntry, response, locator, reading, evidence,
- observation, link⟩
- rcases new with ⟨fixed', stage', refusal', preEntry', response', locator', reading', evidence',
- observation', link'⟩
- simp only [transition.valid_transition, transition.plan_kept, transition.checkpoint_valid,
- transition.pre_entry_kept, transition.response_kept, same_bytes_eq, absent_eq,
- consistent_eq, bind_tc_ok, alloc.vec.Vec.deref, Auths.Product.Recovery.validTransition,
- Auths.Product.Recovery.planKept, Auths.Product.Recovery.checkpointValid,
- Auths.Product.Recovery.preEntryKept, Auths.Product.Recovery.responseKept, viewOf]
- rcases stage with _ | _ | _ | _ | _ | _ <;> rcases stage' with _ | _ | _ | _ | _ | _ <;>
- rcases link with _ | _ | _ <;> rcases link' with _ | _ | _ <;>
- simp [transition.stage_transition_allowed, transition.adds_pre_entry,
- transition.adds_response, transition.link_equal, transition.resolves_unknown, stageOf,
- storedLinkOf, Auths.Product.Recovery.stageTransitionAllowed,
- Auths.Product.Recovery.addsPreEntry, Auths.Product.Recovery.addsResponse,
- Auths.Product.Recovery.resolvesUnknown, List.isEmpty_iff] <;>
- (repeat' split) <;> simp_all <;>
- (repeat' split) <;> simp_all
+ unfold transition.valid_transition
+ rw [TransitionFacts.same_protocol_eq, bind_tc_ok]
+ by_cases protocol : protocolOf old.protocol = protocolOf new.protocol
+ · simp only [protocol, decide_true, if_true]
+ by_cases bounded : old.revision < 9007199254740991#u64
+ · simp only [bounded, if_true]
+ have bounded' : old.revision.val < 9007199254740991 := by simpa using bounded
+ have fits : old.revision.val + (1#u64 : Std.U64).val ≤ UScalar.max .U64 := by
+ scalar_tac
+ obtain ⟨next, nextEq, nextVal⟩ :=
+ spec_imp_exists (UScalar.add_spec (x := old.revision) (y := 1#u64) fits)
+ rw [nextEq, bind_tc_ok]
+ have revisionEq : (new.revision = next) ↔ new.revision.val = old.revision.val + 1 := by
+ rw [UScalar.eq_equiv, nextVal]
+ simp
+ simp only [TransitionFacts.consistent_eq, TransitionFacts.entry_kept_eq,
+ TransitionFacts.plan_kept_eq, TransitionFacts.stage_transition_allowed_eq,
+ TransitionFacts.checkpoint_valid_eq, TransitionFacts.pre_entry_kept_eq,
+ TransitionFacts.response_kept_eq, TransitionFacts.acknowledgement_kept_eq,
+ same_bytes_eq, alloc.vec.Vec.deref, bind_tc_ok, revisionEq, result_guard, spec_ok]
+ simp [Auths.Product.Recovery.validTransition, viewOf, protocol, bounded', Bool.and_assoc]
+ · have bounded' : ¬ old.revision.val < 9007199254740991 := by simpa using bounded
+ simp [bounded, Auths.Product.Recovery.validTransition, viewOf, bounded', spec_ok]
+ · simp [protocol, Auths.Product.Recovery.validTransition, viewOf, spec_ok]
end Transition
diff --git a/formal/Auths/Product/SubmitOrder.lean b/formal/Auths/Product/SubmitOrder.lean
index f1cdc1c16..2f1c50b14 100644
--- a/formal/Auths/Product/SubmitOrder.lean
+++ b/formal/Auths/Product/SubmitOrder.lean
@@ -22,13 +22,6 @@ namespace Auths.Product.SubmitOrder
open Auths.Product.RelativeCeiling (admits)
-/-- Why a lease is taken. -/
-inductive Mode where
- | entry
- | readBack
- | reobserve
- deriving DecidableEq, Repr
-
/-- The declarations the order depends on, fixed before the machine starts. -/
structure Plan where
accountScope : Bool
@@ -48,22 +41,21 @@ inductive Phase where
| scope
| prepare
| claim
- | resume
- | reload (mode : Mode)
- | lease (mode : Mode)
- | prefixGuard (mode : Mode)
- | account (mode : Mode)
- | denied (mode : Mode) (index : Nat)
+ | reload
+ | lease
+ | prefixGuard
+ | account
+ | denied (index : Nat)
| preEntry
| ceiling
| checkpoint
| entryReload
| deadline
+ | grant
| send
| recordResponse
| recordUnknown
| recordNotEntered
- | readBack (mode : Mode)
| done
deriving DecidableEq, Repr
@@ -83,7 +75,7 @@ inductive Verification where
/-- The claim result. -/
inductive ClaimResult where
- | inserted
+ | reserved
| refused
| replay
| unavailable
@@ -118,7 +110,6 @@ inductive CeilingRead where
/-- The provider write result. -/
inductive WriteResult where
- | notEntered
| unknown
| response
deriving DecidableEq, Repr
@@ -126,7 +117,7 @@ inductive WriteResult where
/-- The result of recording the response. -/
inductive ResponseRecord where
| failed
- | recorded (observable : Bool)
+ | recorded
deriving DecidableEq, Repr
/-- The result of the action the previous decision named. -/
@@ -138,7 +129,6 @@ inductive Event where
| scope (bound : Bool)
| preparation (prepared : Bool)
| claim (result : ClaimResult)
- | resume (resumable : Bool)
| reload (unchanged : Bool)
| lease (leased : Bool)
| prefixGuard (allowed : Bool)
@@ -148,6 +138,7 @@ inductive Event where
| ceiling (read : CeilingRead)
| recorded (stored : Bool)
| deadline (within : Bool)
+ | granted (granted : Bool)
| write (result : WriteResult)
| response (record : ResponseRecord)
deriving DecidableEq, Repr
@@ -178,8 +169,8 @@ inductive Stop where
| notEntered
| unknown
| response
- | observed
- | replayRefused
+ | stored
+ | grantUnavailable
| halted
deriving DecidableEq, Repr
@@ -191,22 +182,21 @@ inductive Action where
| bindScope
| prepare
| claim
- | resume
- | reload (mode : Mode)
- | lease (mode : Mode)
- | checkPrefix (mode : Mode)
- | readAccount (mode : Mode)
- | deniedRead (mode : Mode) (index : Nat)
+ | reload
+ | lease
+ | checkPrefix
+ | readAccount
+ | deniedRead (index : Nat)
| preEntryRead
| ceilingRead
| recordCheckpoint
| reloadBeforeEntry
| checkDeadline
+ | grantEntry
| send
| recordResponse
| recordUnknown
| recordNotEntered (refusal : Refusal)
- | readBack (mode : Mode)
| stop (stop : Stop)
deriving DecidableEq, Repr
@@ -232,12 +222,6 @@ def halt (state : State) : Decision := stop state .halted
def notEntered (state : State) (refusal : Refusal) : Decision :=
go state .recordNotEntered (.recordNotEntered refusal)
-def leaseFailed (state : State) (mode : Mode) (refusal : Refusal) : Decision :=
- match mode with
- | .entry => notEntered state refusal
- | .readBack => stop state .response
- | .reobserve => stop state .replayRefused
-
def afterReads (state : State) : Decision :=
if state.plan.preEntry || state.plan.relativeCeiling then
go state .checkpoint .recordCheckpoint
@@ -250,21 +234,15 @@ def ceilingFrom (state : State) : Decision :=
def preEntryFrom (state : State) : Decision :=
if state.plan.preEntry then go state .preEntry .preEntryRead else ceilingFrom state
-def afterCredentials (state : State) (mode : Mode) : Decision :=
- match mode with
- | .entry => preEntryFrom state
- | .readBack => go state (.readBack .readBack) (.readBack .readBack)
- | .reobserve => go state (.readBack .reobserve) (.readBack .reobserve)
-
-def deniedFrom (state : State) (mode : Mode) (index : Nat) : Decision :=
+def deniedFrom (state : State) (index : Nat) : Decision :=
if index < state.plan.deniedReads then
- go state (.denied mode index) (.deniedRead mode index)
+ go state (.denied index) (.deniedRead index)
else
- afterCredentials state mode
+ preEntryFrom state
-def afterPrefix (state : State) (mode : Mode) : Decision :=
- if state.plan.accountRead then go state (.account mode) (.readAccount mode)
- else deniedFrom state mode 0
+def afterPrefix (state : State) : Decision :=
+ if state.plan.accountRead then go state .account .readAccount
+ else deniedFrom state 0
def afterAdmission (state : State) : Decision :=
if state.plan.accountScope then go state .scope .bindScope else go state .prepare .prepare
@@ -273,21 +251,21 @@ def preClaim (state : State) (passed : Bool) (next : Decision) : Decision :=
if passed then next else stop state .refused
def afterClaim (state : State) : ClaimResult → Decision
- | .inserted => go state (.reload .entry) (.reload .entry)
+ | .reserved => go state .reload .reload
| .refused => stop state .claimRefused
- | .replay => go state .resume .resume
+ | .replay => stop state .stored
| .unavailable => stop state .storeUnavailable
-def afterAccount (state : State) (mode : Mode) : AccountResult → Decision
- | .equal => deniedFrom state mode 0
- | .mismatch => leaseFailed state mode .accountMismatch
- | .unavailable => leaseFailed state mode .accountUnavailable
+def afterAccount (state : State) : AccountResult → Decision
+ | .equal => deniedFrom state 0
+ | .mismatch => notEntered state .accountMismatch
+ | .unavailable => notEntered state .accountUnavailable
-def afterDenied (state : State) (mode : Mode) (index : Nat) : DeniedResult → Decision
+def afterDenied (state : State) (index : Nat) : DeniedResult → Decision
| .refused =>
- if index < state.plan.deniedReads then deniedFrom state mode (index + 1) else halt state
- | .answered => leaseFailed state mode .capabilityExcess
- | .unavailable => leaseFailed state mode .capabilityUnavailable
+ if index < state.plan.deniedReads then deniedFrom state (index + 1) else halt state
+ | .answered => notEntered state .capabilityExcess
+ | .unavailable => notEntered state .capabilityUnavailable
def afterPreEntry (state : State) : PreEntryResult → Decision
| .satisfied => ceilingFrom state
@@ -304,23 +282,12 @@ def afterCeiling (state : State) : CeilingRead → Decision
else notEntered state .ceilingAbove
def afterWrite (state : State) : WriteResult → Decision
- | .notEntered => notEntered state .transportNotEntered
| .unknown => go state .recordUnknown .recordUnknown
| .response => go state .recordResponse .recordResponse
def afterResponse (state : State) : ResponseRecord → Decision
| .failed => stop state .unknown
- | .recorded observable =>
- if observable then go state (.reload .readBack) (.reload .readBack)
- else stop state .response
-
-def afterReadBack (state : State) (mode : Mode) (recorded : Bool) : Decision :=
- if recorded then stop state .observed
- else
- match mode with
- | .entry => halt state
- | .readBack => stop state .response
- | .reobserve => stop state .replayRefused
+ | .recorded => stop state .response
/-- The step before the claim. -/
def preClaimStep (state : State) (event : Event) : Decision :=
@@ -339,19 +306,16 @@ def preClaimStep (state : State) (event : Event) : Decision :=
/-- A lease-time step. -/
def leaseStep (state : State) (event : Event) : Decision :=
match state.phase, event with
- | .resume, .resume resumable =>
- if resumable then go state (.reload .reobserve) (.reload .reobserve)
- else stop state .replayRefused
- | .reload mode, .reload unchanged =>
- if unchanged then go state (.lease mode) (.lease mode)
- else leaseFailed state mode .connectionChanged
- | .lease mode, .lease leased =>
- if leased then go state (.prefixGuard mode) (.checkPrefix mode)
- else leaseFailed state mode .credentialUnavailable
- | .prefixGuard mode, .prefixGuard allowed =>
- if allowed then afterPrefix state mode else leaseFailed state mode .modeGuard
- | .account mode, .account result => afterAccount state mode result
- | .denied mode index, .denied result => afterDenied state mode index result
+ | .reload, .reload unchanged =>
+ if unchanged then go state .lease .lease
+ else notEntered state .connectionChanged
+ | .lease, .lease leased =>
+ if leased then go state .prefixGuard .checkPrefix
+ else notEntered state .credentialUnavailable
+ | .prefixGuard, .prefixGuard allowed =>
+ if allowed then afterPrefix state else notEntered state .modeGuard
+ | .account, .account result => afterAccount state result
+ | .denied index, .denied result => afterDenied state index result
| _, _ => halt state
/-- A step between the credential checks and the stop. -/
@@ -365,23 +329,24 @@ def entryStep (state : State) (event : Event) : Decision :=
if unchanged then go state .deadline .checkDeadline
else notEntered state .connectionChanged
| .deadline, .deadline within =>
- if within then go state .send .send else notEntered state .entryDeadline
+ if within then go state .grant .grantEntry else notEntered state .entryDeadline
+ | .grant, .granted granted =>
+ if granted then go state .send .send else stop state .grantUnavailable
| .send, .write result => afterWrite state result
| .recordResponse, .response record => afterResponse state record
| .recordUnknown, .recorded _ => stop state .unknown
| .recordNotEntered, .recorded stored =>
if stored then stop state .notEntered else stop state .unknown
- | .readBack mode, .recorded recorded => afterReadBack state mode recorded
| _, _ => halt state
/-- The next state and action for `event`. -/
def nextStep (state : State) (event : Event) : Decision :=
match state.phase with
| .start | .clock | .verify | .admit | .scope | .prepare | .claim => preClaimStep state event
- | .resume | .reload _ | .lease _ | .prefixGuard _ | .account _ | .denied _ _ =>
+ | .reload | .lease | .prefixGuard | .account | .denied _ =>
leaseStep state event
- | .preEntry | .ceiling | .checkpoint | .entryReload | .deadline | .send | .recordResponse
- | .recordUnknown | .recordNotEntered | .readBack _ => entryStep state event
+ | .preEntry | .ceiling | .checkpoint | .entryReload | .deadline | .grant | .send | .recordResponse
+ | .recordUnknown | .recordNotEntered => entryStep state event
| .done => halt state
/-! ## Traces -/
@@ -421,12 +386,12 @@ def Emitted (trace : List Step) (action : Action) : Prop :=
/-- The actions that write the attempt store. -/
def writesStore : Action → Bool
| .claim | .recordCheckpoint | .recordNotEntered _ | .recordResponse | .recordUnknown
- | .readBack _ => true
+ | .grantEntry => true
| _ => false
/-- The actions that touch the credential or the provider write. -/
def leasesOrSends : Action → Bool
- | .lease _ | .send => true
+ | .lease | .send => true
| _ => false
theorem run_append (state : State) (events : List Event) (event : Event) :
@@ -507,9 +472,9 @@ theorem nextStep_plan (state : State) (event : Event) :
rcases state with ⟨plan, phase, argument⟩
cases phase <;> cases event <;>
simp only [nextStep, preClaimStep, leaseStep, entryStep, go, stop, halt, notEntered,
- leaseFailed, afterReads, ceilingFrom, preEntryFrom, afterCredentials, deniedFrom,
+ afterReads, ceilingFrom, preEntryFrom, deniedFrom,
afterPrefix, afterAdmission, preClaim, afterClaim, afterAccount, afterDenied, afterPreEntry,
- afterCeiling, afterWrite, afterResponse, afterReadBack] <;>
+ afterCeiling, afterWrite, afterResponse] <;>
(repeat' split) <;> rfl
/-- Induction over a trace one event at a time, from the end. -/
@@ -535,21 +500,19 @@ theorem stateAfter_plan (plan : Plan) (events : List Event) :
def Verified (trace : List Step) (argument : Nat) : Prop :=
Answered trace .verify (.verification (.authorized argument))
-/-- The claim a lease mode depends on: an inserted claim for the entry and
-its read-back, a replayed claim for a re-observation. -/
-def Claimed (trace : List Step) : Mode → Prop
- | .entry | .readBack => Answered trace .claim (.claim .inserted)
- | .reobserve => Answered trace .claim (.claim .replay)
+/-- The durable reservation accepted before credential access. -/
+def Claimed (trace : List Step) : Prop :=
+ Answered trace .claim (.claim .reserved)
-/-- Every credential check of a lease in `mode` passed: the lease, the
+/-- Every credential check passed: the lease, the
prefix guard, the account read when declared, and every declared denied
read. -/
-def CredentialsChecked (trace : List Step) (plan : Plan) (mode : Mode) : Prop :=
- Answered trace (.lease mode) (.lease true) ∧
- Answered trace (.prefixGuard mode) (.prefixGuard true) ∧
- (plan.accountRead = true → Answered trace (.account mode) (.account .equal)) ∧
+def CredentialsChecked (trace : List Step) (plan : Plan) : Prop :=
+ Answered trace .lease (.lease true) ∧
+ Answered trace .prefixGuard (.prefixGuard true) ∧
+ (plan.accountRead = true → Answered trace .account (.account .equal)) ∧
∀ index, index < plan.deniedReads →
- Answered trace (.denied mode index) (.denied .refused)
+ Answered trace (.denied index) (.denied .refused)
/-- The relative-ceiling read produced a basis with every bind equal, and
the ratio admits `argument`. -/
@@ -570,46 +533,47 @@ def Facts (trace : List Step) (state : State) : Prop :=
| .prepare | .claim =>
Verified trace state.argument ∧
(state.plan.accountScope = true → Answered trace .scope (.scope true))
- | .resume => Verified trace state.argument ∧ Claimed trace .reobserve
- | .reload mode | .lease mode => Verified trace state.argument ∧ Claimed trace mode
- | .prefixGuard mode =>
- Verified trace state.argument ∧ Claimed trace mode ∧
- Answered trace (.lease mode) (.lease true)
- | .account mode =>
- Verified trace state.argument ∧ Claimed trace mode ∧
- Answered trace (.lease mode) (.lease true) ∧
- Answered trace (.prefixGuard mode) (.prefixGuard true)
- | .denied mode index =>
- Verified trace state.argument ∧ Claimed trace mode ∧
- Answered trace (.lease mode) (.lease true) ∧
- Answered trace (.prefixGuard mode) (.prefixGuard true) ∧
- (state.plan.accountRead = true → Answered trace (.account mode) (.account .equal)) ∧
- ∀ earlier, earlier < index → Answered trace (.denied mode earlier) (.denied .refused)
- | .readBack mode =>
- Verified trace state.argument ∧ Claimed trace mode ∧
- CredentialsChecked trace state.plan mode
+ | .reload | .lease => Verified trace state.argument ∧ Claimed trace
+ | .prefixGuard =>
+ Verified trace state.argument ∧ Claimed trace ∧
+ Answered trace .lease (.lease true)
+ | .account =>
+ Verified trace state.argument ∧ Claimed trace ∧
+ Answered trace .lease (.lease true) ∧
+ Answered trace .prefixGuard (.prefixGuard true)
+ | .denied index =>
+ Verified trace state.argument ∧ Claimed trace ∧
+ Answered trace .lease (.lease true) ∧
+ Answered trace .prefixGuard (.prefixGuard true) ∧
+ (state.plan.accountRead = true → Answered trace .account (.account .equal)) ∧
+ ∀ earlier, earlier < index → Answered trace (.denied earlier) (.denied .refused)
| .preEntry =>
- Verified trace state.argument ∧ Claimed trace .entry ∧
- CredentialsChecked trace state.plan .entry
+ Verified trace state.argument ∧ Claimed trace ∧
+ CredentialsChecked trace state.plan
| .ceiling =>
- Verified trace state.argument ∧ Claimed trace .entry ∧
- CredentialsChecked trace state.plan .entry ∧
+ Verified trace state.argument ∧ Claimed trace ∧
+ CredentialsChecked trace state.plan ∧
(state.plan.preEntry = true → Answered trace .preEntry (.preEntry .satisfied))
| .checkpoint | .entryReload =>
- Verified trace state.argument ∧ Claimed trace .entry ∧
- CredentialsChecked trace state.plan .entry ∧
+ Verified trace state.argument ∧ Claimed trace ∧
+ CredentialsChecked trace state.plan ∧
ReadsPassed trace state.plan state.argument
- | .deadline | .send | .recordResponse | .recordUnknown =>
- Verified trace state.argument ∧ Claimed trace .entry ∧
- CredentialsChecked trace state.plan .entry ∧
+ | .deadline =>
+ Verified trace state.argument ∧ Claimed trace ∧
+ CredentialsChecked trace state.plan ∧
ReadsPassed trace state.plan state.argument ∧
Answered trace .entryReload (.reload true)
-
-theorem denied_extend {trace : List Step} {mode : Mode} {index bound : Nat}
- (prior : ∀ earlier, earlier < index → Answered trace (.denied mode earlier) (.denied .refused))
+ | .grant | .send | .recordResponse | .recordUnknown =>
+ Verified trace state.argument ∧ Claimed trace ∧
+ CredentialsChecked trace state.plan ∧ ReadsPassed trace state.plan state.argument ∧
+ Answered trace .entryReload (.reload true) ∧
+ Answered trace .deadline (.deadline true)
+
+theorem denied_extend {trace : List Step} {index bound : Nat}
+ (prior : ∀ earlier, earlier < index → Answered trace (.denied earlier) (.denied .refused))
(within : bound ≤ index + 1) :
∀ earlier, earlier < bound →
- Answered trace (.denied mode earlier) (.denied .refused) ∨ index = earlier := by
+ Answered trace (.denied earlier) (.denied .refused) ∨ index = earlier := by
intro earlier below
by_cases before : earlier < index
· exact Or.inl (prior earlier before)
@@ -622,9 +586,9 @@ theorem facts_step (trace : List Step) (state : State) (event : Event)
rcases state with ⟨plan, phase, argument⟩
cases phase <;> cases event <;>
simp only [nextStep, preClaimStep, leaseStep, entryStep, go, stop, halt, notEntered,
- leaseFailed, afterReads, ceilingFrom, preEntryFrom, afterCredentials, deniedFrom,
+ afterReads, ceilingFrom, preEntryFrom, deniedFrom,
afterPrefix, afterAdmission, preClaim, afterClaim, afterAccount, afterDenied, afterPreEntry,
- afterCeiling, afterWrite, afterResponse, afterReadBack] at facts ⊢ <;>
+ afterCeiling, afterWrite, afterResponse] at facts ⊢ <;>
(repeat' split) <;>
simp_all [Facts, Verified, Claimed, CredentialsChecked, CeilingAdmitted, ReadsPassed,
answered_append] <;>
@@ -648,15 +612,12 @@ def preClaimAction : Action → Bool
/-- The phases the machine can be in once it named the write. -/
def afterSend : Phase → Bool
- | .send | .recordResponse | .recordUnknown | .recordNotEntered | .done
- | .reload .readBack | .lease .readBack | .prefixGuard .readBack | .account .readBack
- | .denied .readBack _ | .readBack .readBack => true
+ | .send | .recordResponse | .recordUnknown | .done => true
| _ => false
-/-- The phases the machine can be in once a claim replayed. -/
+/-- A replay returns the retained record without another effect. -/
def replaying : Phase → Bool
- | .resume | .reload .reobserve | .lease .reobserve | .prefixGuard .reobserve
- | .account .reobserve | .denied .reobserve _ | .readBack .reobserve | .done => true
+ | .done => true
| _ => false
/-- The ordering invariant of a trace ending in `state`. -/
@@ -686,11 +647,11 @@ set_option hygiene false in
/-- Unfolds one step of the machine and closes each case of the case split. -/
macro "step_cases" : tactic => `(tactic| (
rcases state with ⟨plan, phase, argument⟩
- cases phase <;> (try cases ‹Mode›) <;> cases event <;> (try cases ‹Verification›) <;>
+ cases phase <;> cases event <;> (try cases ‹Verification›) <;>
simp only [nextStep, preClaimStep, leaseStep, entryStep, go, stop, halt, notEntered,
- leaseFailed, afterReads, ceilingFrom, preEntryFrom, afterCredentials, deniedFrom,
+ afterReads, ceilingFrom, preEntryFrom, deniedFrom,
afterPrefix, afterAdmission, preClaim, afterClaim, afterAccount, afterDenied, afterPreEntry,
- afterCeiling, afterWrite, afterResponse, afterReadBack] <;>
+ afterCeiling, afterWrite, afterResponse] <;>
(repeat' split) <;>
simp_all [argumentAfter, beforeVerification, beforeClaim, preClaimAction, afterSend,
replaying]))
@@ -713,7 +674,7 @@ theorem before_claim_step (state : State) (event : Event)
theorem send_step (state : State) (event : Event)
(sent : (nextStep state event).action = .send) :
- state.phase = .deadline ∧ event = .deadline true ∧ (nextStep state event).state.phase = .send := by
+ state.phase = .grant ∧ event = .granted true ∧ (nextStep state event).state.phase = .send := by
revert sent
step_cases
@@ -800,9 +761,9 @@ theorem ordered_step (trace : List Step) (state : State) (event : Event)
rw [phaseEq] at this
cases this
-theorem lease_named (state : State) (event : Event) (mode : Mode)
- (leased : (nextStep state event).action = .lease mode) :
- state.phase = .reload mode ∧ event = .reload true := by
+theorem lease_named (state : State) (event : Event)
+ (leased : (nextStep state event).action = .lease) :
+ state.phase = .reload ∧ event = .reload true := by
revert leased
step_cases
@@ -836,36 +797,31 @@ theorem invariant (plan : Plan) (events : List Event) :
/-! ## Ordering theorems -/
-/-- A credential lease is named only on a successful connection reload in
-the reload phase of its mode, after native verification authorized the
-action, and after the claim it depends on: an inserted claim for the entry
-and its read-back, a replayed claim for a re-observation. -/
+/-- A credential lease requires native authorization and a durable reservation. -/
theorem lease_requires_verified_claim (plan : Plan) (pre : List Event) (event : Event)
- (mode : Mode) (leased : (nextStep (stateAfter plan pre) event).action = .lease mode) :
- event = .reload true ∧ (stateAfter plan pre).phase = .reload mode ∧
+ (leased : (nextStep (stateAfter plan pre) event).action = .lease) :
+ event = .reload true ∧ (stateAfter plan pre).phase = .reload ∧
(∃ argument, Verified (steps plan pre) argument) ∧
- (mode = .reobserve → Answered (steps plan pre) .claim (.claim .replay)) ∧
- (mode ≠ .reobserve → Answered (steps plan pre) .claim (.claim .inserted)) := by
- obtain ⟨phase, eventEq⟩ := lease_named _ _ _ leased
+ Answered (steps plan pre) .claim (.claim .reserved) := by
+ obtain ⟨phase, eventEq⟩ := lease_named _ _ leased
have facts := (invariant plan pre).1
unfold Facts at facts
rw [phase] at facts
- obtain ⟨verified, claimed⟩ := facts
- refine ⟨eventEq, phase, ⟨_, verified⟩, ?_, ?_⟩ <;> intro modeEq <;> cases mode <;>
- simp_all [Claimed]
-
-/-- The write is named only on a successful deadline check, after the entry
-lease succeeded, after the declared pre-entry re-read was satisfied, after
-the connection reload before entry found the record unchanged, and after an
-inserted claim of a verified action. -/
+ exact ⟨eventEq, phase, ⟨_, facts.1⟩, facts.2⟩
+
+/-- The write is named only on a successful durable grant, after the deadline
+check and entry lease, after the declared pre-entry re-read was satisfied,
+after the connection reload found the record unchanged, and after a durable
+reservation of a verified action. -/
theorem send_requires_lease_and_deadline (plan : Plan) (pre : List Event) (event : Event)
(sent : (nextStep (stateAfter plan pre) event).action = .send) :
- event = .deadline true ∧ (stateAfter plan pre).phase = .deadline ∧
+ event = .granted true ∧ (stateAfter plan pre).phase = .grant ∧
(∃ argument, Verified (steps plan pre) argument) ∧
- Answered (steps plan pre) .claim (.claim .inserted) ∧
- Answered (steps plan pre) (.lease .entry) (.lease true) ∧
+ Answered (steps plan pre) .claim (.claim .reserved) ∧
+ Answered (steps plan pre) .lease (.lease true) ∧
(plan.preEntry = true → Answered (steps plan pre) .preEntry (.preEntry .satisfied)) ∧
- Answered (steps plan pre) .entryReload (.reload true) := by
+ Answered (steps plan pre) .entryReload (.reload true) ∧
+ Answered (steps plan pre) .deadline (.deadline true) := by
obtain ⟨phase, eventEq, _⟩ := send_step _ _ sent
have facts := (invariant plan pre).1
unfold Facts at facts
@@ -878,10 +834,10 @@ check: the prefix guard, the account read with equal commitments when
declared, and every declared denied read refused with a declared status. -/
theorem send_requires_credential_checks (plan : Plan) (pre : List Event) (event : Event)
(sent : (nextStep (stateAfter plan pre) event).action = .send) :
- Answered (steps plan pre) (.prefixGuard .entry) (.prefixGuard true) ∧
- (plan.accountRead = true → Answered (steps plan pre) (.account .entry) (.account .equal)) ∧
+ Answered (steps plan pre) .prefixGuard (.prefixGuard true) ∧
+ (plan.accountRead = true → Answered (steps plan pre) .account (.account .equal)) ∧
∀ index, index < plan.deniedReads →
- Answered (steps plan pre) (.denied .entry index) (.denied .refused) := by
+ Answered (steps plan pre) (.denied index) (.denied .refused) := by
obtain ⟨phase, _, _⟩ := send_step _ _ sent
have facts := (invariant plan pre).1
unfold Facts at facts
@@ -948,13 +904,13 @@ theorem pre_claim_refusal_stores_nothing (plan : Plan) (pre : List Event) (event
records, by phase and event; `none` for every other step. -/
def refusalOf (state : State) (event : Event) : Option Refusal :=
match state.phase, event with
- | .reload .entry, .reload false => some .connectionChanged
- | .lease .entry, .lease false => some .credentialUnavailable
- | .prefixGuard .entry, .prefixGuard false => some .modeGuard
- | .account .entry, .account .mismatch => some .accountMismatch
- | .account .entry, .account .unavailable => some .accountUnavailable
- | .denied .entry _, .denied .answered => some .capabilityExcess
- | .denied .entry _, .denied .unavailable => some .capabilityUnavailable
+ | .reload, .reload false => some .connectionChanged
+ | .lease, .lease false => some .credentialUnavailable
+ | .prefixGuard, .prefixGuard false => some .modeGuard
+ | .account, .account .mismatch => some .accountMismatch
+ | .account, .account .unavailable => some .accountUnavailable
+ | .denied _, .denied .answered => some .capabilityExcess
+ | .denied _, .denied .unavailable => some .capabilityUnavailable
| .preEntry, .preEntry .conditionFalse => some .preEntryConditionFalse
| .preEntry, .preEntry .unavailable => some .preEntryUnavailable
| .ceiling, .ceiling .unavailable => some .ceilingUnavailable
@@ -963,7 +919,6 @@ def refusalOf (state : State) (event : Event) : Option Refusal :=
if admits state.argument basis state.plan.basisPoints then none else some .ceilingAbove
| .entryReload, .reload false => some .connectionChanged
| .deadline, .deadline false => some .entryDeadline
- | .send, .write .notEntered => some .transportNotEntered
| _, _ => none
/-- The per-phase table: a step records `not-entered` with a refusal exactly
@@ -982,30 +937,29 @@ theorem not_entered_phase (state : State) (event : Event) (refusal : Refusal)
theorem refusal_claimed (trace : List Step) (state : State) (event : Event) (refusal : Refusal)
(facts : Facts trace state) (assigned : refusalOf state event = some refusal) :
- Answered trace .claim (.claim .inserted) := by
+ Answered trace .claim (.claim .reserved) := by
rcases state with ⟨plan, phase, argument⟩
- cases phase <;> (try cases ‹Mode›) <;> simp_all [Facts, Claimed, refusalOf]
+ cases phase <;> simp_all [Facts, Claimed, refusalOf]
theorem refusal_after_send (state : State) (event : Event) (refusal : Refusal)
(sent : afterSend state.phase = true) (assigned : refusalOf state event = some refusal) :
- refusal = .transportNotEntered := by
+ False := by
rcases state with ⟨plan, phase, argument⟩
- cases phase <;> (try cases ‹Mode›) <;> cases event <;> simp_all [afterSend, refusalOf] <;>
+ cases phase <;> cases event <;> simp_all [afterSend, refusalOf] <;>
(repeat' split at assigned) <;> simp_all
/-- Over every trace: a step records `not-entered` exactly when the claim
-was inserted and the step's phase and event are a failure the table assigns
+was reserved and the step's phase and event are a failure the table assigns
a refusal to, and the recorded refusal is that failure's. Once the write was
-named, the only such failure is the transport's refusal before network
-entry. -/
+named, no such refusal can be emitted. -/
theorem post_claim_refusal_records_not_entered (plan : Plan) (pre : List Event)
(event : Event) (refusal : Refusal) :
((nextStep (stateAfter plan pre) event).action = .recordNotEntered refusal ↔
- Answered (steps plan pre) .claim (.claim .inserted) ∧
+ Answered (steps plan pre) .claim (.claim .reserved) ∧
refusalOf (stateAfter plan pre) event = some refusal) ∧
((nextStep (stateAfter plan pre) event).action = .recordNotEntered refusal →
(nextStep (stateAfter plan pre) event).state.phase = .recordNotEntered ∧
- (Emitted (steps plan pre) .send → refusal = .transportNotEntered)) := by
+ ¬ Emitted (steps plan pre) .send) := by
obtain ⟨facts, _, _, _, _, sent, _⟩ := invariant plan pre
refine ⟨⟨fun recorded => ?_, fun ⟨_, assigned⟩ => ?_⟩, fun recorded => ?_⟩
· have assigned := (not_entered_refusal_table _ _ _).mp recorded
@@ -1014,4 +968,28 @@ theorem post_claim_refusal_records_not_entered (plan : Plan) (pre : List Event)
· refine ⟨not_entered_phase _ _ _ recorded, fun emitted => ?_⟩
exact refusal_after_send _ _ _ (sent emitted) ((not_entered_refusal_table _ _ _).mp recorded)
+/-- A recorded response is a terminal submission result. Observation requires a
+separate authorized operation; the submission machine emits no follow-up read. -/
+theorem response_stops_without_read (state : State) (phase : state.phase = .recordResponse) :
+ nextStep state (.response .recorded) = stop state .response := by
+ rcases state with ⟨plan, phase', argument⟩
+ cases phase
+ rfl
+
+/-- Losing or replaying an existing reservation returns its retained record;
+it cannot resume credential acquisition or issue another entry grant. -/
+theorem replay_returns_stored (state : State) (phase : state.phase = .claim) :
+ nextStep state (.claim .replay) = stop state .stored := by
+ rcases state with ⟨plan, phase', argument⟩
+ cases phase
+ rfl
+
+/-- A failed durable grant stops without sending, and never fabricates a
+not-entered refund after the authority boundary. -/
+theorem refused_grant_stops_without_send (state : State) (phase : state.phase = .grant) :
+ nextStep state (.granted false) = stop state .grantUnavailable := by
+ rcases state with ⟨plan, phase', argument⟩
+ cases phase
+ rfl
+
end Auths.Product.SubmitOrder
diff --git a/formal/Auths/Theorems.lean b/formal/Auths/Theorems.lean
index ede497928..5c390bcdb 100644
--- a/formal/Auths/Theorems.lean
+++ b/formal/Auths/Theorems.lean
@@ -182,6 +182,10 @@ def theoremInventory : List Lean.Name :=
`Auths.Product.Recovery.class_matches_declarations,
`Auths.Product.Recovery.write_never_conditional,
`Auths.Product.Recovery.unknown_resolves_only_linked,
+ `Auths.Product.Recovery.unknown_exits_with_bound_evidence,
+ `Auths.Product.Recovery.transition_ownership,
+ `Auths.Product.Recovery.late_response_retains_source,
+ `Auths.Product.Recovery.late_acknowledgement_retains_source,
`Auths.Product.Recovery.provider_claim_requires_evidence,
`Auths.Product.Recovery.terminal_stages_final,
`Auths.Product.Recovery.transitions_preserve_identity,
@@ -191,6 +195,9 @@ def theoremInventory : List Lean.Name :=
`Auths.Product.SubmitOrder.send_requires_relative_ceiling,
`Auths.Product.SubmitOrder.account_scope_requires_binding,
`Auths.Product.SubmitOrder.at_most_one_send,
+ `Auths.Product.SubmitOrder.response_stops_without_read,
+ `Auths.Product.SubmitOrder.replay_returns_stored,
+ `Auths.Product.SubmitOrder.refused_grant_stops_without_send,
`Auths.Product.SubmitOrder.pre_claim_refusal_stores_nothing,
`Auths.Product.SubmitOrder.not_entered_refusal_table,
`Auths.Product.SubmitOrder.post_claim_refusal_records_not_entered,
diff --git a/formal/README.md b/formal/README.md
index c94b8ae7d..6254e1fcd 100644
--- a/formal/README.md
+++ b/formal/README.md
@@ -125,3 +125,52 @@ the translation, toolchain, source, and evidence inputs still match. Changes
to those inputs require two fresh, byte-identical reproductions. The complete
qualified path remains in hosted CI; contributors do not need to install the
formal toolchains locally to receive proof diagnostics.
+
+## AP074 gateway correspondence review
+
+The AP074 repair replaces the earlier implicit-read submission model with the
+shipping `order::next_step` machine. A successful deadline check names durable
+grant creation; only `Granted(true)` names `Send`. A replay stops with `Stored`,
+and recording a response stops with `Response`. The ordering proofs retain
+native verification, durable reservation, credential checks, pre-entry reads,
+relative ceilings, and at most one send. A post-send path cannot produce a
+refunding `NotEntered` result.
+
+The attempt projection now includes its protocol, bounded revision, entry grant,
+and recipient acknowledgement. `translated_valid_transition_refines_model`
+proves exact correspondence for every pair of translated views, including
+revision-overflow refusal. The natural model proves protocol retention,
+revision advancement, grant retention, and terminal-stage finality. Its unknown
+exit table distinguishes three evidence routes:
+
+- A late source response retains the original grant and records response facts,
+ without asserting provider observation.
+- A late recipient acknowledgement is restricted to disclosure, retains the
+ original grant, and cannot claim a provider response or observation.
+- An independently authorized linked observation can establish
+ `ObservedByProvider` from unknown only with the retained verified-locator link
+ and recorded provider evidence.
+
+The first two routes have separate `late_response_retains_source` and
+`late_acknowledgement_retains_source` theorems. The qualified unknown-observation
+claim is deliberately conditional on the target being `ObservedByProvider`;
+`unknown_exits_with_bound_evidence` proves the complete exit table. The old
+statement that *every* unknown exit must be provider observation is no longer
+true of the shipping contract and is not advertised as such.
+
+These are proofs about the decision leaves and canonical byte projections.
+They do not prove signature verification, exact recipient/challenge binding,
+source provenance, store atomicity, or external I/O. The runtime must establish
+those facts before constructing the projection and must authorize recovery as a
+separate operation. Nonempty acknowledgement bytes in the model are not a
+substitute for the runtime's authentication checks. Process, store, transport,
+and cryptographic tests remain required alongside these proofs.
+
+The reviewed gateway translation inventory is 126 local functions with zero
+external or opaque local functions. AP074 adds `acknowledgement_kept`,
+`acknowledgement_stage`, `entry_kept`, `protocol_consistent`, `requires_entry`,
+and `same_protocol`; all are closed transition checks covered by refinement.
+It removes `after_credentials`, `after_read_back`, and `lease_failed`, which
+implemented the obsolete mode-dependent implicit-read route. Regeneration uses
+the unchanged pinned Charon/Aeneas/Lean toolchains and two clean translations;
+the inventory change is not an exception to the translation or axiom gates.
diff --git a/formal/assurance-manifest-v1.toml b/formal/assurance-manifest-v1.toml
index ee29fb643..37d89c524 100644
--- a/formal/assurance-manifest-v1.toml
+++ b/formal/assurance-manifest-v1.toml
@@ -1,8 +1,8 @@
schema = "auths-proof-formal-assurance/v1"
lean_toolchain = "leanprover/lean4:v4.31.0"
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
-statement_review = "native-k-of-n-approvals-2026-09-30"
-statement_inventory_sha256 = "717939bc5edda0732568a53e598d571b4c3e7584267b23dd1536de29ca0b9838"
+statement_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
+statement_inventory_sha256 = "09e7c08798aa84244d364c0de8c4d5258978f57ef5a4a62f6b1409e00540db40"
allowed_axioms = [
"Classical.choice",
"Quot.sound",
@@ -15,7 +15,7 @@ claim_text = "Lean proves the rich authority property: finiteSet subset refl."
claim_status = "proved"
lean_declaration = "Auths.Rich.finiteSet_subset_refl"
lean_statement_sha256 = "cac9d4b80390ce990484b56aee1f0d20fcb8d1b2b2b8f07ef1fa819caca6982c"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -77,7 +77,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -97,7 +97,7 @@ claim_text = "Lean proves the rich authority property: finiteSet subset trans."
claim_status = "proved"
lean_declaration = "Auths.Rich.finiteSet_subset_trans"
lean_statement_sha256 = "620ac85f39efb4108727cc41c2b3b7663eea53ffde7ef5d782ee2e5897e161da"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -159,7 +159,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -179,7 +179,7 @@ claim_text = "Lean proves the rich authority property: finiteSet subset antisymm
claim_status = "proved"
lean_declaration = "Auths.Rich.finiteSet_subset_antisymm"
lean_statement_sha256 = "04c8f87dfb0552e7abf656cf19ca4e29a5a81369e1178eb7d9aa634966a115be"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -241,7 +241,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -261,7 +261,7 @@ claim_text = "Lean proves the rich authority property: finiteSet membership mono
claim_status = "proved"
lean_declaration = "Auths.Rich.finiteSet_membership_monotone"
lean_statement_sha256 = "526bdf55ccea3aa1a07aef06197ef303b18a5c083994c97056422713106028a2"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -323,7 +323,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -343,7 +343,7 @@ claim_text = "Lean proves the rich authority property: window contained refl."
claim_status = "proved"
lean_declaration = "Auths.Rich.window_contained_refl"
lean_statement_sha256 = "52b61c239b76a3ec75ebc375c19ba060e170f9c88ab845652188da7e1f32c715"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -405,7 +405,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -422,7 +422,7 @@ claim_text = "Lean proves the rich authority property: window contained trans."
claim_status = "proved"
lean_declaration = "Auths.Rich.window_contained_trans"
lean_statement_sha256 = "ed63f2b8060617d8e042f212bea8d9df13b58494cd94ef01081d7c849e19de7b"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -484,7 +484,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -505,7 +505,7 @@ claim_text = "Lean proves the rich authority property: window contained antisymm
claim_status = "proved"
lean_declaration = "Auths.Rich.window_contained_antisymm"
lean_statement_sha256 = "a263d89be575fd98940d6a3324ec862360cfe26b2024dd38a13b828b763aa74a"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -567,7 +567,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -588,7 +588,7 @@ claim_text = "Lean proves the rich authority property: window coverage monotone.
claim_status = "proved"
lean_declaration = "Auths.Rich.window_coverage_monotone"
lean_statement_sha256 = "e8d8a9503bde767273a9b55883a18365323b88ae798ebd4de744ffd822ee9dd7"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -650,7 +650,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -671,7 +671,7 @@ claim_text = "Lean proves the rich authority property: action constraint refl."
claim_status = "proved"
lean_declaration = "Auths.Rich.action_constraint_refl"
lean_statement_sha256 = "6b41a97add7b56d69dde91e7008ca41e34f568ca4081d8b81ca5b728342e5c1d"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -733,7 +733,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -753,7 +753,7 @@ claim_text = "Lean proves the rich authority property: action constraint trans."
claim_status = "proved"
lean_declaration = "Auths.Rich.action_constraint_trans"
lean_statement_sha256 = "1510dabf7d9cf6d4c1bdb99de295ced7f32f3c44678519aef38aa3fb58dcdee0"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -815,7 +815,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -836,7 +836,7 @@ claim_text = "Lean proves the rich authority property: action constraint antisym
claim_status = "proved"
lean_declaration = "Auths.Rich.action_constraint_antisymm"
lean_statement_sha256 = "59a7fb2812568c3a9eabd67248452de770baf4d2e1daf9ebf2dbea0baa42c81d"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -898,7 +898,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -919,7 +919,7 @@ claim_text = "Lean proves the rich authority property: action constraint canonic
claim_status = "proved"
lean_declaration = "Auths.Rich.action_constraint_canonical_antisymm"
lean_statement_sha256 = "47d8802981c338d34dedf567e68555c3cca9f298dd13c681ba4b3baa927a6813"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -981,7 +981,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -1002,7 +1002,7 @@ claim_text = "Lean proves the rich authority property: action constraint allows
claim_status = "proved"
lean_declaration = "Auths.Rich.action_constraint_allows_monotone"
lean_statement_sha256 = "49927a15dec0e627c64244b981bd81ca171589f5ada904616bcec9811746efe2"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -1064,7 +1064,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -1085,7 +1085,7 @@ claim_text = "Lean proves the rich authority property: budget refl."
claim_status = "proved"
lean_declaration = "Auths.Rich.budget_refl"
lean_statement_sha256 = "54ffb1f592e48a7292cb925afd2f0e1c4c2412a46b7b5d8aeec7113fe58fc09a"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -1147,7 +1147,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -1164,7 +1164,7 @@ claim_text = "Lean proves the rich authority property: budget trans."
claim_status = "proved"
lean_declaration = "Auths.Rich.budget_trans"
lean_statement_sha256 = "313fe44a88fa3efd9bc1800bf0e6169039c5bcec051b847ea86c71d27318e0ae"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -1226,7 +1226,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -1243,7 +1243,7 @@ claim_text = "Lean proves the rich authority property: budget antisymm."
claim_status = "proved"
lean_declaration = "Auths.Rich.budget_antisymm"
lean_statement_sha256 = "569b03bd90e823b38193e8a9681faf27341786455644132e4861c779c8c3395b"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -1305,7 +1305,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -1322,7 +1322,7 @@ claim_text = "Lean proves the rich authority property: budget coverage monotone.
claim_status = "proved"
lean_declaration = "Auths.Rich.budget_coverage_monotone"
lean_statement_sha256 = "594cf3b8740d3b68643756cf665f0be72b2b5b010227f5efc0407ce78fc91520"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -1384,7 +1384,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -1401,7 +1401,7 @@ claim_text = "Lean proves the rich authority property: budget action coverage mo
claim_status = "proved"
lean_declaration = "Auths.Rich.budget_action_coverage_monotone"
lean_statement_sha256 = "986470888869bfa94623fd5d5f4a7a72820170912004d0710b39127565ab0379"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -1463,7 +1463,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -1480,7 +1480,7 @@ claim_text = "An expressible profile with no budget request under a bounded ceil
claim_status = "proved"
lean_declaration = "Auths.Rich.budgetCoversAction_expressible_absent_bounded"
lean_statement_sha256 = "9b30941cc8a5c3b3aa4c70e4703fb0e902ed07d7d0f5d07536f087234423a0b4"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -1542,7 +1542,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 budget coverage over opaque identity carriers."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -1559,7 +1559,7 @@ claim_text = "Lean proves the rich authority property: extensions refl."
claim_status = "proved"
lean_declaration = "Auths.Rich.extensions_refl"
lean_statement_sha256 = "3deb6e01a441a7e322859de8eadaec18124a1cb370b09a38322fb4e17dfcb4cf"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -1621,7 +1621,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -1638,7 +1638,7 @@ claim_text = "Lean proves the rich authority property: extensions trans."
claim_status = "proved"
lean_declaration = "Auths.Rich.extensions_trans"
lean_statement_sha256 = "90aa05bf5d3bdaa2f0a8df2b7738628d66fdc848f306b392a015a2e1e12ba9e4"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -1700,7 +1700,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -1717,7 +1717,7 @@ claim_text = "Lean proves the rich authority property: extensions le pinned iff.
claim_status = "proved"
lean_declaration = "Auths.Rich.extensions_le_pinned_iff"
lean_statement_sha256 = "98fe643cba56dcda0d75b3dc27de128995feff749567004d74f0bf6a8ecf7e37"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -1779,7 +1779,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -1796,7 +1796,7 @@ claim_text = "Lean proves the rich authority property: extensions le false of dr
claim_status = "proved"
lean_declaration = "Auths.Rich.extensions_le_false_of_dropped"
lean_statement_sha256 = "46a605cf7dea5af58e69b6f0bbc4ffed2ed34084cda7ea99b680ca975ac47f3a"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -1858,7 +1858,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -1875,7 +1875,7 @@ claim_text = "Lean proves a child that strips a pinned critical-extension identi
claim_status = "proved"
lean_declaration = "Auths.Rich.extensions_le_false_of_stripped"
lean_statement_sha256 = "92f064c6de7f9415d2483d26e88473b421a0b266b431f939de9f4178952ef6cd"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -1937,7 +1937,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -1954,7 +1954,7 @@ claim_text = "Lean proves a child whose payload for a pinned identifier its law
claim_status = "proved"
lean_declaration = "Auths.Rich.extensions_le_false_of_refused"
lean_statement_sha256 = "4639baf3f61027f8833b78149c26c0af128dc39852f0360464e854e34acb944d"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -2016,7 +2016,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -2033,7 +2033,7 @@ claim_text = "Lean proves a child that adds an identifier whose law refuses addi
claim_status = "proved"
lean_declaration = "Auths.Rich.extensions_le_false_of_refused_addition"
lean_statement_sha256 = "7c57bc9d5691c1cb44f74f95653c02818b7a982cc8ed01c05d43aaed876b38a6"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -2095,7 +2095,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -2112,7 +2112,7 @@ claim_text = "Lean proves the rich authority property: extensions le refuses a d
claim_status = "proved"
lean_declaration = "Auths.Rich.extensions_le_refuses_a_dropped_singleton"
lean_statement_sha256 = "b0e6555b3ea818523fe6e8df539811645a6a7ee61e8a03c464b8852fb184829a"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -2174,7 +2174,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -2194,7 +2194,7 @@ claim_text = "Lean proves per-identifier extension attenuation admits no world t
claim_status = "proved"
lean_declaration = "Auths.Rich.extensions_le_narrows"
lean_statement_sha256 = "b84fc2cf70c513867276c1398608b399a6e3ca9f2c01905b3c339810dd1496d8"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -2256,7 +2256,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -2273,7 +2273,7 @@ claim_text = "Lean proves the rich authority property: status refl."
claim_status = "proved"
lean_declaration = "Auths.Rich.status_refl"
lean_statement_sha256 = "7d5310e68d00575620e99f894080ddc254f48770537a07e1a7e4acdae0090d63"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -2335,7 +2335,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -2352,7 +2352,7 @@ claim_text = "Lean proves the rich authority property: status trans."
claim_status = "proved"
lean_declaration = "Auths.Rich.status_trans"
lean_statement_sha256 = "7e9d6580025807c2b2c984e8b167029b84ac610f9dcd280179e4c176872a04a1"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -2414,7 +2414,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -2431,7 +2431,7 @@ claim_text = "Lean proves the rich authority property: status antisymm."
claim_status = "proved"
lean_declaration = "Auths.Rich.status_antisymm"
lean_statement_sha256 = "a3badc32c1fcd22da760494c3c6eb256e637312f34fa7691591e7283f504867e"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -2493,7 +2493,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -2513,7 +2513,7 @@ claim_text = "Lean proves the rich authority property: status satisfaction monot
claim_status = "proved"
lean_declaration = "Auths.Rich.status_satisfaction_monotone"
lean_statement_sha256 = "7f5f3cfc2e90f4a2bdf90499b6b596a0711d27f3668d3d99bcf8c5a0a146e022"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -2575,7 +2575,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -2592,7 +2592,7 @@ claim_text = "Lean proves the rich authority property: profile refl."
claim_status = "proved"
lean_declaration = "Auths.Rich.profile_refl"
lean_statement_sha256 = "c256b640094e0458a6cb336c78c358b8f35906d7838c86142f649c4d7c3cf9ea"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -2654,7 +2654,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -2674,7 +2674,7 @@ claim_text = "Lean proves the rich authority property: profile trans."
claim_status = "proved"
lean_declaration = "Auths.Rich.profile_trans"
lean_statement_sha256 = "bc702fae63c302ea1b30323b66eaa633f2c489dd14a84bda0802852888da4e5d"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -2736,7 +2736,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -2756,7 +2756,7 @@ claim_text = "Lean proves the rich authority property: profile antisymm."
claim_status = "proved"
lean_declaration = "Auths.Rich.profile_antisymm"
lean_statement_sha256 = "306251093aa948294a625b5fd0e2707c52d9d391f172185ca36949ee0bd022ba"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -2818,7 +2818,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -2838,7 +2838,7 @@ claim_text = "Lean proves the rich authority property: profile coverage monotone
claim_status = "proved"
lean_declaration = "Auths.Rich.profile_coverage_monotone"
lean_statement_sha256 = "fa6e1ec9008031df18e845bb6457030c5169bff159a30e5a30e4be9d7b445c16"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -2900,7 +2900,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -2920,7 +2920,7 @@ claim_text = "Lean proves the rich authority property: structural scope le refl.
claim_status = "proved"
lean_declaration = "Auths.Rich.structural_scope_le_refl"
lean_statement_sha256 = "d62dc8e9720475e1f9aef453def070f12382d75b5d1836e28810c50a123fe1b4"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -2982,7 +2982,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -3002,7 +3002,7 @@ claim_text = "Lean proves the rich authority property: structural scope le trans
claim_status = "proved"
lean_declaration = "Auths.Rich.structural_scope_le_trans"
lean_statement_sha256 = "6cbce72d938dea6cd1a9f4814ba78cb52e8c5f4c7cddd2469a6526bb1662bb94"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -3064,7 +3064,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -3085,7 +3085,7 @@ claim_text = "Lean proves the rich authority property: scope le canonical antisy
claim_status = "proved"
lean_declaration = "Auths.Rich.scope_le_canonical_antisymmetry"
lean_statement_sha256 = "197fe00cdc067ce9188485dd0e9b76f9561669b95ca225396561d8ce855f4542"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -3147,7 +3147,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -3168,7 +3168,7 @@ claim_text = "Lean proves the rich authority property: action coverage downward
claim_status = "proved"
lean_declaration = "Auths.Rich.action_coverage_downward_closed"
lean_statement_sha256 = "f76fefd95f4dae279b6fcfbe014a5e8fdd7a9a7c1641f42c19045b7b16a9c486"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -3230,7 +3230,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -3251,7 +3251,7 @@ claim_text = "Lean proves the rich authority property: evidence requirements dow
claim_status = "proved"
lean_declaration = "Auths.Rich.evidence_requirements_downward_closed"
lean_statement_sha256 = "3fca86b2826d0c6807da5e6d6c742bb51084ff7bd9914f1cc40003a0eaae08b4"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -3313,7 +3313,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -3333,7 +3333,7 @@ claim_text = "Lean proves the rich authority property: complete admission downwa
claim_status = "proved"
lean_declaration = "Auths.Rich.complete_admission_downward_closed"
lean_statement_sha256 = "4894f6f4a2eb7bef73ad797ae1601d7c07994e4d1bff74e873b8f1df4ba6657b"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -3395,7 +3395,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -3416,7 +3416,7 @@ claim_text = "Lean proves the rich authority property: semantic attenuation preo
claim_status = "proved"
lean_declaration = "Auths.Rich.semantic_attenuation_preorder_refl"
lean_statement_sha256 = "7ae1373433d76ff6136509d9662dab1b8e34867c4b4b067f1242499fb06ba0fd"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -3478,7 +3478,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -3498,7 +3498,7 @@ claim_text = "Lean proves the rich authority property: semantic attenuation preo
claim_status = "proved"
lean_declaration = "Auths.Rich.semantic_attenuation_preorder_trans"
lean_statement_sha256 = "9b916eba63a367eacf049d8c8cb710908e7ca06d6e624f1b3d474bf6cb3fced9"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -3560,7 +3560,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -3580,7 +3580,7 @@ claim_text = "Lean proves the rich authority property: structural scope le impli
claim_status = "proved"
lean_declaration = "Auths.Rich.structural_scope_le_implies_semantic_attenuation"
lean_statement_sha256 = "5218e5c6169c0fc0af43cb425d0a837930029032da4962d4168fd025f932d431"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -3642,7 +3642,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -3663,7 +3663,7 @@ claim_text = "Lean proves the rich authority property: scope semantic equivalenc
claim_status = "proved"
lean_declaration = "Auths.Rich.scope_semantic_equivalence"
lean_statement_sha256 = "ab0d42af5b5cf85e211b3660953239b56f008cd416570dd57f51d3d0c60175d2"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -3725,7 +3725,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -3745,7 +3745,7 @@ claim_text = "Lean proves the rich authority property: structural scope le decid
claim_status = "proved"
lean_declaration = "Auths.Rich.structural_scope_le_decides_declared_v1_relation"
lean_statement_sha256 = "d21c9865f8d29bd1fef0778402eddb386af93dcea402725cc09a63a41c590529"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -3807,7 +3807,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -3828,7 +3828,7 @@ claim_text = "Lean proves the rich authority property: delegate implies scope le
claim_status = "proved"
lean_declaration = "Auths.Rich.delegate_implies_scope_le"
lean_statement_sha256 = "492a39565c693001c3f03fbb40573d3d3c6f08ddacf36e7b5f996ab04df0109f"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -3890,7 +3890,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -3910,7 +3910,7 @@ claim_text = "Lean proves the rich authority property: delegate preserves root."
claim_status = "proved"
lean_declaration = "Auths.Rich.delegate_preserves_root"
lean_statement_sha256 = "6b5a9d3d8a41d6753cf6787bd5df8afeffa2c2679fa0217b2fdfa59d8a1f46b2"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -3972,7 +3972,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -3992,7 +3992,7 @@ claim_text = "Lean proves the rich authority property: delegate updates subject
claim_status = "proved"
lean_declaration = "Auths.Rich.delegate_updates_subject_and_parent"
lean_statement_sha256 = "73aea20a45df311686451843eb0a05632aee895859fd5b12a91d950b9f30b581"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -4054,7 +4054,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -4074,7 +4074,7 @@ claim_text = "Lean proves the rich authority property: anchored chain preserves
claim_status = "proved"
lean_declaration = "Auths.Rich.anchored_chain_preserves_provenance"
lean_statement_sha256 = "854d2853a17bef32a429e1da533bb6ea9dec3428bf7a6649fc0dc5711331b9ee"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -4136,7 +4136,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -4156,7 +4156,7 @@ claim_text = "Lean proves the rich authority property: singleton context selects
claim_status = "proved"
lean_declaration = "Auths.Rich.singleton_context_selects_fresh_self_rooted_start"
lean_statement_sha256 = "f110660181e70518b73b5360585a5eab8dad1d5ec2af11ac92b85c15569aabc8"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -4218,7 +4218,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -4238,7 +4238,7 @@ claim_text = "Lean proves the rich authority property: chain descends from one r
claim_status = "proved"
lean_declaration = "Auths.Authority.chain_descends_from_one_root"
lean_statement_sha256 = "854d2853a17bef32a429e1da533bb6ea9dec3428bf7a6649fc0dc5711331b9ee"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -4300,7 +4300,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -4320,7 +4320,7 @@ claim_text = "Lean proves the rich authority property: delegate strict depth."
claim_status = "proved"
lean_declaration = "Auths.Rich.delegate_strict_depth"
lean_statement_sha256 = "996384a43e290522990e26bb761fc181d37e551f3b2c11190e7ef4a55cd5a930"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -4382,7 +4382,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -4402,7 +4402,7 @@ claim_text = "Lean proves the rich authority property: delegate never widens."
claim_status = "proved"
lean_declaration = "Auths.Rich.delegate_never_widens"
lean_statement_sha256 = "492a39565c693001c3f03fbb40573d3d3c6f08ddacf36e7b5f996ab04df0109f"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -4464,7 +4464,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -4484,7 +4484,7 @@ claim_text = "Lean proves the rich authority property: remaining depth well foun
claim_status = "proved"
lean_declaration = "Auths.Rich.remaining_depth_well_founded"
lean_statement_sha256 = "dcf9ed58cbeb7f0b6d9c7fa332d1b3f696f0114cee5ab770d2ad31b27d8698eb"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -4546,7 +4546,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -4566,7 +4566,7 @@ claim_text = "Lean proves the rich authority property: finite delegation chain."
claim_status = "proved"
lean_declaration = "Auths.Rich.finite_delegation_chain"
lean_statement_sha256 = "c54457dda6f3f87e19a7224a424211e04f7da7706c7a6b950084ba630f0741f5"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -4628,7 +4628,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -4648,7 +4648,7 @@ claim_text = "Lean proves the rich authority property: chain transitive attenuat
claim_status = "proved"
lean_declaration = "Auths.Rich.chain_transitive_attenuation"
lean_statement_sha256 = "941e203039eef42df31f5c80208d2de0ac1345ff01f73b04ce139c9d57995c83"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -4710,7 +4710,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -4731,7 +4731,7 @@ claim_text = "Lean proves the rich authority property: authorized action covered
claim_status = "proved"
lean_declaration = "Auths.Rich.authorized_action_covered"
lean_statement_sha256 = "7a3a3213af6f0aed9678912e472e816a9bbfb0f9b0040452f3fcde79d32152b0"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -4793,7 +4793,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -4814,7 +4814,7 @@ claim_text = "Lean proves the rich authority property: extensions dimension is e
claim_status = "proved"
lean_declaration = "Auths.Rich.extensions_dimension_is_exact"
lean_statement_sha256 = "ade313b34026661e080eb2b6ae656b39acb694efc597028eb0352fcc3c119bb8"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -4876,7 +4876,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -4897,7 +4897,7 @@ claim_text = "Lean proves the generated extension dimension is false for a grant
claim_status = "proved"
lean_declaration = "Auths.Rich.extensions_dimension_false_of_stripped"
lean_statement_sha256 = "978bd5383ecf2efee06ce1391d5355821d353f4e454cbe15d8a0bec2d925aea2"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -4959,7 +4959,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -4980,7 +4980,7 @@ claim_text = "Lean proves the rich authority property: altered extensions deny e
claim_status = "proved"
lean_declaration = "Auths.Rich.altered_extensions_deny_every_projection"
lean_statement_sha256 = "02f306f478ccb5919a689482ba1e101e9723e5d365d3f7bd4c71d303a37c5383"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -5042,7 +5042,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -5063,7 +5063,7 @@ claim_text = "Lean proves every accepted edge carries the grant's extension set,
claim_status = "proved"
lean_declaration = "Auths.Rich.delegate_attenuates_pinned_extensions"
lean_statement_sha256 = "24acb38fc7066db51bd66b9e07dd5b6376010f969e2466acd191d080bbf01040"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -5125,7 +5125,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -5145,7 +5145,7 @@ claim_text = "Lean proves every state reachable along a delegation chain carries
claim_status = "proved"
lean_declaration = "Auths.Rich.chain_retains_pinned_extension_identifiers"
lean_statement_sha256 = "cd2cd919349232f6b3b498154eae98ab6070edd0936d891646946968b3cc6ac4"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -5207,7 +5207,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -5227,7 +5227,7 @@ claim_text = "Lean proves an accepted delegation never widens authority, includi
claim_status = "proved"
lean_declaration = "Auths.Rich.delegate_never_widens_authority"
lean_statement_sha256 = "459fd80f79b0474a17b23f31d1ea894170a59be8af9ca849fa17b930850855f5"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -5289,7 +5289,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -5310,7 +5310,7 @@ claim_text = "Lean proves no state reachable along a delegation chain admits an
claim_status = "proved"
lean_declaration = "Auths.Rich.chain_never_widens_authority"
lean_statement_sha256 = "badba34df0ad02950772de9a16a08ed704b00d13161ca3adb9690156fa9528d9"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -5372,7 +5372,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -5393,7 +5393,7 @@ claim_text = "Lean proves the rich authority property: rich projection accepts i
claim_status = "proved"
lean_declaration = "Auths.Rich.rich_projection_accepts_iff_root_and_scope_depth_checks"
lean_statement_sha256 = "8701007d8b4c42f2059cd7e02a841fe632ae4380470f3b2a396d0702e8371f46"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -5455,7 +5455,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -5476,7 +5476,7 @@ claim_text = "Lean proves the rich authority property: apply grant success iff l
claim_status = "proved"
lean_declaration = "Auths.Rich.apply_grant_success_iff_linked_and_projection"
lean_statement_sha256 = "d1e837214c3e2d92a11750a81de8f7a132d3615cd87b3d826e3952408a44d1b5"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -5538,7 +5538,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -5559,7 +5559,7 @@ claim_text = "Lean proves the rich authority property: apply grant success iff d
claim_status = "proved"
lean_declaration = "Auths.Rich.apply_grant_success_iff_delegates"
lean_statement_sha256 = "60adbaa9a7b69216f9a63e2d01e0d46c696e2e5c06b47a6b4b4a0bdac13e6550"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -5621,7 +5621,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -5642,7 +5642,7 @@ claim_text = "Lean proves the rich authority property: apply grant success uniqu
claim_status = "proved"
lean_declaration = "Auths.Rich.apply_grant_success_unique"
lean_statement_sha256 = "e5f71557dd61f8b0a5479564c6418fd399bf7ab28dc31f18d2fcb8acd3ee169d"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -5704,7 +5704,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -5725,7 +5725,7 @@ claim_text = "Lean proves the rich authority property: authority delegate diagno
claim_status = "proved"
lean_declaration = "Auths.Rich.authority_delegate_diagnostic_sound_complete"
lean_statement_sha256 = "1b03026088d20c617111ac18f843d1a98eff2a0d88aee42703c1f9bbfbfc6202"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -5787,7 +5787,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -5808,7 +5808,7 @@ claim_text = "Lean proves the rich authority property: authority delegate first
claim_status = "proved"
lean_declaration = "Auths.Rich.authority_delegate_first_failure"
lean_statement_sha256 = "4520e0f85b7efe3d629a9e1beaf601ea16ba7d9728759bab9dfbd2f208ddffcf"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -5870,7 +5870,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -5891,7 +5891,7 @@ claim_text = "Lean proves the rich authority property: author planning diagnosti
claim_status = "proved"
lean_declaration = "Auths.Rich.author_planning_diagnostic_sound_complete"
lean_statement_sha256 = "b6e9eff61245c569032bf56f08d2e4024342839385feca01eec87393f6fb6e58"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -5953,7 +5953,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -5974,7 +5974,7 @@ claim_text = "Lean proves the rich authority property: coverage decision ok iff
claim_status = "proved"
lean_declaration = "Auths.Rich.coverage_decision_ok_iff_covers"
lean_statement_sha256 = "bba6affe418888d125471704869356f8eda4861176df0aa5e75b72e873bb4a31"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -6036,7 +6036,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -6056,7 +6056,7 @@ claim_text = "Lean proves the rich authority property: rooted of origin."
claim_status = "proved"
lean_declaration = "Auths.Rich.rooted_of_origin"
lean_statement_sha256 = "273f922682334f0d8d4d60194e8517e3c842a615da43e0ec829e231893fbed43"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -6118,7 +6118,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -6138,7 +6138,7 @@ claim_text = "Lean proves the rich authority property: reachable accepted."
claim_status = "proved"
lean_declaration = "Auths.Rich.reachable_accepted"
lean_statement_sha256 = "9072a66c7332af7b2d27fea1b83fae0893fc14f9158106ce72fd5c7fb60cce48"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -6200,7 +6200,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -6220,7 +6220,7 @@ claim_text = "Lean proves the rich authority property: rooted of reachable."
claim_status = "proved"
lean_declaration = "Auths.Rich.rooted_of_reachable"
lean_statement_sha256 = "ff692ede01812e450cc8984ab3fef696b0e1d71fb93086a53dd1117c27eefdd9"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -6282,7 +6282,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -6302,7 +6302,7 @@ claim_text = "Lean proves the rich authority property: registry coverage decisio
claim_status = "proved"
lean_declaration = "Auths.Rich.registry_coverage_decision_ok_iff_covers"
lean_statement_sha256 = "75c5f00e47dfec8a229951c0220026fd984afd3ffce7b03b32f61bcc8a50672c"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -6364,7 +6364,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -6384,7 +6384,7 @@ claim_text = "Lean proves the rich authority property: extensions not forgeable.
claim_status = "proved"
lean_declaration = "Auths.Rich.CertifiedProjection.extensions_not_forgeable"
lean_statement_sha256 = "77fa3bfc160297a578bb99027f43b51d5f93820a10000030043dbcb37f577bcf"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -6446,7 +6446,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -6467,7 +6467,7 @@ claim_text = "Lean proves the rich authority property: root not forgeable."
claim_status = "proved"
lean_declaration = "Auths.Rich.CertifiedProjection.root_not_forgeable"
lean_statement_sha256 = "5e5412a76d954a30bb8e5581001dc229b77ddde589cfc4209756972e667e02a6"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -6529,7 +6529,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -6550,7 +6550,7 @@ claim_text = "Lean proves the rich authority property: validity start direction.
claim_status = "proved"
lean_declaration = "Auths.Rich.Mutations.validity_start_direction"
lean_statement_sha256 = "b131c8323adfe2d1fdfb276e8b57d8ecbeeadc7ba76c42e142e9e60a1c7c8bd0"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -6612,7 +6612,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -6633,7 +6633,7 @@ claim_text = "Lean proves the rich authority property: validity end direction."
claim_status = "proved"
lean_declaration = "Auths.Rich.Mutations.validity_end_direction"
lean_statement_sha256 = "64a8422660d220c34ccf2286b3d2b656caee30d180efb50390b98109f012cb9b"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -6695,7 +6695,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -6716,7 +6716,7 @@ claim_text = "Lean proves the rich authority property: permission subset directi
claim_status = "proved"
lean_declaration = "Auths.Rich.Mutations.permission_subset_direction"
lean_statement_sha256 = "460b9c2c8b40e2c47974aa6f40439a0921401af0a741e636dca7e8ce1e1128cb"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -6778,7 +6778,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -6799,7 +6799,7 @@ claim_text = "Lean proves the rich authority property: audience subset direction
claim_status = "proved"
lean_declaration = "Auths.Rich.Mutations.audience_subset_direction"
lean_statement_sha256 = "1efdeced0d7cc63e210e3eed78bf72af7188d0b4b89988645067c065df38d4ca"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -6861,7 +6861,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -6882,7 +6882,7 @@ claim_text = "Lean proves the rich authority property: permission membership dec
claim_status = "proved"
lean_declaration = "Auths.Rich.Mutations.permission_membership_decision"
lean_statement_sha256 = "5562c113f4b1600e9fa2e351daefdf7ad676c95399913cfaf5366a4d02693859"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -6944,7 +6944,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -6964,7 +6964,7 @@ claim_text = "Lean proves the rich authority property: audience membership decis
claim_status = "proved"
lean_declaration = "Auths.Rich.Mutations.audience_membership_decision"
lean_statement_sha256 = "5562c113f4b1600e9fa2e351daefdf7ad676c95399913cfaf5366a4d02693859"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -7026,7 +7026,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -7046,7 +7046,7 @@ claim_text = "Lean proves the rich authority property: body digest membership de
claim_status = "proved"
lean_declaration = "Auths.Rich.Mutations.body_digest_membership_decision"
lean_statement_sha256 = "40e8f579841509c93f52d9b1e8ceeb0068d9deee142398402b148d5dcab74f52"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -7108,7 +7108,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -7128,7 +7128,7 @@ claim_text = "Lean proves the rich authority property: body digest subset direct
claim_status = "proved"
lean_declaration = "Auths.Rich.Mutations.body_digest_subset_direction"
lean_statement_sha256 = "6f6d71aafc31d6e2f5f5d6b41450fab50ccbcc47ea24ff1a4d10a06f547e7a3b"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -7190,7 +7190,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -7211,7 +7211,7 @@ claim_text = "Lean proves the rich authority property: action exact equality."
claim_status = "proved"
lean_declaration = "Auths.Rich.Mutations.action_exact_equality"
lean_statement_sha256 = "12d3ee8161fd73a2b9a4c289c074a70dbe838e01889bae915eb1a28a5e0b5cbf"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -7273,7 +7273,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -7293,7 +7293,7 @@ claim_text = "Lean proves the rich authority property: action constructor fallba
claim_status = "proved"
lean_declaration = "Auths.Rich.Mutations.action_constructor_fallback"
lean_statement_sha256 = "52a8700eb9047be465715d24eb7cce83a544f68beb675fc0ca746b20adcef29d"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -7355,7 +7355,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -7376,7 +7376,7 @@ claim_text = "Lean proves the rich authority property: action singleton exact re
claim_status = "proved"
lean_declaration = "Auths.Rich.Mutations.action_singleton_exact_rejection"
lean_statement_sha256 = "f748a54bcd87a1105b376fd789d539dcba915c55b382c4d84d0764697544e403"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -7438,7 +7438,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -7459,7 +7459,7 @@ claim_text = "Lean proves the rich authority property: budget value direction."
claim_status = "proved"
lean_declaration = "Auths.Rich.Mutations.budget_value_direction"
lean_statement_sha256 = "e5c3c785f61156754ca58b613f917897b7204e66cf2904be240c6ae56dc234d5"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -7521,7 +7521,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -7542,7 +7542,7 @@ claim_text = "Lean proves the rich authority property: budget algebra equality."
claim_status = "proved"
lean_declaration = "Auths.Rich.Mutations.budget_algebra_equality"
lean_statement_sha256 = "856f719d1d4fe6a2b19dab9c3bcfd40873e74a55abaf10f945308fbb05b95652"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -7604,7 +7604,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -7625,7 +7625,7 @@ claim_text = "Lean proves the rich authority property: optional budget bounded p
claim_status = "proved"
lean_declaration = "Auths.Rich.Mutations.optional_budget_bounded_parent"
lean_statement_sha256 = "35b7c28a001e2386a652f5e174baa301a265a3eadd74a8c68c3b43d6d0cde025"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -7687,7 +7687,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -7708,7 +7708,7 @@ claim_text = "Lean proves the rich authority property: optional budget no reques
claim_status = "proved"
lean_declaration = "Auths.Rich.Mutations.optional_budget_no_request"
lean_statement_sha256 = "b661893af19ee64775f156871b07d71fd53478d22876735146530e3d0d624299"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -7770,7 +7770,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -7790,7 +7790,7 @@ claim_text = "Lean proves the rich authority property: delegation depth strictne
claim_status = "proved"
lean_declaration = "Auths.Rich.Mutations.delegation_depth_strictness"
lean_statement_sha256 = "7c203a21398399f9e247a106c21958e147051c105775efb5696c6103d3fcef53"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -7852,7 +7852,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -7873,7 +7873,7 @@ claim_text = "Lean proves the rich authority property: status age direction."
claim_status = "proved"
lean_declaration = "Auths.Rich.Mutations.status_age_direction"
lean_statement_sha256 = "bcd2dc8a61a687aa05525983f0a82d8eb2c30f82bd8714e98f4538219ffa4018"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -7935,7 +7935,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -7956,7 +7956,7 @@ claim_text = "Lean proves the rich authority property: status method equality."
claim_status = "proved"
lean_declaration = "Auths.Rich.Mutations.status_method_equality"
lean_statement_sha256 = "65eb51069ff53dc4a198806e1fd34df0d06c718b805fe98af2c9ce6b81818c8f"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -8018,7 +8018,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -8039,7 +8039,7 @@ claim_text = "Lean proves the rich authority property: assurance equality."
claim_status = "proved"
lean_declaration = "Auths.Rich.Mutations.assurance_equality"
lean_statement_sha256 = "682a5d15cb1ed38aafddcb2a5e0e987609428320b263e5bbae4cbea456d9d8ca"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -8101,7 +8101,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -8122,7 +8122,7 @@ claim_text = "Lean kills the mutation that accepts a child critical extension wi
claim_status = "proved"
lean_declaration = "Auths.Rich.Mutations.critical_extension_law"
lean_statement_sha256 = "23f565b7d27ba7241114c8362d3604ec74ed08a4e05b82d5a97dce2500c87970"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -8184,7 +8184,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -8205,7 +8205,7 @@ claim_text = "Lean proves the rich authority property: critical extensions rever
claim_status = "proved"
lean_declaration = "Auths.Rich.Mutations.critical_extensions_reversed_not_sorted"
lean_statement_sha256 = "1ac76c3db510279b30c76d84b5538b47ad6c5f46fd2e3ae71a45976fea3131dc"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -8267,7 +8267,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -8284,7 +8284,7 @@ claim_text = "Lean proves the rich authority property: profile version equality.
claim_status = "proved"
lean_declaration = "Auths.Rich.Mutations.profile_version_equality"
lean_statement_sha256 = "d061c3229df6d475f2f0dfc7a492dc1dce9c34555a20bd03b530e579a1b34d72"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -8346,7 +8346,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -8367,7 +8367,7 @@ claim_text = "Lean proves the rich authority property: principal linkage equalit
claim_status = "proved"
lean_declaration = "Auths.Rich.Mutations.principal_linkage_equality"
lean_statement_sha256 = "399766c14541a0cf9533a13cbb97e40432b86f182ea9b99bfc835aa86556998d"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -8429,7 +8429,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -8450,7 +8450,7 @@ claim_text = "Lean proves the rich authority property: grant linkage equality."
claim_status = "proved"
lean_declaration = "Auths.Rich.Mutations.grant_linkage_equality"
lean_statement_sha256 = "78a60f2b43f1db4ae1ea88249aa8edab996999d314cc0d09a55151ae5170fb0c"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -8512,7 +8512,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -8533,7 +8533,7 @@ claim_text = "Lean proves the rich authority property: coverage diagnostic sound
claim_status = "proved"
lean_declaration = "Auths.Rich.coverage_diagnostic_sound_complete"
lean_statement_sha256 = "77658775391d0d2d6b38529f58c690b9aa08a5146284dbb022765523106519e5"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -8595,7 +8595,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -8615,7 +8615,7 @@ claim_text = "The mechanically translated production author-scope evaluator retu
claim_status = "qualified"
lean_declaration = "Auths.Refinement.translated_rust_refines_rich_spec"
lean_statement_sha256 = "195ddcaa796803794574579d593c5a78f7fbd9611d32492de2a095e044fe2ca8"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = [
"auths_authority::evaluate_author_scope_view",
"auths_model::profile_ref_equal",
@@ -8696,7 +8696,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The exact translated pre-signing scope evaluator over validated bounded views, including all ten scope/depth dimensions and ordered diagnostics, for every set of critical-extension laws."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, the listed foundational axioms, and the theorem's explicit representation-validity premises are trusted.",
@@ -8733,7 +8733,7 @@ sha256 = "2bb401ffca0136622cd79bb14a5a38852061649a06cc46100503870c22d300c7"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-055"
@@ -8741,7 +8741,7 @@ claim_text = "The mechanically translated terminal-coverage evaluator returns ex
claim_status = "qualified"
lean_declaration = "Auths.Refinement.translated_coverage_refines_rich_spec"
lean_statement_sha256 = "e8deba0035a09886221c46411be908f1ac8c30e0e3755ca956a62884a416a0fa"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = [
"auths_authority::evaluate_action_coverage_view",
"auths_authority::root_linkage",
@@ -8818,7 +8818,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The crate-private raw translated evaluator over validated authority/action views. The public EffectiveAuthority::authorizes boundary resolves budget expressibility from AcceptedRegistries and the exact action profile before calling it."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, the listed foundational axioms, and the theorem's explicit representation-validity premises are trusted.",
@@ -8854,7 +8854,7 @@ sha256 = "2bb401ffca0136622cd79bb14a5a38852061649a06cc46100503870c22d300c7"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-056"
@@ -8862,7 +8862,7 @@ claim_text = "The mechanically translated delegation evaluator returns exactly t
claim_status = "qualified"
lean_declaration = "Auths.Refinement.translated_delegation_refines_rich_spec"
lean_statement_sha256 = "45782f85b6ded4714d9ca4e4b6a8ee2dc0e70c99395691101d834087a89c2fe1"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = [
"auths_authority::evaluate_grant_view",
"auths_authority::root_linkage",
@@ -8952,7 +8952,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The crate-private raw evaluator over validated parent/grant views, including root linkage, strict depth, every attenuation dimension, and the unique accepted transition, for every set of critical-extension laws."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, the listed foundational axioms, and the theorem's explicit representation-validity premises are trusted.",
@@ -8989,7 +8989,7 @@ sha256 = "2bb401ffca0136622cd79bb14a5a38852061649a06cc46100503870c22d300c7"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-104"
@@ -8997,7 +8997,7 @@ claim_text = "The regenerated Aeneas translation of optional_budget_covers answe
claim_status = "proved"
lean_declaration = "Auths.Refinement.translated_bounded_ceiling_denies_absent_request"
lean_statement_sha256 = "9c4ea11fa46f39a993ab4038263180284eca394df782a70259346c8167d7e4fc"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -9059,7 +9059,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -9080,7 +9080,7 @@ claim_text = "Lean proves the rich authority property: translated expressible ab
claim_status = "proved"
lean_declaration = "Auths.Refinement.translated_expressible_absent_request_is_denied"
lean_statement_sha256 = "90aa025faf3b73862f51432d5b07d0ef69c1bd67826d739083b6e42d65f00e4a"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -9142,7 +9142,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -9163,7 +9163,7 @@ claim_text = "Lean proves the rich authority property: translated inexpressible
claim_status = "proved"
lean_declaration = "Auths.Refinement.translated_inexpressible_absent_request_is_covered"
lean_statement_sha256 = "294ac2be47b25a2f72d528f66b974cb3acc2dfd8a330a8c6696b8dad12fb8d97"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -9225,7 +9225,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -9246,7 +9246,7 @@ claim_text = "Lean proves the rich authority property: translated absent ceiling
claim_status = "proved"
lean_declaration = "Auths.Refinement.translated_absent_ceiling_covers_anything"
lean_statement_sha256 = "90df3548fb3d56bed30812c94acb18b2d6b0028f77aff11cd44aade8e2889961"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -9308,7 +9308,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -9329,7 +9329,7 @@ claim_text = "Lean proves the bounded product-policy property: configuration mat
claim_status = "proved"
lean_declaration = "Auths.Product.configuration_match_refl"
lean_statement_sha256 = "1d3b3a5ed6156f4e5616d877b35e9942f824f876e6db3aa3cd39729305a8ed4e"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -9391,7 +9391,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure V1 bounded product-policy commitments, checked arithmetic, configuration gating, and eligibility."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -9408,7 +9408,7 @@ claim_text = "Lean proves the bounded product-policy property: configuration mat
claim_status = "proved"
lean_declaration = "Auths.Product.configuration_match_deterministic"
lean_statement_sha256 = "0b97aea7619fab2ffebd442c8d6376a02779d0e4c9673837e48f45c99488c8e0"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -9470,7 +9470,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure V1 bounded product-policy commitments, checked arithmetic, configuration gating, and eligibility."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -9487,7 +9487,7 @@ claim_text = "Lean proves the bounded product-policy property: checked add never
claim_status = "proved"
lean_declaration = "Auths.Product.checked_add_never_wraps"
lean_statement_sha256 = "f2e86c112c2021bab2444682e85973ad5b3f3de473bf5e093a47c65ebd74cf06"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -9549,7 +9549,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure V1 bounded product-policy commitments, checked arithmetic, configuration gating, and eligibility."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -9566,7 +9566,7 @@ claim_text = "Lean proves the bounded product-policy property: checked sub never
claim_status = "proved"
lean_declaration = "Auths.Product.checked_sub_never_underflows"
lean_statement_sha256 = "fc97e83a1f988edafe17d5d775cafb333138ff36e1ee3ac65609f48844129fa5"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -9628,7 +9628,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure V1 bounded product-policy commitments, checked arithmetic, configuration gating, and eligibility."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -9645,7 +9645,7 @@ claim_text = "Lean proves the bounded product-policy property: checked div rejec
claim_status = "proved"
lean_declaration = "Auths.Product.checked_div_rejects_zero"
lean_statement_sha256 = "004c9c8a21f3a6416181d4d561ae3dd38aebca10d640f288eb8af0e4bba34ba4"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -9707,7 +9707,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure V1 bounded product-policy commitments, checked arithmetic, configuration gating, and eligibility."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -9724,7 +9724,7 @@ claim_text = "Lean proves the bounded product-policy property: three way partiti
claim_status = "proved"
lean_declaration = "Auths.Product.three_way_partition"
lean_statement_sha256 = "fa4bfacdd689bfe94265f7c2627598a8d758ae6ac3667bc6270f581219151e3e"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -9786,7 +9786,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure V1 bounded product-policy commitments, checked arithmetic, configuration gating, and eligibility."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -9803,7 +9803,7 @@ claim_text = "Lean proves the bounded product-policy property: product contract
claim_status = "proved"
lean_declaration = "Auths.Product.product_contract_configuration_safety"
lean_statement_sha256 = "18957d267951e86ccae5c49b3b704e2e45d87c427d023f4e0a3684b895e461d5"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -9865,7 +9865,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure V1 bounded product-policy commitments, checked arithmetic, configuration gating, and eligibility."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -9882,7 +9882,7 @@ claim_text = "Lean proves the bounded product-policy property: product contract
claim_status = "proved"
lean_declaration = "Auths.Product.product_contract_output_completeness"
lean_statement_sha256 = "b4fa5d97dd43b72133cc583f8834921ed1c968fde6ac3eb4f7c4b5ecc49f1c08"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -9944,7 +9944,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure V1 bounded product-policy commitments, checked arithmetic, configuration gating, and eligibility."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -9964,7 +9964,7 @@ claim_text = "Lean proves the bounded product-policy property: translated config
claim_status = "proved"
lean_declaration = "Auths.Product.Refinement.translated_configuration_refines_projection"
lean_statement_sha256 = "a988d91b2f65fea6451fdbe5ebc0b75ea7324b27c0b4031b1b49144eb15ef297"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -10026,7 +10026,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure V1 bounded product-policy commitments, checked arithmetic, configuration gating, and eligibility."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -10043,7 +10043,7 @@ claim_text = "Lean proves the bounded product-policy property: translated checke
claim_status = "proved"
lean_declaration = "Auths.Product.Refinement.translated_checked_add_refines_nat"
lean_statement_sha256 = "eed21c6bdc5e0ba56c0c8dd9249ee1d51053fa42b8df3958a762399ba0bf7ab5"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -10105,7 +10105,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure V1 bounded product-policy commitments, checked arithmetic, configuration gating, and eligibility."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -10126,7 +10126,7 @@ claim_text = "Lean proves the bounded product-policy property: translated checke
claim_status = "proved"
lean_declaration = "Auths.Product.Refinement.translated_checked_sub_refines_nat"
lean_statement_sha256 = "fed19d1701dac7a6f928957a78799fcf63eaf80d24ea0cbed66d77e490ed879c"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -10188,7 +10188,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure V1 bounded product-policy commitments, checked arithmetic, configuration gating, and eligibility."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -10209,7 +10209,7 @@ claim_text = "Lean proves the bounded product-policy property: translated checke
claim_status = "proved"
lean_declaration = "Auths.Product.Refinement.translated_checked_mul_refines_nat"
lean_statement_sha256 = "f2e5d31dcc8e7f7651c0223bfdabccadc850be49cb8431aebf966aee9f7acc35"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -10271,7 +10271,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure V1 bounded product-policy commitments, checked arithmetic, configuration gating, and eligibility."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -10292,7 +10292,7 @@ claim_text = "Lean proves the bounded product-policy property: translated checke
claim_status = "proved"
lean_declaration = "Auths.Product.Refinement.translated_checked_div_rejects_zero"
lean_statement_sha256 = "cb30a894abfc0b326ba704c9d0979adb434cb7fc94f0134e14442c313e8670fb"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -10354,7 +10354,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure V1 bounded product-policy commitments, checked arithmetic, configuration gating, and eligibility."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -10375,7 +10375,7 @@ claim_text = "The mechanically translated window index returns exactly the natur
claim_status = "qualified"
lean_declaration = "Auths.Product.Refinement.translated_window_index_refines_nat"
lean_statement_sha256 = "2d4076b355c9fd4faad8b9663b8a68e1c49f619da2cf11c74e9bdb30f7d480da"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_bounded_policy::kernel::window_index"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -10437,7 +10437,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Every pair of u64 time and window length."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, and the listed foundational axioms are trusted; the theorem has no premise.",
@@ -10463,7 +10463,7 @@ sha256 = "ba7aea214063b8a698fbd2c9c8cd95a717bf0ec0eaef70a92844444bf7e9f22e"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-263"
@@ -10471,7 +10471,7 @@ claim_text = "The mechanically translated digest comparison returns exactly the
claim_status = "qualified"
lean_declaration = "Auths.Product.Refinement.BoundedPolicy.translated_digest_equal_refines_model"
lean_statement_sha256 = "d8411964df44b9e93b4d097961e55e60bef00a8ef71c89d7b3bec08963599b85"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_model::bounded_policy::digest_equal"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -10533,7 +10533,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Every pair of 32-byte digests."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, and the listed foundational axioms are trusted; the theorem has no premise.",
@@ -10559,7 +10559,7 @@ sha256 = "d4d36b72718a4076b7c0561abfa2edc006a1f9a4491762979f86c3dd99a71fdc"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-264"
@@ -10567,7 +10567,7 @@ claim_text = "The mechanically translated parent-link check returns exactly the
claim_status = "qualified"
lean_declaration = "Auths.Product.Refinement.BoundedPolicy.translated_bounded_policy_link_accepts_refines_model"
lean_statement_sha256 = "7c7c5aec926901e72f22a2b9c4b1c6c4bb833c010440f657fcb79d7e827144d7"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = [
"auths_model::bounded_policy::bounded_policy_link_accepts",
"auths_model::bounded_policy::digest_equal",
@@ -10632,7 +10632,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Every optional child link and optional parent digest."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, and the listed foundational axioms are trusted; the theorem has no premise.",
@@ -10658,7 +10658,7 @@ sha256 = "d4d36b72718a4076b7c0561abfa2edc006a1f9a4491762979f86c3dd99a71fdc"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-186"
@@ -10666,7 +10666,7 @@ claim_text = "Lean proves the translated argument-ceiling evaluation leaf comput
claim_status = "proved"
lean_declaration = "Auths.Product.Refinement.translated_ceiling_count_refines_model"
lean_statement_sha256 = "ada2be3af3def88e58de1ff9aef4485651996241a5c2a0f8ce0dd50d27728b1e"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -10728,7 +10728,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The Aeneas-translated argument-ceiling window-count leaves of auths-bounded-policy, under the pinned translation."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -10748,7 +10748,7 @@ claim_text = "Lean proves the translated argument-ceiling tightening leaf comput
claim_status = "proved"
lean_declaration = "Auths.Product.Refinement.translated_ceiling_count_tightens_refines_model"
lean_statement_sha256 = "9477f8ea177a765c5ac61d23ce7032ca8156309063f5963bad218c829d08981c"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -10810,7 +10810,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The Aeneas-translated argument-ceiling window-count leaves of auths-bounded-policy, under the pinned translation."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -10830,7 +10830,7 @@ claim_text = "The mechanically translated count leaf returns exactly the model's
claim_status = "qualified"
lean_declaration = "Auths.Product.Refinement.translated_chain_counts_admit_refines_model"
lean_statement_sha256 = "f71baa54d41b494f3c2a69f250c2183339389bbe1d5d94fd64d8b5ed9afb7d67"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_bounded_policy::kernel::chain_counts_admit"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -10892,7 +10892,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Every pair of u64 slices of counter values and capacities, including slices of different lengths."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, and the listed foundational axioms are trusted; the theorem has no premise.",
@@ -10918,7 +10918,7 @@ sha256 = "ba7aea214063b8a698fbd2c9c8cd95a717bf0ec0eaef70a92844444bf7e9f22e"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-237"
@@ -10926,7 +10926,7 @@ claim_text = "The mechanically translated sum leaf returns exactly the model's s
claim_status = "qualified"
lean_declaration = "Auths.Product.Refinement.translated_chain_sums_admit_refines_model"
lean_statement_sha256 = "c75fcec112558e4eca8125e5996cd4dea6d1e2a76cd45bf1b1accbf1c0dfc5ee"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_bounded_policy::kernel::chain_sums_admit"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -10988,7 +10988,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Every u64 argument and pair of u64 slices of running sums and capacities, including slices of different lengths."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, and the listed foundational axioms are trusted; the theorem has no premise.",
@@ -11014,7 +11014,7 @@ sha256 = "ba7aea214063b8a698fbd2c9c8cd95a717bf0ec0eaef70a92844444bf7e9f22e"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-238"
@@ -11022,7 +11022,7 @@ claim_text = "The mechanically translated tightening decider of policy /2 return
claim_status = "qualified"
lean_declaration = "Auths.Product.Refinement.translated_argument_policy_tightens_refines_model"
lean_statement_sha256 = "f20279e5c8f3b5e9d3f9e56408c91d6f0151b2106434969f9890a5ef5389c337"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = [
"auths_bounded_policy::kernel::argument_policy_tightens",
"auths_bounded_policy::kernel::ceiling_count_tightens",
@@ -11087,7 +11087,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Every pair of policy members: argument names, ceilings, windows, counts, optional sums with optional partitions, and optional scopes with value lists of any length."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, and the listed foundational axioms are trusted; the theorem has no premise.",
@@ -11113,7 +11113,7 @@ sha256 = "ba7aea214063b8a698fbd2c9c8cd95a717bf0ec0eaef70a92844444bf7e9f22e"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-175"
@@ -11121,7 +11121,7 @@ claim_text = "Lean proves the argument-ceiling evaluation is eligible exactly wh
claim_status = "proved"
lean_declaration = "Auths.Product.CeilingCount.code_eligible_iff"
lean_statement_sha256 = "ca2f3dfef380a22f6280465f8320e1aca773e3ca0290cc38f7babe12f6642e63"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -11183,7 +11183,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The gateway's registered argument-ceiling window-count evaluator /2: fixed-context evaluation over verified unsigned and text arguments and the counters of one fixed, epoch-aligned window, semantic tightening with the sum, partition, and scope rules, and the registered tightening decider."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -11200,7 +11200,7 @@ claim_text = "Lean proves the argument-ceiling tightening decider of policy /2 d
claim_status = "proved"
lean_declaration = "Auths.Product.CeilingCount.decider_iff_tightens"
lean_statement_sha256 = "ae8a39a1e41e905d9464cc4dad8638cf1047017ff0ad58d77cceba6498a1a2a3"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -11262,7 +11262,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The gateway's registered argument-ceiling window-count evaluator /2: fixed-context evaluation over verified unsigned and text arguments and the counters of one fixed, epoch-aligned window, semantic tightening with the sum, partition, and scope rules, and the registered tightening decider."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -11282,7 +11282,7 @@ claim_text = "Lean proves a tighter argument-ceiling policy never admits a conte
claim_status = "proved"
lean_declaration = "Auths.Product.CeilingCount.tightening_never_admits_more"
lean_statement_sha256 = "fbdda26118f2596fdad003444f236e79a1bf12754a1a6beb0d61e30dd66e6287"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -11344,7 +11344,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The gateway's registered argument-ceiling window-count evaluator /2: fixed-context evaluation over verified unsigned and text arguments and the counters of one fixed, epoch-aligned window, semantic tightening with the sum, partition, and scope rules, and the registered tightening decider."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -11365,7 +11365,7 @@ claim_text = "Lean proves the argument-ceiling tightening decider is sound: a ch
claim_status = "proved"
lean_declaration = "Auths.Product.CeilingCount.decider_sound"
lean_statement_sha256 = "9be80ea95233697c96ff6a9b5dce9aca5f84988261dcb57119c5e7bf0b7af0e8"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -11427,7 +11427,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The gateway's registered argument-ceiling window-count evaluator /2: fixed-context evaluation over verified unsigned and text arguments and the counters of one fixed, epoch-aligned window, semantic tightening with the sum, partition, and scope rules, and the registered tightening decider."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -11448,7 +11448,7 @@ claim_text = "Lean proves a smaller argument ceiling never authorizes more."
claim_status = "proved"
lean_declaration = "Auths.Product.CeilingCount.smaller_ceiling_never_authorizes_more"
lean_statement_sha256 = "9e56daaf4f09d2a01150daf784d6584c590f59545639139ae91bfa6be4025690"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -11510,7 +11510,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The gateway's registered argument-ceiling window-count evaluator /2: fixed-context evaluation over verified unsigned and text arguments and the counters of one fixed, epoch-aligned window, semantic tightening with the sum, partition, and scope rules, and the registered tightening decider."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -11531,7 +11531,7 @@ claim_text = "Lean proves a smaller count per fixed window never authorizes more
claim_status = "proved"
lean_declaration = "Auths.Product.CeilingCount.smaller_count_never_authorizes_more"
lean_statement_sha256 = "ddcd9f8d9141338018e9fb5a68fca37b7e157619f0ca917c0550cef5c8a828a6"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -11593,7 +11593,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The gateway's registered argument-ceiling window-count evaluator /2: fixed-context evaluation over verified unsigned and text arguments and the counters of one fixed, epoch-aligned window, semantic tightening with the sum, partition, and scope rules, and the registered tightening decider."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -11614,7 +11614,7 @@ claim_text = "Lean proves argument-ceiling tightening is reflexive."
claim_status = "proved"
lean_declaration = "Auths.Product.CeilingCount.tightens_refl"
lean_statement_sha256 = "3c2b30687d72f28f07c13e1069f9e8f49ad2ca604c8d0b9561ad86e64090dec6"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -11676,7 +11676,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The gateway's registered argument-ceiling window-count evaluator /2: fixed-context evaluation over verified unsigned and text arguments and the counters of one fixed, epoch-aligned window, semantic tightening with the sum, partition, and scope rules, and the registered tightening decider."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -11696,7 +11696,7 @@ claim_text = "Lean proves argument-ceiling tightening is transitive."
claim_status = "proved"
lean_declaration = "Auths.Product.CeilingCount.tightens_trans"
lean_statement_sha256 = "27bddfea8d4eaa275419f7461e14c30f42db2e6aa67ed049e2baafe0c113b7fd"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -11758,7 +11758,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The gateway's registered argument-ceiling window-count evaluator /2: fixed-context evaluation over verified unsigned and text arguments and the counters of one fixed, epoch-aligned window, semantic tightening with the sum, partition, and scope rules, and the registered tightening decider."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -11775,7 +11775,7 @@ claim_text = "Lean proves the bounded-policy extension law the gateway enforces
claim_status = "proved"
lean_declaration = "Auths.Product.CeilingCount.bounded_policy_law_lawful"
lean_statement_sha256 = "b5311babd266c58b0a167f530675436fd804cf4ba0af4a66d33f6b8ba37fa749"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -11837,7 +11837,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Product-layer premise of the bounded-policy-commitment-v1 extension. The kernel checks only that a delegated commitment links the digest of its parent's extension bytes; narrowing is discharged here by the registered evaluator's tightening decider, which the gateway runs on every linked pair before eligibility. This is not a kernel theorem."
residual_assumptions = [
"Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted.",
@@ -11861,7 +11861,7 @@ claim_text = "Lean proves the bounded-policy extension law accepts adding a boun
claim_status = "proved"
lean_declaration = "Auths.Product.CeilingCount.bounded_policy_law_accepts_addition"
lean_statement_sha256 = "1025dc58f2d798f02e8a8a643c4c3a76c544e76302970c2d7ce99ed6274cf89b"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -11923,7 +11923,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Product-layer premise of the bounded-policy-commitment-v1 extension. The kernel checks only that a delegated commitment links the digest of its parent's extension bytes; narrowing is discharged here by the registered evaluator's tightening decider, which the gateway runs on every linked pair before eligibility. This is not a kernel theorem."
residual_assumptions = [
"Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted.",
@@ -11946,7 +11946,7 @@ claim_text = "Lean proves the argument-ceiling evaluator satisfies the shared fi
claim_status = "proved"
lean_declaration = "Auths.Product.CeilingCount.ceiling_count_fixed_context_tightening"
lean_statement_sha256 = "30accc61db6512783cf3b568eefa9d09579eccc72e3ce2d0ca196d3e27efbd40"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -12008,7 +12008,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The gateway's registered argument-ceiling window-count evaluator /2: fixed-context evaluation over verified unsigned and text arguments and the counters of one fixed, epoch-aligned window, semantic tightening with the sum, partition, and scope rules, and the registered tightening decider."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -12029,7 +12029,7 @@ claim_text = "Lean proves reserving a chain's counters is all or none: either no
claim_status = "proved"
lean_declaration = "Auths.Product.CeilingCount.reserve_all_or_none"
lean_statement_sha256 = "6ab5bb8963257698a595bbad3d04e33cf7a4284948372b3c1217489fd81cb0ef"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -12091,7 +12091,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The gateway's chain admission and reservation over the counters of one fixed, epoch-aligned window: each bounded grant of the authorized branch is a link whose count and sum counters are keyed by its subject, so every admitted action of the subject and of each delegate charges them. Store atomicity and the unauthenticated gateway clock are residual assumptions."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -12111,7 +12111,7 @@ claim_text = "Lean proves that, for every arrival order in one fixed window, a l
claim_status = "proved"
lean_declaration = "Auths.Product.CeilingCount.aggregate_count_bound"
lean_statement_sha256 = "0f5accd437257509195f6034a9d50f0713a21aeab2c655b0fecb95527124eb8d"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -12173,7 +12173,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The gateway's chain admission and reservation over the counters of one fixed, epoch-aligned window: each bounded grant of the authorized branch is a link whose count and sum counters are keyed by its subject, so every admitted action of the subject and of each delegate charges them. Store atomicity and the unauthenticated gateway clock are residual assumptions."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -12193,7 +12193,7 @@ claim_text = "Lean proves that, for every arrival order in one fixed window, the
claim_status = "proved"
lean_declaration = "Auths.Product.CeilingCount.aggregate_spend_bound"
lean_statement_sha256 = "8bc18ab4c3404924341910f3b156124dad2a029e37e2fdef41b52a5117a13340"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -12255,7 +12255,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The gateway's chain admission and reservation over the counters of one fixed, epoch-aligned window: each bounded grant of the authorized branch is a link whose count and sum counters are keyed by its subject, so every admitted action of the subject and of each delegate charges them. Store atomicity and the unauthenticated gateway clock are residual assumptions."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -12275,7 +12275,7 @@ claim_text = "Lean proves that, for every arrival order in one fixed window, a l
claim_status = "proved"
lean_declaration = "Auths.Product.CeilingCount.aggregate_sum_bound"
lean_statement_sha256 = "37700c8ea564cd8346ffd48a8f33774fae3436abd135c0584bc7704afe7ff07f"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -12337,7 +12337,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The gateway's chain admission and reservation over the counters of one fixed, epoch-aligned window: each bounded grant of the authorized branch is a link whose count and sum counters are keyed by its subject, so every admitted action of the subject and of each delegate charges them. Store atomicity and the unauthenticated gateway clock are residual assumptions."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -12357,7 +12357,7 @@ claim_text = "Lean proves delegation never multiplies a count: in one fixed wind
claim_status = "proved"
lean_declaration = "Auths.Product.CeilingCount.delegation_never_multiplies"
lean_statement_sha256 = "f09ba1275f7a50f260a64fc78ed3663ea70f459ba704b3acb4598f685fdd9184"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -12419,7 +12419,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The gateway's chain admission and reservation over the counters of one fixed, epoch-aligned window: each bounded grant of the authorized branch is a link whose count and sum counters are keyed by its subject, so every admitted action of the subject and of each delegate charges them. Store atomicity and the unauthenticated gateway clock are residual assumptions."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -12439,7 +12439,7 @@ claim_text = "Lean proves delegation never multiplies a sum: in one fixed window
claim_status = "proved"
lean_declaration = "Auths.Product.CeilingCount.sum_delegation_never_multiplies"
lean_statement_sha256 = "289fea63130a02da951fd05c805bb6df285781948437b27519116a3e1b2defaf"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -12501,7 +12501,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The gateway's chain admission and reservation over the counters of one fixed, epoch-aligned window: each bounded grant of the authorized branch is a link whose count and sum counters are keyed by its subject, so every admitted action of the subject and of each delegate charges them. Store atomicity and the unauthenticated gateway clock are residual assumptions."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -12521,7 +12521,7 @@ claim_text = "Lean proves an admitted chain's scope and partition values are in
claim_status = "proved"
lean_declaration = "Auths.Product.CeilingCount.scope_admits_only_listed"
lean_statement_sha256 = "cad1fcf6fbaf3064577c894a1a8ea868c3e4085020bb580625abbb168f84a49f"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -12583,7 +12583,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The gateway's chain admission and reservation over the counters of one fixed, epoch-aligned window: each bounded grant of the authorized branch is a link whose count and sum counters are keyed by its subject, so every admitted action of the subject and of each delegate charges them. Store atomicity and the unauthenticated gateway clock are residual assumptions."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -12603,7 +12603,7 @@ claim_text = "Lean proves the count leaf's model, over a chain's distinct count
claim_status = "proved"
lean_declaration = "Auths.Product.CeilingCount.chain_counts_part"
lean_statement_sha256 = "e93ff36670cc1ff2df9e0af15ff3bef912520b07c55b67ade09e6b84ce314f2a"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -12665,7 +12665,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The gateway's chain admission and reservation over the counters of one fixed, epoch-aligned window: each bounded grant of the authorized branch is a link whose count and sum counters are keyed by its subject, so every admitted action of the subject and of each delegate charges them. Store atomicity and the unauthenticated gateway clock are residual assumptions."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -12686,7 +12686,7 @@ claim_text = "Lean proves the sum leaf's model, over a chain's distinct sum coun
claim_status = "proved"
lean_declaration = "Auths.Product.CeilingCount.chain_sums_part"
lean_statement_sha256 = "f10a2dd1bc798c06920ade27ff1bfeba59a7fbfa6653d81664b258ad231ed7f2"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -12748,7 +12748,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The gateway's chain admission and reservation over the counters of one fixed, epoch-aligned window: each bounded grant of the authorized branch is a link whose count and sum counters are keyed by its subject, so every admitted action of the subject and of each delegate charges them. Store atomicity and the unauthenticated gateway clock are residual assumptions."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -12769,7 +12769,7 @@ claim_text = "Lean proves chain admission is exactly the per-link argument, ceil
claim_status = "proved"
lean_declaration = "Auths.Product.CeilingCount.chain_admits_parts"
lean_statement_sha256 = "3b815dca82bdd32ac7a47c9a5b5c29a70942dffb774dbe00ed4940ff34eeb880"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -12831,7 +12831,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The gateway's chain admission and reservation over the counters of one fixed, epoch-aligned window: each bounded grant of the authorized branch is a link whose count and sum counters are keyed by its subject, so every admitted action of the subject and of each delegate charges them. Store atomicity and the unauthenticated gateway clock are residual assumptions."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -12851,7 +12851,7 @@ claim_text = "Lean proves the gateway recipe property: capability total determin
claim_status = "proved"
lean_declaration = "Auths.Product.Recovery.capability_total_deterministic"
lean_statement_sha256 = "6f4fc1fec3bdae5a436cbf7894e76aa835bec591fc37b7dbf002aca033c13667"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -12913,7 +12913,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure gateway recipe semantics: the recovery capability a recipe's declarations determine, and closed request construction over every compiled plan and argument list."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -12922,7 +12922,7 @@ axioms = []
[[claims.evidence]]
kind = "lean-proof"
artifact = "formal/Auths/Product/Recovery.lean"
-sha256 = "54b0c7fa26dfa03b5a9659875100a4b5bc845dfe010f73d0ac96bc981c208b70"
+sha256 = "04c16abce67bebe1cb1a395a1a025a5c5cc9f84f6e2e34327bae0af88cb48d5c"
[[claims]]
claim_id = "AP-FORMAL-RICH-203"
@@ -12930,7 +12930,7 @@ claim_text = "Lean proves the gateway recipe property: class matches declaration
claim_status = "proved"
lean_declaration = "Auths.Product.Recovery.class_matches_declarations"
lean_statement_sha256 = "117884bcb89669e3376db0b00e75a96265c1bfdb3392000fde1430b59ea3a078"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -12992,7 +12992,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure gateway recipe semantics: the recovery capability a recipe's declarations determine, and closed request construction over every compiled plan and argument list."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -13001,7 +13001,7 @@ axioms = ["propext"]
[[claims.evidence]]
kind = "lean-proof"
artifact = "formal/Auths/Product/Recovery.lean"
-sha256 = "54b0c7fa26dfa03b5a9659875100a4b5bc845dfe010f73d0ac96bc981c208b70"
+sha256 = "04c16abce67bebe1cb1a395a1a025a5c5cc9f84f6e2e34327bae0af88cb48d5c"
[[claims]]
claim_id = "AP-FORMAL-RICH-204"
@@ -13009,7 +13009,7 @@ claim_text = "Lean proves the gateway recipe property: write never conditional."
claim_status = "proved"
lean_declaration = "Auths.Product.Recovery.write_never_conditional"
lean_statement_sha256 = "a29b4d1636de3467039d88777f399f89ed9748f66ca901ced1ebe540275f3832"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -13071,7 +13071,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure gateway recipe semantics: the recovery capability a recipe's declarations determine, and closed request construction over every compiled plan and argument list."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -13080,15 +13080,15 @@ axioms = []
[[claims.evidence]]
kind = "lean-proof"
artifact = "formal/Auths/Product/Recovery.lean"
-sha256 = "54b0c7fa26dfa03b5a9659875100a4b5bc845dfe010f73d0ac96bc981c208b70"
+sha256 = "04c16abce67bebe1cb1a395a1a025a5c5cc9f84f6e2e34327bae0af88cb48d5c"
[[claims]]
claim_id = "AP-FORMAL-RICH-215"
-claim_text = "Lean proves the gateway recipe property: unknown resolves only linked."
+claim_text = "Lean proves that an unknown reaching provider observation retains a verified-locator link and records provider evidence; late source responses and recipient acknowledgements are distinct routes."
claim_status = "proved"
lean_declaration = "Auths.Product.Recovery.unknown_resolves_only_linked"
-lean_statement_sha256 = "375c42a9e70d6c1717fd9e451f43612cae8141cfff4f9cdd00c28f38970b6383"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+lean_statement_sha256 = "1a756a3cb1a5eb1f238e64bf9724e78fc92671f995385e2a9f8e1dbd4a6bf6a6"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -13150,7 +13150,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure gateway recipe semantics: the recovery capability a recipe's declarations determine, closed request construction over every compiled plan and argument list, the attempt-record transition rule over every sequence of records, the admission order over every event trace, and the exact relative ceiling."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -13162,15 +13162,331 @@ axioms = [
[[claims.evidence]]
kind = "lean-proof"
artifact = "formal/Auths/Product/Recovery.lean"
-sha256 = "54b0c7fa26dfa03b5a9659875100a4b5bc845dfe010f73d0ac96bc981c208b70"
+sha256 = "04c16abce67bebe1cb1a395a1a025a5c5cc9f84f6e2e34327bae0af88cb48d5c"
+
+[[claims]]
+claim_id = "AP-FORMAL-RICH-313"
+claim_text = "Lean proves the gateway recipe property: unknown exits with bound evidence."
+claim_status = "proved"
+lean_declaration = "Auths.Product.Recovery.unknown_exits_with_bound_evidence"
+lean_statement_sha256 = "0da149d7ad7a06229ef88d294abdccd8a8b4e80df4356246b05ada0b733afc78"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
+rust_symbols = []
+semantic_source_closure = [
+ "formal/Auths.lean",
+ "formal/Auths/Approval.lean",
+ "formal/Auths/AssuranceAudit.lean",
+ "formal/Auths/Attenuation.lean",
+ "formal/Auths/Authority.lean",
+ "formal/Auths/Base.lean",
+ "formal/Auths/Composition.lean",
+ "formal/Auths/Diversity.lean",
+ "formal/Auths/ExtensionLaw.lean",
+ "formal/Auths/Generated/Algebra.lean",
+ "formal/Auths/Lifecycle/Refinement.lean",
+ "formal/Auths/Lifecycle/Semantics.lean",
+ "formal/Auths/Lifecycle/Theorems.lean",
+ "formal/Auths/Observation.lean",
+ "formal/Auths/Product/Arithmetic.lean",
+ "formal/Auths/Product/CeilingCount.lean",
+ "formal/Auths/Product/Commitment.lean",
+ "formal/Auths/Product/ConnectionGenerations.lean",
+ "formal/Auths/Product/Eligibility.lean",
+ "formal/Auths/Product/Recovery.lean",
+ "formal/Auths/Product/Refinement.lean",
+ "formal/Auths/Product/Refinement/BoundedPolicy.lean",
+ "formal/Auths/Product/Refinement/Connections.lean",
+ "formal/Auths/Product/Refinement/Gateway.lean",
+ "formal/Auths/Product/RelativeCeiling.lean",
+ "formal/Auths/Product/RequestConstruction.lean",
+ "formal/Auths/Product/SubmitOrder.lean",
+ "formal/Auths/Product/Theorems.lean",
+ "formal/Auths/Product/Tightening.lean",
+ "formal/Auths/Refinement/Approval.lean",
+ "formal/Auths/Refinement/Observation.lean",
+ "formal/Auths/Refinement/Production.lean",
+ "formal/Auths/Rich/Mutations.lean",
+ "formal/Auths/Rich/Semantics.lean",
+ "formal/Auths/Rich/Theorems.lean",
+ "formal/Auths/Rich/Types.lean",
+ "formal/Auths/Theorems.lean",
+ "formal/algebra-contract-v1.toml",
+ "formal/lake-manifest.json",
+ "formal/lakefile.toml",
+ "formal/lean-toolchain",
+ "formal/qualification/aeneas/generated/algebra/Funs.lean",
+ "formal/qualification/aeneas/generated/algebra/Types.lean",
+ "formal/qualification/aeneas/generated/authority/Funs.lean",
+ "formal/qualification/aeneas/generated/authority/FunsExternal.lean",
+ "formal/qualification/aeneas/generated/authority/Types.lean",
+ "formal/qualification/aeneas/generated/authority/TypesExternal.lean",
+ "formal/qualification/aeneas/generated/bounded_policy/Funs.lean",
+ "formal/qualification/aeneas/generated/bounded_policy/Types.lean",
+ "formal/qualification/aeneas/generated/connections/Funs.lean",
+ "formal/qualification/aeneas/generated/connections/Types.lean",
+ "formal/qualification/aeneas/generated/gateway/Funs.lean",
+ "formal/qualification/aeneas/generated/gateway/Types.lean",
+ "formal/qualification/aeneas/generated/lifecycle/Funs.lean",
+ "formal/qualification/aeneas/generated/lifecycle/Types.lean",
+ "formal/qualification/aeneas/generated/model/Funs.lean",
+ "formal/qualification/aeneas/generated/model/FunsExternal.lean",
+ "formal/qualification/aeneas/generated/model/Types.lean",
+]
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
+scope = "Pure gateway recipe semantics: the recovery capability a recipe's declarations determine, closed request construction over every compiled plan and argument list, the attempt-record transition rule over every sequence of records, the admission order over every event trace, and the exact relative ceiling."
+residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
+toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
+axioms = ["propext"]
+
+[[claims.evidence]]
+kind = "lean-proof"
+artifact = "formal/Auths/Product/Recovery.lean"
+sha256 = "04c16abce67bebe1cb1a395a1a025a5c5cc9f84f6e2e34327bae0af88cb48d5c"
+
+[[claims]]
+claim_id = "AP-FORMAL-RICH-314"
+claim_text = "Lean proves the gateway recipe property: transition ownership."
+claim_status = "proved"
+lean_declaration = "Auths.Product.Recovery.transition_ownership"
+lean_statement_sha256 = "504422bb822a66679164c3ddc04582e541e483297ea461e19f919fd04cde85e9"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
+rust_symbols = []
+semantic_source_closure = [
+ "formal/Auths.lean",
+ "formal/Auths/Approval.lean",
+ "formal/Auths/AssuranceAudit.lean",
+ "formal/Auths/Attenuation.lean",
+ "formal/Auths/Authority.lean",
+ "formal/Auths/Base.lean",
+ "formal/Auths/Composition.lean",
+ "formal/Auths/Diversity.lean",
+ "formal/Auths/ExtensionLaw.lean",
+ "formal/Auths/Generated/Algebra.lean",
+ "formal/Auths/Lifecycle/Refinement.lean",
+ "formal/Auths/Lifecycle/Semantics.lean",
+ "formal/Auths/Lifecycle/Theorems.lean",
+ "formal/Auths/Observation.lean",
+ "formal/Auths/Product/Arithmetic.lean",
+ "formal/Auths/Product/CeilingCount.lean",
+ "formal/Auths/Product/Commitment.lean",
+ "formal/Auths/Product/ConnectionGenerations.lean",
+ "formal/Auths/Product/Eligibility.lean",
+ "formal/Auths/Product/Recovery.lean",
+ "formal/Auths/Product/Refinement.lean",
+ "formal/Auths/Product/Refinement/BoundedPolicy.lean",
+ "formal/Auths/Product/Refinement/Connections.lean",
+ "formal/Auths/Product/Refinement/Gateway.lean",
+ "formal/Auths/Product/RelativeCeiling.lean",
+ "formal/Auths/Product/RequestConstruction.lean",
+ "formal/Auths/Product/SubmitOrder.lean",
+ "formal/Auths/Product/Theorems.lean",
+ "formal/Auths/Product/Tightening.lean",
+ "formal/Auths/Refinement/Approval.lean",
+ "formal/Auths/Refinement/Observation.lean",
+ "formal/Auths/Refinement/Production.lean",
+ "formal/Auths/Rich/Mutations.lean",
+ "formal/Auths/Rich/Semantics.lean",
+ "formal/Auths/Rich/Theorems.lean",
+ "formal/Auths/Rich/Types.lean",
+ "formal/Auths/Theorems.lean",
+ "formal/algebra-contract-v1.toml",
+ "formal/lake-manifest.json",
+ "formal/lakefile.toml",
+ "formal/lean-toolchain",
+ "formal/qualification/aeneas/generated/algebra/Funs.lean",
+ "formal/qualification/aeneas/generated/algebra/Types.lean",
+ "formal/qualification/aeneas/generated/authority/Funs.lean",
+ "formal/qualification/aeneas/generated/authority/FunsExternal.lean",
+ "formal/qualification/aeneas/generated/authority/Types.lean",
+ "formal/qualification/aeneas/generated/authority/TypesExternal.lean",
+ "formal/qualification/aeneas/generated/bounded_policy/Funs.lean",
+ "formal/qualification/aeneas/generated/bounded_policy/Types.lean",
+ "formal/qualification/aeneas/generated/connections/Funs.lean",
+ "formal/qualification/aeneas/generated/connections/Types.lean",
+ "formal/qualification/aeneas/generated/gateway/Funs.lean",
+ "formal/qualification/aeneas/generated/gateway/Types.lean",
+ "formal/qualification/aeneas/generated/lifecycle/Funs.lean",
+ "formal/qualification/aeneas/generated/lifecycle/Types.lean",
+ "formal/qualification/aeneas/generated/model/Funs.lean",
+ "formal/qualification/aeneas/generated/model/FunsExternal.lean",
+ "formal/qualification/aeneas/generated/model/Types.lean",
+]
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
+scope = "Pure gateway recipe semantics: the recovery capability a recipe's declarations determine, closed request construction over every compiled plan and argument list, the attempt-record transition rule over every sequence of records, the admission order over every event trace, and the exact relative ceiling."
+residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
+toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
+axioms = ["propext"]
+
+[[claims.evidence]]
+kind = "lean-proof"
+artifact = "formal/Auths/Product/Recovery.lean"
+sha256 = "04c16abce67bebe1cb1a395a1a025a5c5cc9f84f6e2e34327bae0af88cb48d5c"
+
+[[claims]]
+claim_id = "AP-FORMAL-RICH-315"
+claim_text = "Lean proves the gateway recipe property: late response retains source."
+claim_status = "proved"
+lean_declaration = "Auths.Product.Recovery.late_response_retains_source"
+lean_statement_sha256 = "ceebf2b4398ba515895efd291b628ae9aeda3c65342bb40129ffc79b025a06ce"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
+rust_symbols = []
+semantic_source_closure = [
+ "formal/Auths.lean",
+ "formal/Auths/Approval.lean",
+ "formal/Auths/AssuranceAudit.lean",
+ "formal/Auths/Attenuation.lean",
+ "formal/Auths/Authority.lean",
+ "formal/Auths/Base.lean",
+ "formal/Auths/Composition.lean",
+ "formal/Auths/Diversity.lean",
+ "formal/Auths/ExtensionLaw.lean",
+ "formal/Auths/Generated/Algebra.lean",
+ "formal/Auths/Lifecycle/Refinement.lean",
+ "formal/Auths/Lifecycle/Semantics.lean",
+ "formal/Auths/Lifecycle/Theorems.lean",
+ "formal/Auths/Observation.lean",
+ "formal/Auths/Product/Arithmetic.lean",
+ "formal/Auths/Product/CeilingCount.lean",
+ "formal/Auths/Product/Commitment.lean",
+ "formal/Auths/Product/ConnectionGenerations.lean",
+ "formal/Auths/Product/Eligibility.lean",
+ "formal/Auths/Product/Recovery.lean",
+ "formal/Auths/Product/Refinement.lean",
+ "formal/Auths/Product/Refinement/BoundedPolicy.lean",
+ "formal/Auths/Product/Refinement/Connections.lean",
+ "formal/Auths/Product/Refinement/Gateway.lean",
+ "formal/Auths/Product/RelativeCeiling.lean",
+ "formal/Auths/Product/RequestConstruction.lean",
+ "formal/Auths/Product/SubmitOrder.lean",
+ "formal/Auths/Product/Theorems.lean",
+ "formal/Auths/Product/Tightening.lean",
+ "formal/Auths/Refinement/Approval.lean",
+ "formal/Auths/Refinement/Observation.lean",
+ "formal/Auths/Refinement/Production.lean",
+ "formal/Auths/Rich/Mutations.lean",
+ "formal/Auths/Rich/Semantics.lean",
+ "formal/Auths/Rich/Theorems.lean",
+ "formal/Auths/Rich/Types.lean",
+ "formal/Auths/Theorems.lean",
+ "formal/algebra-contract-v1.toml",
+ "formal/lake-manifest.json",
+ "formal/lakefile.toml",
+ "formal/lean-toolchain",
+ "formal/qualification/aeneas/generated/algebra/Funs.lean",
+ "formal/qualification/aeneas/generated/algebra/Types.lean",
+ "formal/qualification/aeneas/generated/authority/Funs.lean",
+ "formal/qualification/aeneas/generated/authority/FunsExternal.lean",
+ "formal/qualification/aeneas/generated/authority/Types.lean",
+ "formal/qualification/aeneas/generated/authority/TypesExternal.lean",
+ "formal/qualification/aeneas/generated/bounded_policy/Funs.lean",
+ "formal/qualification/aeneas/generated/bounded_policy/Types.lean",
+ "formal/qualification/aeneas/generated/connections/Funs.lean",
+ "formal/qualification/aeneas/generated/connections/Types.lean",
+ "formal/qualification/aeneas/generated/gateway/Funs.lean",
+ "formal/qualification/aeneas/generated/gateway/Types.lean",
+ "formal/qualification/aeneas/generated/lifecycle/Funs.lean",
+ "formal/qualification/aeneas/generated/lifecycle/Types.lean",
+ "formal/qualification/aeneas/generated/model/Funs.lean",
+ "formal/qualification/aeneas/generated/model/FunsExternal.lean",
+ "formal/qualification/aeneas/generated/model/Types.lean",
+]
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
+scope = "Pure gateway recipe semantics: the recovery capability a recipe's declarations determine, closed request construction over every compiled plan and argument list, the attempt-record transition rule over every sequence of records, the admission order over every event trace, and the exact relative ceiling."
+residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
+toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
+axioms = ["propext"]
+
+[[claims.evidence]]
+kind = "lean-proof"
+artifact = "formal/Auths/Product/Recovery.lean"
+sha256 = "04c16abce67bebe1cb1a395a1a025a5c5cc9f84f6e2e34327bae0af88cb48d5c"
+
+[[claims]]
+claim_id = "AP-FORMAL-RICH-316"
+claim_text = "Lean proves the gateway recipe property: late acknowledgement retains source."
+claim_status = "proved"
+lean_declaration = "Auths.Product.Recovery.late_acknowledgement_retains_source"
+lean_statement_sha256 = "d5c65641ea4deca74f99e636eb849180b5ade14f3212c5f04b67180f97eb1cb2"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
+rust_symbols = []
+semantic_source_closure = [
+ "formal/Auths.lean",
+ "formal/Auths/Approval.lean",
+ "formal/Auths/AssuranceAudit.lean",
+ "formal/Auths/Attenuation.lean",
+ "formal/Auths/Authority.lean",
+ "formal/Auths/Base.lean",
+ "formal/Auths/Composition.lean",
+ "formal/Auths/Diversity.lean",
+ "formal/Auths/ExtensionLaw.lean",
+ "formal/Auths/Generated/Algebra.lean",
+ "formal/Auths/Lifecycle/Refinement.lean",
+ "formal/Auths/Lifecycle/Semantics.lean",
+ "formal/Auths/Lifecycle/Theorems.lean",
+ "formal/Auths/Observation.lean",
+ "formal/Auths/Product/Arithmetic.lean",
+ "formal/Auths/Product/CeilingCount.lean",
+ "formal/Auths/Product/Commitment.lean",
+ "formal/Auths/Product/ConnectionGenerations.lean",
+ "formal/Auths/Product/Eligibility.lean",
+ "formal/Auths/Product/Recovery.lean",
+ "formal/Auths/Product/Refinement.lean",
+ "formal/Auths/Product/Refinement/BoundedPolicy.lean",
+ "formal/Auths/Product/Refinement/Connections.lean",
+ "formal/Auths/Product/Refinement/Gateway.lean",
+ "formal/Auths/Product/RelativeCeiling.lean",
+ "formal/Auths/Product/RequestConstruction.lean",
+ "formal/Auths/Product/SubmitOrder.lean",
+ "formal/Auths/Product/Theorems.lean",
+ "formal/Auths/Product/Tightening.lean",
+ "formal/Auths/Refinement/Approval.lean",
+ "formal/Auths/Refinement/Observation.lean",
+ "formal/Auths/Refinement/Production.lean",
+ "formal/Auths/Rich/Mutations.lean",
+ "formal/Auths/Rich/Semantics.lean",
+ "formal/Auths/Rich/Theorems.lean",
+ "formal/Auths/Rich/Types.lean",
+ "formal/Auths/Theorems.lean",
+ "formal/algebra-contract-v1.toml",
+ "formal/lake-manifest.json",
+ "formal/lakefile.toml",
+ "formal/lean-toolchain",
+ "formal/qualification/aeneas/generated/algebra/Funs.lean",
+ "formal/qualification/aeneas/generated/algebra/Types.lean",
+ "formal/qualification/aeneas/generated/authority/Funs.lean",
+ "formal/qualification/aeneas/generated/authority/FunsExternal.lean",
+ "formal/qualification/aeneas/generated/authority/Types.lean",
+ "formal/qualification/aeneas/generated/authority/TypesExternal.lean",
+ "formal/qualification/aeneas/generated/bounded_policy/Funs.lean",
+ "formal/qualification/aeneas/generated/bounded_policy/Types.lean",
+ "formal/qualification/aeneas/generated/connections/Funs.lean",
+ "formal/qualification/aeneas/generated/connections/Types.lean",
+ "formal/qualification/aeneas/generated/gateway/Funs.lean",
+ "formal/qualification/aeneas/generated/gateway/Types.lean",
+ "formal/qualification/aeneas/generated/lifecycle/Funs.lean",
+ "formal/qualification/aeneas/generated/lifecycle/Types.lean",
+ "formal/qualification/aeneas/generated/model/Funs.lean",
+ "formal/qualification/aeneas/generated/model/FunsExternal.lean",
+ "formal/qualification/aeneas/generated/model/Types.lean",
+]
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
+scope = "Pure gateway recipe semantics: the recovery capability a recipe's declarations determine, closed request construction over every compiled plan and argument list, the attempt-record transition rule over every sequence of records, the admission order over every event trace, and the exact relative ceiling."
+residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
+toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
+axioms = ["propext"]
+
+[[claims.evidence]]
+kind = "lean-proof"
+artifact = "formal/Auths/Product/Recovery.lean"
+sha256 = "04c16abce67bebe1cb1a395a1a025a5c5cc9f84f6e2e34327bae0af88cb48d5c"
[[claims]]
claim_id = "AP-FORMAL-RICH-216"
claim_text = "Lean proves the gateway recipe property: provider claim requires evidence."
claim_status = "proved"
lean_declaration = "Auths.Product.Recovery.provider_claim_requires_evidence"
-lean_statement_sha256 = "def91039f3a07a5683a962f3ee33d303addaae9a5689dda0f604fcd20b6167d0"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+lean_statement_sha256 = "7b62af78f9946db240bf59c1e04cd40c271023404a326a7478d20c3be87f33c4"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -13232,7 +13548,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure gateway recipe semantics: the recovery capability a recipe's declarations determine, closed request construction over every compiled plan and argument list, the attempt-record transition rule over every sequence of records, the admission order over every event trace, and the exact relative ceiling."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -13245,15 +13561,15 @@ axioms = [
[[claims.evidence]]
kind = "lean-proof"
artifact = "formal/Auths/Product/Recovery.lean"
-sha256 = "54b0c7fa26dfa03b5a9659875100a4b5bc845dfe010f73d0ac96bc981c208b70"
+sha256 = "04c16abce67bebe1cb1a395a1a025a5c5cc9f84f6e2e34327bae0af88cb48d5c"
[[claims]]
claim_id = "AP-FORMAL-RICH-217"
claim_text = "Lean proves the gateway recipe property: terminal stages final."
claim_status = "proved"
lean_declaration = "Auths.Product.Recovery.terminal_stages_final"
-lean_statement_sha256 = "9d46111d329a79a4788b23cc5a64f81e8853791104d67e3fd1992bf904f80684"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+lean_statement_sha256 = "65dc627394212edcac36e0f8391e85611724b9ba1c114b29c9dac2954c10daa8"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -13315,7 +13631,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure gateway recipe semantics: the recovery capability a recipe's declarations determine, closed request construction over every compiled plan and argument list, the attempt-record transition rule over every sequence of records, the admission order over every event trace, and the exact relative ceiling."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -13327,7 +13643,7 @@ axioms = [
[[claims.evidence]]
kind = "lean-proof"
artifact = "formal/Auths/Product/Recovery.lean"
-sha256 = "54b0c7fa26dfa03b5a9659875100a4b5bc845dfe010f73d0ac96bc981c208b70"
+sha256 = "04c16abce67bebe1cb1a395a1a025a5c5cc9f84f6e2e34327bae0af88cb48d5c"
[[claims]]
claim_id = "AP-FORMAL-RICH-218"
@@ -13335,7 +13651,7 @@ claim_text = "Lean proves the gateway recipe property: transitions preserve iden
claim_status = "proved"
lean_declaration = "Auths.Product.Recovery.transitions_preserve_identity"
lean_statement_sha256 = "ecbc5c29d6d4e0eeeb25fe5f4b2267aa8fc7b10bf316d501eac0e2f32af18bd1"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -13397,7 +13713,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure gateway recipe semantics: the recovery capability a recipe's declarations determine, closed request construction over every compiled plan and argument list, the attempt-record transition rule over every sequence of records, the admission order over every event trace, and the exact relative ceiling."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -13410,15 +13726,15 @@ axioms = [
[[claims.evidence]]
kind = "lean-proof"
artifact = "formal/Auths/Product/Recovery.lean"
-sha256 = "54b0c7fa26dfa03b5a9659875100a4b5bc845dfe010f73d0ac96bc981c208b70"
+sha256 = "04c16abce67bebe1cb1a395a1a025a5c5cc9f84f6e2e34327bae0af88cb48d5c"
[[claims]]
claim_id = "AP-FORMAL-RICH-219"
claim_text = "Lean proves the gateway recipe property: lease requires verified claim."
claim_status = "proved"
lean_declaration = "Auths.Product.SubmitOrder.lease_requires_verified_claim"
-lean_statement_sha256 = "c909a8624ee83d30cce1508a08b66b2f2844cffe645297855e5b6eb63f537092"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+lean_statement_sha256 = "ff698cb0bfd03fc9e57c0efe4ce166518ffb9d6750434de0ceeb957dd9a2341b"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -13480,7 +13796,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure gateway recipe semantics: the recovery capability a recipe's declarations determine, closed request construction over every compiled plan and argument list, the attempt-record transition rule over every sequence of records, the admission order over every event trace, and the exact relative ceiling."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -13492,15 +13808,15 @@ axioms = [
[[claims.evidence]]
kind = "lean-proof"
artifact = "formal/Auths/Product/SubmitOrder.lean"
-sha256 = "6b2fc0797bc4b8ccc5c5e53072c8388fc3267fc4131786cfdb761be02b50f722"
+sha256 = "c3a785617ab737b1a9df676404241163616a426827ffbb55c5e26369f59099dc"
[[claims]]
claim_id = "AP-FORMAL-RICH-220"
-claim_text = "Lean proves the gateway recipe property: send requires lease and deadline."
+claim_text = "Lean proves that sending requires a successful durable entry grant after native verification, reservation, credential lease, declared pre-entry checks, connection reload, and deadline validation."
claim_status = "proved"
lean_declaration = "Auths.Product.SubmitOrder.send_requires_lease_and_deadline"
-lean_statement_sha256 = "b7308523e47d7da1dd32e3799a7794b90d040b26648e64f8200af2bbbc3b93f3"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+lean_statement_sha256 = "726a3d5cb31a8fc6a47c978ceb36144dacad3dfc363ff931fd306cfc8cf5ba01"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -13562,7 +13878,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure gateway recipe semantics: the recovery capability a recipe's declarations determine, closed request construction over every compiled plan and argument list, the attempt-record transition rule over every sequence of records, the admission order over every event trace, and the exact relative ceiling."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -13574,15 +13890,15 @@ axioms = [
[[claims.evidence]]
kind = "lean-proof"
artifact = "formal/Auths/Product/SubmitOrder.lean"
-sha256 = "6b2fc0797bc4b8ccc5c5e53072c8388fc3267fc4131786cfdb761be02b50f722"
+sha256 = "c3a785617ab737b1a9df676404241163616a426827ffbb55c5e26369f59099dc"
[[claims]]
claim_id = "AP-FORMAL-RICH-221"
claim_text = "Lean proves the gateway recipe property: send requires credential checks."
claim_status = "proved"
lean_declaration = "Auths.Product.SubmitOrder.send_requires_credential_checks"
-lean_statement_sha256 = "dee08ab850d6d85d371a7e39dce3e204411a2e20d9619258d27bf67b313e5039"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+lean_statement_sha256 = "f704e163d11883021606a89f21440d172e6d0e06b87e91b9c4f0214dd7496bdf"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -13644,7 +13960,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure gateway recipe semantics: the recovery capability a recipe's declarations determine, closed request construction over every compiled plan and argument list, the attempt-record transition rule over every sequence of records, the admission order over every event trace, and the exact relative ceiling."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -13656,7 +13972,7 @@ axioms = [
[[claims.evidence]]
kind = "lean-proof"
artifact = "formal/Auths/Product/SubmitOrder.lean"
-sha256 = "6b2fc0797bc4b8ccc5c5e53072c8388fc3267fc4131786cfdb761be02b50f722"
+sha256 = "c3a785617ab737b1a9df676404241163616a426827ffbb55c5e26369f59099dc"
[[claims]]
claim_id = "AP-FORMAL-RICH-222"
@@ -13664,7 +13980,7 @@ claim_text = "Lean proves the gateway recipe property: send requires relative ce
claim_status = "proved"
lean_declaration = "Auths.Product.SubmitOrder.send_requires_relative_ceiling"
lean_statement_sha256 = "bb8d3c138ad18d9b9963c22bf746dbf1de793b9a6f64a37658db9b8854d22986"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -13726,7 +14042,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure gateway recipe semantics: the recovery capability a recipe's declarations determine, closed request construction over every compiled plan and argument list, the attempt-record transition rule over every sequence of records, the admission order over every event trace, and the exact relative ceiling."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -13738,7 +14054,7 @@ axioms = [
[[claims.evidence]]
kind = "lean-proof"
artifact = "formal/Auths/Product/SubmitOrder.lean"
-sha256 = "6b2fc0797bc4b8ccc5c5e53072c8388fc3267fc4131786cfdb761be02b50f722"
+sha256 = "c3a785617ab737b1a9df676404241163616a426827ffbb55c5e26369f59099dc"
[[claims]]
claim_id = "AP-FORMAL-RICH-223"
@@ -13746,7 +14062,7 @@ claim_text = "Lean proves the gateway recipe property: account scope requires bi
claim_status = "proved"
lean_declaration = "Auths.Product.SubmitOrder.account_scope_requires_binding"
lean_statement_sha256 = "246b93cf15db6b83675a162d97172251edb9bef8f6d9131fd4cecc7884ce6914"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -13808,7 +14124,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure gateway recipe semantics: the recovery capability a recipe's declarations determine, closed request construction over every compiled plan and argument list, the attempt-record transition rule over every sequence of records, the admission order over every event trace, and the exact relative ceiling."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -13820,7 +14136,7 @@ axioms = [
[[claims.evidence]]
kind = "lean-proof"
artifact = "formal/Auths/Product/SubmitOrder.lean"
-sha256 = "6b2fc0797bc4b8ccc5c5e53072c8388fc3267fc4131786cfdb761be02b50f722"
+sha256 = "c3a785617ab737b1a9df676404241163616a426827ffbb55c5e26369f59099dc"
[[claims]]
claim_id = "AP-FORMAL-RICH-224"
@@ -13828,7 +14144,7 @@ claim_text = "Lean proves the gateway recipe property: at most one send."
claim_status = "proved"
lean_declaration = "Auths.Product.SubmitOrder.at_most_one_send"
lean_statement_sha256 = "19162c884a8ed60f9a15dfbb26e95aa35015ede017313498984ffada67a3ff50"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -13890,7 +14206,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure gateway recipe semantics: the recovery capability a recipe's declarations determine, closed request construction over every compiled plan and argument list, the attempt-record transition rule over every sequence of records, the admission order over every event trace, and the exact relative ceiling."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -13902,7 +14218,244 @@ axioms = [
[[claims.evidence]]
kind = "lean-proof"
artifact = "formal/Auths/Product/SubmitOrder.lean"
-sha256 = "6b2fc0797bc4b8ccc5c5e53072c8388fc3267fc4131786cfdb761be02b50f722"
+sha256 = "c3a785617ab737b1a9df676404241163616a426827ffbb55c5e26369f59099dc"
+
+[[claims]]
+claim_id = "AP-FORMAL-RICH-317"
+claim_text = "Lean proves the gateway recipe property: response stops without read."
+claim_status = "proved"
+lean_declaration = "Auths.Product.SubmitOrder.response_stops_without_read"
+lean_statement_sha256 = "b41dcd2e88d39fdb460f461f623c58be530fd77e3ec7cc28a55dbf4c6e07b032"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
+rust_symbols = []
+semantic_source_closure = [
+ "formal/Auths.lean",
+ "formal/Auths/Approval.lean",
+ "formal/Auths/AssuranceAudit.lean",
+ "formal/Auths/Attenuation.lean",
+ "formal/Auths/Authority.lean",
+ "formal/Auths/Base.lean",
+ "formal/Auths/Composition.lean",
+ "formal/Auths/Diversity.lean",
+ "formal/Auths/ExtensionLaw.lean",
+ "formal/Auths/Generated/Algebra.lean",
+ "formal/Auths/Lifecycle/Refinement.lean",
+ "formal/Auths/Lifecycle/Semantics.lean",
+ "formal/Auths/Lifecycle/Theorems.lean",
+ "formal/Auths/Observation.lean",
+ "formal/Auths/Product/Arithmetic.lean",
+ "formal/Auths/Product/CeilingCount.lean",
+ "formal/Auths/Product/Commitment.lean",
+ "formal/Auths/Product/ConnectionGenerations.lean",
+ "formal/Auths/Product/Eligibility.lean",
+ "formal/Auths/Product/Recovery.lean",
+ "formal/Auths/Product/Refinement.lean",
+ "formal/Auths/Product/Refinement/BoundedPolicy.lean",
+ "formal/Auths/Product/Refinement/Connections.lean",
+ "formal/Auths/Product/Refinement/Gateway.lean",
+ "formal/Auths/Product/RelativeCeiling.lean",
+ "formal/Auths/Product/RequestConstruction.lean",
+ "formal/Auths/Product/SubmitOrder.lean",
+ "formal/Auths/Product/Theorems.lean",
+ "formal/Auths/Product/Tightening.lean",
+ "formal/Auths/Refinement/Approval.lean",
+ "formal/Auths/Refinement/Observation.lean",
+ "formal/Auths/Refinement/Production.lean",
+ "formal/Auths/Rich/Mutations.lean",
+ "formal/Auths/Rich/Semantics.lean",
+ "formal/Auths/Rich/Theorems.lean",
+ "formal/Auths/Rich/Types.lean",
+ "formal/Auths/Theorems.lean",
+ "formal/algebra-contract-v1.toml",
+ "formal/lake-manifest.json",
+ "formal/lakefile.toml",
+ "formal/lean-toolchain",
+ "formal/qualification/aeneas/generated/algebra/Funs.lean",
+ "formal/qualification/aeneas/generated/algebra/Types.lean",
+ "formal/qualification/aeneas/generated/authority/Funs.lean",
+ "formal/qualification/aeneas/generated/authority/FunsExternal.lean",
+ "formal/qualification/aeneas/generated/authority/Types.lean",
+ "formal/qualification/aeneas/generated/authority/TypesExternal.lean",
+ "formal/qualification/aeneas/generated/bounded_policy/Funs.lean",
+ "formal/qualification/aeneas/generated/bounded_policy/Types.lean",
+ "formal/qualification/aeneas/generated/connections/Funs.lean",
+ "formal/qualification/aeneas/generated/connections/Types.lean",
+ "formal/qualification/aeneas/generated/gateway/Funs.lean",
+ "formal/qualification/aeneas/generated/gateway/Types.lean",
+ "formal/qualification/aeneas/generated/lifecycle/Funs.lean",
+ "formal/qualification/aeneas/generated/lifecycle/Types.lean",
+ "formal/qualification/aeneas/generated/model/Funs.lean",
+ "formal/qualification/aeneas/generated/model/FunsExternal.lean",
+ "formal/qualification/aeneas/generated/model/Types.lean",
+]
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
+scope = "Pure gateway recipe semantics: the recovery capability a recipe's declarations determine, closed request construction over every compiled plan and argument list, the attempt-record transition rule over every sequence of records, the admission order over every event trace, and the exact relative ceiling."
+residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
+toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
+axioms = ["propext"]
+
+[[claims.evidence]]
+kind = "lean-proof"
+artifact = "formal/Auths/Product/SubmitOrder.lean"
+sha256 = "c3a785617ab737b1a9df676404241163616a426827ffbb55c5e26369f59099dc"
+
+[[claims]]
+claim_id = "AP-FORMAL-RICH-318"
+claim_text = "Lean proves the gateway recipe property: replay returns stored."
+claim_status = "proved"
+lean_declaration = "Auths.Product.SubmitOrder.replay_returns_stored"
+lean_statement_sha256 = "7bcd77b4b74a511e910a7f93a59fa82bdb64b0bef69bcab062fd7c797eeaed13"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
+rust_symbols = []
+semantic_source_closure = [
+ "formal/Auths.lean",
+ "formal/Auths/Approval.lean",
+ "formal/Auths/AssuranceAudit.lean",
+ "formal/Auths/Attenuation.lean",
+ "formal/Auths/Authority.lean",
+ "formal/Auths/Base.lean",
+ "formal/Auths/Composition.lean",
+ "formal/Auths/Diversity.lean",
+ "formal/Auths/ExtensionLaw.lean",
+ "formal/Auths/Generated/Algebra.lean",
+ "formal/Auths/Lifecycle/Refinement.lean",
+ "formal/Auths/Lifecycle/Semantics.lean",
+ "formal/Auths/Lifecycle/Theorems.lean",
+ "formal/Auths/Observation.lean",
+ "formal/Auths/Product/Arithmetic.lean",
+ "formal/Auths/Product/CeilingCount.lean",
+ "formal/Auths/Product/Commitment.lean",
+ "formal/Auths/Product/ConnectionGenerations.lean",
+ "formal/Auths/Product/Eligibility.lean",
+ "formal/Auths/Product/Recovery.lean",
+ "formal/Auths/Product/Refinement.lean",
+ "formal/Auths/Product/Refinement/BoundedPolicy.lean",
+ "formal/Auths/Product/Refinement/Connections.lean",
+ "formal/Auths/Product/Refinement/Gateway.lean",
+ "formal/Auths/Product/RelativeCeiling.lean",
+ "formal/Auths/Product/RequestConstruction.lean",
+ "formal/Auths/Product/SubmitOrder.lean",
+ "formal/Auths/Product/Theorems.lean",
+ "formal/Auths/Product/Tightening.lean",
+ "formal/Auths/Refinement/Approval.lean",
+ "formal/Auths/Refinement/Observation.lean",
+ "formal/Auths/Refinement/Production.lean",
+ "formal/Auths/Rich/Mutations.lean",
+ "formal/Auths/Rich/Semantics.lean",
+ "formal/Auths/Rich/Theorems.lean",
+ "formal/Auths/Rich/Types.lean",
+ "formal/Auths/Theorems.lean",
+ "formal/algebra-contract-v1.toml",
+ "formal/lake-manifest.json",
+ "formal/lakefile.toml",
+ "formal/lean-toolchain",
+ "formal/qualification/aeneas/generated/algebra/Funs.lean",
+ "formal/qualification/aeneas/generated/algebra/Types.lean",
+ "formal/qualification/aeneas/generated/authority/Funs.lean",
+ "formal/qualification/aeneas/generated/authority/FunsExternal.lean",
+ "formal/qualification/aeneas/generated/authority/Types.lean",
+ "formal/qualification/aeneas/generated/authority/TypesExternal.lean",
+ "formal/qualification/aeneas/generated/bounded_policy/Funs.lean",
+ "formal/qualification/aeneas/generated/bounded_policy/Types.lean",
+ "formal/qualification/aeneas/generated/connections/Funs.lean",
+ "formal/qualification/aeneas/generated/connections/Types.lean",
+ "formal/qualification/aeneas/generated/gateway/Funs.lean",
+ "formal/qualification/aeneas/generated/gateway/Types.lean",
+ "formal/qualification/aeneas/generated/lifecycle/Funs.lean",
+ "formal/qualification/aeneas/generated/lifecycle/Types.lean",
+ "formal/qualification/aeneas/generated/model/Funs.lean",
+ "formal/qualification/aeneas/generated/model/FunsExternal.lean",
+ "formal/qualification/aeneas/generated/model/Types.lean",
+]
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
+scope = "Pure gateway recipe semantics: the recovery capability a recipe's declarations determine, closed request construction over every compiled plan and argument list, the attempt-record transition rule over every sequence of records, the admission order over every event trace, and the exact relative ceiling."
+residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
+toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
+axioms = ["propext"]
+
+[[claims.evidence]]
+kind = "lean-proof"
+artifact = "formal/Auths/Product/SubmitOrder.lean"
+sha256 = "c3a785617ab737b1a9df676404241163616a426827ffbb55c5e26369f59099dc"
+
+[[claims]]
+claim_id = "AP-FORMAL-RICH-319"
+claim_text = "Lean proves the gateway recipe property: refused grant stops without send."
+claim_status = "proved"
+lean_declaration = "Auths.Product.SubmitOrder.refused_grant_stops_without_send"
+lean_statement_sha256 = "4c8415d14c2229561309e9c1e978adfb395285233fd507d0eed3f7ed8a76754c"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
+rust_symbols = []
+semantic_source_closure = [
+ "formal/Auths.lean",
+ "formal/Auths/Approval.lean",
+ "formal/Auths/AssuranceAudit.lean",
+ "formal/Auths/Attenuation.lean",
+ "formal/Auths/Authority.lean",
+ "formal/Auths/Base.lean",
+ "formal/Auths/Composition.lean",
+ "formal/Auths/Diversity.lean",
+ "formal/Auths/ExtensionLaw.lean",
+ "formal/Auths/Generated/Algebra.lean",
+ "formal/Auths/Lifecycle/Refinement.lean",
+ "formal/Auths/Lifecycle/Semantics.lean",
+ "formal/Auths/Lifecycle/Theorems.lean",
+ "formal/Auths/Observation.lean",
+ "formal/Auths/Product/Arithmetic.lean",
+ "formal/Auths/Product/CeilingCount.lean",
+ "formal/Auths/Product/Commitment.lean",
+ "formal/Auths/Product/ConnectionGenerations.lean",
+ "formal/Auths/Product/Eligibility.lean",
+ "formal/Auths/Product/Recovery.lean",
+ "formal/Auths/Product/Refinement.lean",
+ "formal/Auths/Product/Refinement/BoundedPolicy.lean",
+ "formal/Auths/Product/Refinement/Connections.lean",
+ "formal/Auths/Product/Refinement/Gateway.lean",
+ "formal/Auths/Product/RelativeCeiling.lean",
+ "formal/Auths/Product/RequestConstruction.lean",
+ "formal/Auths/Product/SubmitOrder.lean",
+ "formal/Auths/Product/Theorems.lean",
+ "formal/Auths/Product/Tightening.lean",
+ "formal/Auths/Refinement/Approval.lean",
+ "formal/Auths/Refinement/Observation.lean",
+ "formal/Auths/Refinement/Production.lean",
+ "formal/Auths/Rich/Mutations.lean",
+ "formal/Auths/Rich/Semantics.lean",
+ "formal/Auths/Rich/Theorems.lean",
+ "formal/Auths/Rich/Types.lean",
+ "formal/Auths/Theorems.lean",
+ "formal/algebra-contract-v1.toml",
+ "formal/lake-manifest.json",
+ "formal/lakefile.toml",
+ "formal/lean-toolchain",
+ "formal/qualification/aeneas/generated/algebra/Funs.lean",
+ "formal/qualification/aeneas/generated/algebra/Types.lean",
+ "formal/qualification/aeneas/generated/authority/Funs.lean",
+ "formal/qualification/aeneas/generated/authority/FunsExternal.lean",
+ "formal/qualification/aeneas/generated/authority/Types.lean",
+ "formal/qualification/aeneas/generated/authority/TypesExternal.lean",
+ "formal/qualification/aeneas/generated/bounded_policy/Funs.lean",
+ "formal/qualification/aeneas/generated/bounded_policy/Types.lean",
+ "formal/qualification/aeneas/generated/connections/Funs.lean",
+ "formal/qualification/aeneas/generated/connections/Types.lean",
+ "formal/qualification/aeneas/generated/gateway/Funs.lean",
+ "formal/qualification/aeneas/generated/gateway/Types.lean",
+ "formal/qualification/aeneas/generated/lifecycle/Funs.lean",
+ "formal/qualification/aeneas/generated/lifecycle/Types.lean",
+ "formal/qualification/aeneas/generated/model/Funs.lean",
+ "formal/qualification/aeneas/generated/model/FunsExternal.lean",
+ "formal/qualification/aeneas/generated/model/Types.lean",
+]
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
+scope = "Pure gateway recipe semantics: the recovery capability a recipe's declarations determine, closed request construction over every compiled plan and argument list, the attempt-record transition rule over every sequence of records, the admission order over every event trace, and the exact relative ceiling."
+residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
+toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
+axioms = ["propext"]
+
+[[claims.evidence]]
+kind = "lean-proof"
+artifact = "formal/Auths/Product/SubmitOrder.lean"
+sha256 = "c3a785617ab737b1a9df676404241163616a426827ffbb55c5e26369f59099dc"
[[claims]]
claim_id = "AP-FORMAL-RICH-225"
@@ -13910,7 +14463,7 @@ claim_text = "Lean proves the gateway recipe property: pre claim refusal stores
claim_status = "proved"
lean_declaration = "Auths.Product.SubmitOrder.pre_claim_refusal_stores_nothing"
lean_statement_sha256 = "d8266ccd00249eba927b1fbbf7e505db7e3ca06f100684c3b07d2fdc8006ec30"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -13972,7 +14525,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure gateway recipe semantics: the recovery capability a recipe's declarations determine, closed request construction over every compiled plan and argument list, the attempt-record transition rule over every sequence of records, the admission order over every event trace, and the exact relative ceiling."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -13984,7 +14537,7 @@ axioms = [
[[claims.evidence]]
kind = "lean-proof"
artifact = "formal/Auths/Product/SubmitOrder.lean"
-sha256 = "6b2fc0797bc4b8ccc5c5e53072c8388fc3267fc4131786cfdb761be02b50f722"
+sha256 = "c3a785617ab737b1a9df676404241163616a426827ffbb55c5e26369f59099dc"
[[claims]]
claim_id = "AP-FORMAL-RICH-226"
@@ -13992,7 +14545,7 @@ claim_text = "Lean proves the gateway recipe property: not entered refusal table
claim_status = "proved"
lean_declaration = "Auths.Product.SubmitOrder.not_entered_refusal_table"
lean_statement_sha256 = "177dc8c4accfb1fced3209eccd723ab3b00afdf8b4c71396393a772a862ff947"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -14054,7 +14607,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure gateway recipe semantics: the recovery capability a recipe's declarations determine, closed request construction over every compiled plan and argument list, the attempt-record transition rule over every sequence of records, the admission order over every event trace, and the exact relative ceiling."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -14066,15 +14619,15 @@ axioms = [
[[claims.evidence]]
kind = "lean-proof"
artifact = "formal/Auths/Product/SubmitOrder.lean"
-sha256 = "6b2fc0797bc4b8ccc5c5e53072c8388fc3267fc4131786cfdb761be02b50f722"
+sha256 = "c3a785617ab737b1a9df676404241163616a426827ffbb55c5e26369f59099dc"
[[claims]]
claim_id = "AP-FORMAL-RICH-227"
claim_text = "Lean proves the gateway recipe property: post claim refusal records not entered."
claim_status = "proved"
lean_declaration = "Auths.Product.SubmitOrder.post_claim_refusal_records_not_entered"
-lean_statement_sha256 = "192a3f34693f8a77fa76cfd8218a26fb32cd4e000c260923dc49f212bcda9ebe"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+lean_statement_sha256 = "6608158b0cdeba4e352b3b9c138f5708428ebb0bbefbeea7e4dcc19b65962cb4"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -14136,7 +14689,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure gateway recipe semantics: the recovery capability a recipe's declarations determine, closed request construction over every compiled plan and argument list, the attempt-record transition rule over every sequence of records, the admission order over every event trace, and the exact relative ceiling."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -14148,7 +14701,7 @@ axioms = [
[[claims.evidence]]
kind = "lean-proof"
artifact = "formal/Auths/Product/SubmitOrder.lean"
-sha256 = "6b2fc0797bc4b8ccc5c5e53072c8388fc3267fc4131786cfdb761be02b50f722"
+sha256 = "c3a785617ab737b1a9df676404241163616a426827ffbb55c5e26369f59099dc"
[[claims]]
claim_id = "AP-FORMAL-RICH-228"
@@ -14156,7 +14709,7 @@ claim_text = "Lean proves the gateway recipe property: relative ceiling exact."
claim_status = "proved"
lean_declaration = "Auths.Product.RelativeCeiling.relative_ceiling_exact"
lean_statement_sha256 = "ccaec2583a29ff660d59afea3cbc28bc6c2d81811b173ca661a752950597cf61"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -14218,7 +14771,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure gateway recipe semantics: the recovery capability a recipe's declarations determine, closed request construction over every compiled plan and argument list, the attempt-record transition rule over every sequence of records, the admission order over every event trace, and the exact relative ceiling."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -14235,7 +14788,7 @@ claim_text = "Lean proves the gateway recipe property: relative ceiling never ex
claim_status = "proved"
lean_declaration = "Auths.Product.RelativeCeiling.relative_ceiling_never_exceeds_ratio"
lean_statement_sha256 = "ddc05e7aac856e77fe875dc9327bb2f3de7d3c89a1c998ba9d0579239d51e192"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -14297,7 +14850,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure gateway recipe semantics: the recovery capability a recipe's declarations determine, closed request construction over every compiled plan and argument list, the attempt-record transition rule over every sequence of records, the admission order over every event trace, and the exact relative ceiling."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -14314,7 +14867,7 @@ claim_text = "Lean proves the gateway recipe property: relative ceiling monotone
claim_status = "proved"
lean_declaration = "Auths.Product.RelativeCeiling.relative_ceiling_monotone"
lean_statement_sha256 = "0d104c40d0dd9c30ee55e8a8708d8c13f59c95a81c2ef0fea5335ca5db706848"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -14376,7 +14929,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure gateway recipe semantics: the recovery capability a recipe's declarations determine, closed request construction over every compiled plan and argument list, the attempt-record transition rule over every sequence of records, the admission order over every event trace, and the exact relative ceiling."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -14397,7 +14950,7 @@ claim_text = "Lean proves the gateway recipe property: relative basis never nega
claim_status = "proved"
lean_declaration = "Auths.Product.RelativeCeiling.relative_basis_never_negative"
lean_statement_sha256 = "4ed5440adb7ca11dfa8288e59aaad29468d967cb1c8654cfc5d3331fbbc57b4b"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -14459,7 +15012,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure gateway recipe semantics: the recovery capability a recipe's declarations determine, closed request construction over every compiled plan and argument list, the attempt-record transition rule over every sequence of records, the admission order over every event trace, and the exact relative ceiling."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -14480,7 +15033,7 @@ claim_text = "Lean proves the gateway recipe property: method and origin fixed."
claim_status = "proved"
lean_declaration = "Auths.Product.RequestConstruction.method_and_origin_fixed"
lean_statement_sha256 = "7bcf7156c75568a5207b8689d2b3be7a890a5007f80deb6dfb4242603d1abd5e"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -14542,7 +15095,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure gateway recipe semantics: the recovery capability a recipe's declarations determine, and closed request construction over every compiled plan and argument list."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -14559,7 +15112,7 @@ claim_text = "Lean proves the gateway recipe property: path from declared segmen
claim_status = "proved"
lean_declaration = "Auths.Product.RequestConstruction.path_from_declared_segments"
lean_statement_sha256 = "80f90e7a0f1a764d352c074a03cb29c5a8f0190e431604b216eca137f5cdce0a"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -14621,7 +15174,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure gateway recipe semantics: the recovery capability a recipe's declarations determine, and closed request construction over every compiled plan and argument list."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -14641,7 +15194,7 @@ claim_text = "Lean proves the gateway recipe property: headers within declaratio
claim_status = "proved"
lean_declaration = "Auths.Product.RequestConstruction.headers_within_declaration"
lean_statement_sha256 = "1767a7004a5b634fa8269968caa485cf114ffcdf718ae1c2a1218f06c950ce98"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -14703,7 +15256,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure gateway recipe semantics: the recovery capability a recipe's declarations determine, and closed request construction over every compiled plan and argument list."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -14720,7 +15273,7 @@ claim_text = "Lean proves the gateway recipe property: account scope header exac
claim_status = "proved"
lean_declaration = "Auths.Product.RequestConstruction.account_scope_header_exact"
lean_statement_sha256 = "10c9fb25aece47a2f6c82a3c90711b6fe27ee7b0b0ab111e86168e1548ec93c5"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -14782,7 +15335,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure gateway recipe semantics: the recovery capability a recipe's declarations determine, and closed request construction over every compiled plan and argument list."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -14799,7 +15352,7 @@ claim_text = "Lean proves the gateway recipe property: credential reads fixed."
claim_status = "proved"
lean_declaration = "Auths.Product.RequestConstruction.credential_reads_fixed"
lean_statement_sha256 = "02766f5e9149ca84b35952a07902a0d150524636fc1f4d9c50789dd03a159abb"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -14861,7 +15414,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure gateway recipe semantics: the recovery capability a recipe's declarations determine, and closed request construction over every compiled plan and argument list."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -14878,7 +15431,7 @@ claim_text = "Lean proves the gateway recipe property: body bounded."
claim_status = "proved"
lean_declaration = "Auths.Product.RequestConstruction.body_bounded"
lean_statement_sha256 = "b9ff16f851122675e417ffb5cf618b0960a474aba638473d9fd6c139cb4fa182"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -14940,7 +15493,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure gateway recipe semantics: the recovery capability a recipe's declarations determine, and closed request construction over every compiled plan and argument list."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -14960,7 +15513,7 @@ claim_text = "The mechanically translated initial submit state returns exactly t
claim_status = "qualified"
lean_declaration = "Auths.Product.Refinement.Gateway.translated_start_refines_model"
lean_statement_sha256 = "0352a3ca2e917c34188e99770f714387e9fda7d0420d45b399c294fae164fce5"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_gateway_kernel::order::start"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -15022,7 +15575,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Every submit plan."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, and the listed foundational axioms are trusted; the theorem has no premise.",
@@ -15038,17 +15591,17 @@ axioms = [
[[claims.evidence]]
kind = "lean-refinement"
artifact = "formal/Auths/Product/Refinement/Gateway.lean"
-sha256 = "59a2ad78eca15d40e4ca249d653cdbb870131ee2f24446bbff2f5f6144c72e13"
+sha256 = "25a1770ec00c4ddcdaaa6cd6995d4c1e47afad17bc7e3164bf1bc4b0babafed6"
[[claims.evidence]]
kind = "mechanical-translation"
artifact = "formal/qualification/aeneas/generated/gateway/Funs.lean"
-sha256 = "c8c86db058f313ab46ee86370d79ee70130c1ead3860dd7dabd49f2a89af1636"
+sha256 = "8c6c2ab1a63280e94f4084566888d727b13a5890b9de4c75aa90322406c5c601"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-211"
@@ -15056,7 +15609,7 @@ claim_text = "The mechanically translated recovery-capability leaf returns exact
claim_status = "qualified"
lean_declaration = "Auths.Product.Refinement.Gateway.translated_recovery_capability_refines_model"
lean_statement_sha256 = "427ce8baf2a9bf87c27a0b185a87ef6d7cf57dfaaf3b89f6e43bfd50e7de8c92"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_gateway_kernel::recovery::recovery_capability"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -15118,7 +15671,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Every observation locator, echo flag, idempotency declaration, and pre-entry flag."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the listed foundational axioms, and the theorem's explicit representation-validity premise (every buffer the leaf builds is shorter than u32::MAX bytes) are trusted.",
@@ -15133,17 +15686,17 @@ axioms = [
[[claims.evidence]]
kind = "lean-refinement"
artifact = "formal/Auths/Product/Refinement/Gateway.lean"
-sha256 = "59a2ad78eca15d40e4ca249d653cdbb870131ee2f24446bbff2f5f6144c72e13"
+sha256 = "25a1770ec00c4ddcdaaa6cd6995d4c1e47afad17bc7e3164bf1bc4b0babafed6"
[[claims.evidence]]
kind = "mechanical-translation"
artifact = "formal/qualification/aeneas/generated/gateway/Funs.lean"
-sha256 = "c8c86db058f313ab46ee86370d79ee70130c1ead3860dd7dabd49f2a89af1636"
+sha256 = "8c6c2ab1a63280e94f4084566888d727b13a5890b9de4c75aa90322406c5c601"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-212"
@@ -15151,7 +15704,7 @@ claim_text = "The mechanically translated write construction returns exactly the
claim_status = "qualified"
lean_declaration = "Auths.Product.Refinement.Gateway.translated_construct_write_refines_model"
lean_statement_sha256 = "4a1b4fdc57bc755532f723f0e489a52178d1e1e905b9665f6c5f92cb0afeb155"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_gateway_kernel::construct::construct_write"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -15213,7 +15766,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Every write plan and argument list: path validation and percent-encoding, the canonical JSON or form body with the echo, version, account-scope, and idempotency headers, and the path and body bounds."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the listed foundational axioms, and the theorem's explicit representation-validity premise (every buffer the leaf builds is shorter than u32::MAX bytes) are trusted.",
@@ -15229,17 +15782,17 @@ axioms = [
[[claims.evidence]]
kind = "lean-refinement"
artifact = "formal/Auths/Product/Refinement/Gateway.lean"
-sha256 = "59a2ad78eca15d40e4ca249d653cdbb870131ee2f24446bbff2f5f6144c72e13"
+sha256 = "25a1770ec00c4ddcdaaa6cd6995d4c1e47afad17bc7e3164bf1bc4b0babafed6"
[[claims.evidence]]
kind = "mechanical-translation"
artifact = "formal/qualification/aeneas/generated/gateway/Funs.lean"
-sha256 = "c8c86db058f313ab46ee86370d79ee70130c1ead3860dd7dabd49f2a89af1636"
+sha256 = "8c6c2ab1a63280e94f4084566888d727b13a5890b9de4c75aa90322406c5c601"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-213"
@@ -15247,7 +15800,7 @@ claim_text = "The mechanically translated action-read construction returns exact
claim_status = "qualified"
lean_declaration = "Auths.Product.Refinement.Gateway.translated_construct_action_read_refines_model"
lean_statement_sha256 = "bb5c93a4929e5313fe181638a0bd81b0e4f2dffb612feed83dc75a3f4434fc22"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_gateway_kernel::construct::construct_action_read"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -15309,7 +15862,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Every action-read plan and argument list: the encoded path and the version and account-scope headers."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the listed foundational axioms, and the theorem's explicit representation-validity premise (every buffer the leaf builds is shorter than u32::MAX bytes) are trusted.",
@@ -15325,17 +15878,17 @@ axioms = [
[[claims.evidence]]
kind = "lean-refinement"
artifact = "formal/Auths/Product/Refinement/Gateway.lean"
-sha256 = "59a2ad78eca15d40e4ca249d653cdbb870131ee2f24446bbff2f5f6144c72e13"
+sha256 = "25a1770ec00c4ddcdaaa6cd6995d4c1e47afad17bc7e3164bf1bc4b0babafed6"
[[claims.evidence]]
kind = "mechanical-translation"
artifact = "formal/qualification/aeneas/generated/gateway/Funs.lean"
-sha256 = "c8c86db058f313ab46ee86370d79ee70130c1ead3860dd7dabd49f2a89af1636"
+sha256 = "8c6c2ab1a63280e94f4084566888d727b13a5890b9de4c75aa90322406c5c601"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-214"
@@ -15343,7 +15896,7 @@ claim_text = "The mechanically translated credential-read construction returns e
claim_status = "qualified"
lean_declaration = "Auths.Product.Refinement.Gateway.translated_construct_credential_read_refines_model"
lean_statement_sha256 = "9e69f024fad8ed4a8f21ccaf9bab552a377296555f66063bed13270869dab59f"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_gateway_kernel::construct::construct_credential_read"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -15405,7 +15958,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Every credential-read plan: the fixed path and the version headers."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the listed foundational axioms, and the theorem's explicit representation-validity premise (every buffer the leaf builds is shorter than u32::MAX bytes) are trusted.",
@@ -15421,17 +15974,17 @@ axioms = [
[[claims.evidence]]
kind = "lean-refinement"
artifact = "formal/Auths/Product/Refinement/Gateway.lean"
-sha256 = "59a2ad78eca15d40e4ca249d653cdbb870131ee2f24446bbff2f5f6144c72e13"
+sha256 = "25a1770ec00c4ddcdaaa6cd6995d4c1e47afad17bc7e3164bf1bc4b0babafed6"
[[claims.evidence]]
kind = "mechanical-translation"
artifact = "formal/qualification/aeneas/generated/gateway/Funs.lean"
-sha256 = "c8c86db058f313ab46ee86370d79ee70130c1ead3860dd7dabd49f2a89af1636"
+sha256 = "8c6c2ab1a63280e94f4084566888d727b13a5890b9de4c75aa90322406c5c601"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-232"
@@ -15439,7 +15992,7 @@ claim_text = "The mechanically translated admission-order step machine returns e
claim_status = "qualified"
lean_declaration = "Auths.Product.Refinement.Gateway.translated_next_step_refines_model"
lean_statement_sha256 = "4f6f78fb06ad5e6bdabeafe6530141df9d68a0fbedaa02213e7e830e6c491989"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = [
"auths_gateway_kernel::order::next_step",
"auths_gateway_kernel::ratio::relative_ceiling_admits",
@@ -15504,7 +16057,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Every plan, phase, verified argument, and event, including the denied-read index arithmetic and the relative-ceiling comparison on the verified argument."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, and the listed foundational axioms are trusted; the theorem has no premise.",
@@ -15520,17 +16073,17 @@ axioms = [
[[claims.evidence]]
kind = "lean-refinement"
artifact = "formal/Auths/Product/Refinement/Gateway.lean"
-sha256 = "59a2ad78eca15d40e4ca249d653cdbb870131ee2f24446bbff2f5f6144c72e13"
+sha256 = "25a1770ec00c4ddcdaaa6cd6995d4c1e47afad17bc7e3164bf1bc4b0babafed6"
[[claims.evidence]]
kind = "mechanical-translation"
artifact = "formal/qualification/aeneas/generated/gateway/Funs.lean"
-sha256 = "c8c86db058f313ab46ee86370d79ee70130c1ead3860dd7dabd49f2a89af1636"
+sha256 = "8c6c2ab1a63280e94f4084566888d727b13a5890b9de4c75aa90322406c5c601"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-233"
@@ -15538,7 +16091,7 @@ claim_text = "The mechanically translated attempt-record transition rule returns
claim_status = "qualified"
lean_declaration = "Auths.Product.Refinement.Gateway.translated_valid_transition_refines_model"
lean_statement_sha256 = "a112e2fa7fad786aa452f674aeca7331f57dc51d5d01c9023571e09b6c4f1833"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_gateway_kernel::transition::valid_transition"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -15600,7 +16153,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Every pair of transition views: the unchangeable bytes, the plan summary, the closed stage table, the checkpoint, the pre-entry and response fields, and stage consistency."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, and the listed foundational axioms are trusted; the theorem has no premise.",
@@ -15616,17 +16169,17 @@ axioms = [
[[claims.evidence]]
kind = "lean-refinement"
artifact = "formal/Auths/Product/Refinement/Gateway.lean"
-sha256 = "59a2ad78eca15d40e4ca249d653cdbb870131ee2f24446bbff2f5f6144c72e13"
+sha256 = "25a1770ec00c4ddcdaaa6cd6995d4c1e47afad17bc7e3164bf1bc4b0babafed6"
[[claims.evidence]]
kind = "mechanical-translation"
artifact = "formal/qualification/aeneas/generated/gateway/Funs.lean"
-sha256 = "c8c86db058f313ab46ee86370d79ee70130c1ead3860dd7dabd49f2a89af1636"
+sha256 = "8c6c2ab1a63280e94f4084566888d727b13a5890b9de4c75aa90322406c5c601"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-234"
@@ -15634,7 +16187,7 @@ claim_text = "The mechanically translated relative basis returns exactly the mod
claim_status = "qualified"
lean_declaration = "Auths.Product.Refinement.Gateway.translated_relative_basis_refines_model"
lean_statement_sha256 = "01d60aceb109109bfa03e8afc89b12399bf3ff7435df7643a695c05f4a3eb0db"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_gateway_kernel::ratio::relative_basis"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -15696,7 +16249,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Every u64 value and optional u64 subtrahend."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, and the listed foundational axioms are trusted; the theorem has no premise.",
@@ -15712,17 +16265,17 @@ axioms = [
[[claims.evidence]]
kind = "lean-refinement"
artifact = "formal/Auths/Product/Refinement/Gateway.lean"
-sha256 = "59a2ad78eca15d40e4ca249d653cdbb870131ee2f24446bbff2f5f6144c72e13"
+sha256 = "25a1770ec00c4ddcdaaa6cd6995d4c1e47afad17bc7e3164bf1bc4b0babafed6"
[[claims.evidence]]
kind = "mechanical-translation"
artifact = "formal/qualification/aeneas/generated/gateway/Funs.lean"
-sha256 = "c8c86db058f313ab46ee86370d79ee70130c1ead3860dd7dabd49f2a89af1636"
+sha256 = "8c6c2ab1a63280e94f4084566888d727b13a5890b9de4c75aa90322406c5c601"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-235"
@@ -15730,7 +16283,7 @@ claim_text = "The mechanically translated relative-ceiling check returns exactly
claim_status = "qualified"
lean_declaration = "Auths.Product.Refinement.Gateway.translated_relative_ceiling_admits_refines_model"
lean_statement_sha256 = "b74b4a9484b06e370376182e5eb9d0160797fbd60246ee5d54bb4d02d935841b"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_gateway_kernel::ratio::relative_ceiling_admits"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -15792,7 +16345,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Every u64 argument and basis and u16 basis points; neither u128 product overflows."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, and the listed foundational axioms are trusted; the theorem has no premise.",
@@ -15808,17 +16361,17 @@ axioms = [
[[claims.evidence]]
kind = "lean-refinement"
artifact = "formal/Auths/Product/Refinement/Gateway.lean"
-sha256 = "59a2ad78eca15d40e4ca249d653cdbb870131ee2f24446bbff2f5f6144c72e13"
+sha256 = "25a1770ec00c4ddcdaaa6cd6995d4c1e47afad17bc7e3164bf1bc4b0babafed6"
[[claims.evidence]]
kind = "mechanical-translation"
artifact = "formal/qualification/aeneas/generated/gateway/Funs.lean"
-sha256 = "c8c86db058f313ab46ee86370d79ee70130c1ead3860dd7dabd49f2a89af1636"
+sha256 = "8c6c2ab1a63280e94f4084566888d727b13a5890b9de4c75aa90322406c5c601"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-249"
@@ -15826,7 +16379,7 @@ claim_text = "Lean proves the connection credential-generation property: retaine
claim_status = "proved"
lean_declaration = "Auths.Product.ConnectionGenerations.retained_generation_exact"
lean_statement_sha256 = "5a5fe438f56954d871f9657a5a3d382ded930c94ba4e4ff14d8786624db91d65"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -15888,7 +16441,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure connection-record generation semantics over every sequence of install, join, rotate, disable, enable, and revoke across any number of processes sharing one record: the credential generation, the retained-entry lease rule, and revocation."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -15908,7 +16461,7 @@ claim_text = "Lean proves the connection credential-generation property: lease g
claim_status = "proved"
lean_declaration = "Auths.Product.ConnectionGenerations.lease_generation_exact"
lean_statement_sha256 = "00f5f2fe7894d82e1bd31d1008f7d6c616bb2fdc44df3d770efd72cb3328b8cc"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -15970,7 +16523,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure connection-record generation semantics over every sequence of install, join, rotate, disable, enable, and revoke across any number of processes sharing one record: the credential generation, the retained-entry lease rule, and revocation."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -15990,7 +16543,7 @@ claim_text = "Lean proves the connection credential-generation property: credent
claim_status = "proved"
lean_declaration = "Auths.Product.ConnectionGenerations.credential_generation_le_generation"
lean_statement_sha256 = "9238e8945c3c2eefb361bdb5948e24e74aff8c8d75bdc993f254fae718423f21"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -16052,7 +16605,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure connection-record generation semantics over every sequence of install, join, rotate, disable, enable, and revoke across any number of processes sharing one record: the credential generation, the retained-entry lease rule, and revocation."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -16072,7 +16625,7 @@ claim_text = "Lean proves the connection credential-generation property: state c
claim_status = "proved"
lean_declaration = "Auths.Product.ConnectionGenerations.state_change_preserves_credential_generation"
lean_statement_sha256 = "21d808e879ca61e1b963ee47434d437f62a4fddfd931e9bddb526fd64fc60d2d"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -16134,7 +16687,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure connection-record generation semantics over every sequence of install, join, rotate, disable, enable, and revoke across any number of processes sharing one record: the credential generation, the retained-entry lease rule, and revocation."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -16154,7 +16707,7 @@ claim_text = "Lean proves the connection credential-generation property: lease s
claim_status = "proved"
lean_declaration = "Auths.Product.ConnectionGenerations.lease_selects_credential_generation"
lean_statement_sha256 = "fcd8e4b1f6cc2d26b557d112ddd23e106297ded39fd605786a19c571a77a566f"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -16216,7 +16769,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure connection-record generation semantics over every sequence of install, join, rotate, disable, enable, and revoke across any number of processes sharing one record: the credential generation, the retained-entry lease rule, and revocation."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -16236,7 +16789,7 @@ claim_text = "Lean proves the connection credential-generation property: join af
claim_status = "proved"
lean_declaration = "Auths.Product.ConnectionGenerations.join_after_state_changes"
lean_statement_sha256 = "cb7c333deb7e0a658e774caa74b3ea3f815fe51dcaf51a2df9577a0d1cb5a8a7"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -16298,7 +16851,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure connection-record generation semantics over every sequence of install, join, rotate, disable, enable, and revoke across any number of processes sharing one record: the credential generation, the retained-entry lease rule, and revocation."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -16318,7 +16871,7 @@ claim_text = "Lean proves the connection credential-generation property: revoked
claim_status = "proved"
lean_declaration = "Auths.Product.ConnectionGenerations.revoked_never_leases"
lean_statement_sha256 = "cd6bed579be40bdfdc14ebdefd4b7f46fbe2499409a8d791466649e46331f49d"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -16380,7 +16933,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure connection-record generation semantics over every sequence of install, join, rotate, disable, enable, and revoke across any number of processes sharing one record: the credential generation, the retained-entry lease rule, and revocation."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -16397,7 +16950,7 @@ claim_text = "The mechanically translated next generation returns exactly the mo
claim_status = "qualified"
lean_declaration = "Auths.Product.Refinement.Connections.translated_next_generation_refines_model"
lean_statement_sha256 = "50f902a37ddd5aa8eb769d37c32b135d114611b85d83dffd4241faa4527575fc"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_connections::kernel::next_generation"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -16459,7 +17012,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Every u64 generation."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, and the listed foundational axioms are trusted; the theorem has no premise.",
@@ -16485,7 +17038,7 @@ sha256 = "dfa1998e3506f06dce142222dc383116d0952155517edc399f5d94e688e22912"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-257"
@@ -16493,7 +17046,7 @@ claim_text = "The mechanically translated state change returns exactly the model
claim_status = "qualified"
lean_declaration = "Auths.Product.Refinement.Connections.translated_state_change_refines_model"
lean_statement_sha256 = "ec42028ad8f3b27ddda3a7529c79097a496ad20d6bfa4b529ebbaa27c97b1b5c"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = [
"auths_connections::kernel::state_change",
"auths_connections::kernel::next_generation",
@@ -16558,7 +17111,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Every pair of u64 generations."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, and the listed foundational axioms are trusted; the theorem has no premise.",
@@ -16584,7 +17137,7 @@ sha256 = "dfa1998e3506f06dce142222dc383116d0952155517edc399f5d94e688e22912"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-258"
@@ -16592,7 +17145,7 @@ claim_text = "The mechanically translated rotation returns exactly the model's g
claim_status = "qualified"
lean_declaration = "Auths.Product.Refinement.Connections.translated_rotation_refines_model"
lean_statement_sha256 = "ae9b8de24e57ae7e8116c3d3a5fe6e35be3cbe350807822b1a42bb0f2cb5c250"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = [
"auths_connections::kernel::rotation",
"auths_connections::kernel::next_generation",
@@ -16657,7 +17210,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Every pair of u64 generations."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, and the listed foundational axioms are trusted; the theorem has no premise.",
@@ -16683,7 +17236,7 @@ sha256 = "dfa1998e3506f06dce142222dc383116d0952155517edc399f5d94e688e22912"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-259"
@@ -16691,7 +17244,7 @@ claim_text = "The mechanically translated retained-entry search returns exactly
claim_status = "qualified"
lean_declaration = "Auths.Product.Refinement.Connections.translated_retained_generation_refines_model"
lean_statement_sha256 = "0f34a598bce76fd894dbb859b226616c03433fa08660f8179f7a08b77d815fd3"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_connections::kernel::retained_generation"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -16753,7 +17306,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Every u64 record generation and every slice of stored generations, in any order."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, and the listed foundational axioms are trusted; the theorem has no premise.",
@@ -16779,7 +17332,7 @@ sha256 = "dfa1998e3506f06dce142222dc383116d0952155517edc399f5d94e688e22912"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-260"
@@ -16787,7 +17340,7 @@ claim_text = "The mechanically translated lease leaf returns exactly the model's
claim_status = "qualified"
lean_declaration = "Auths.Product.Refinement.Connections.translated_lease_generation_refines_model"
lean_statement_sha256 = "0b055274a8b03d3b97de987af664c1ea51b66bfc09e0edc61606483c4c826911"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = [
"auths_connections::kernel::lease_generation",
"auths_connections::kernel::retained_generation",
@@ -16852,7 +17405,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Every u64 record and credential generation and every slice of stored generations."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, and the listed foundational axioms are trusted; the theorem has no premise.",
@@ -16878,7 +17431,7 @@ sha256 = "dfa1998e3506f06dce142222dc383116d0952155517edc399f5d94e688e22912"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-070"
@@ -16886,7 +17439,7 @@ claim_text = "Lean proves the reservation/execution lifecycle property: additive
claim_status = "proved"
lean_declaration = "Auths.Lifecycle.additive_capacity_success_positive"
lean_statement_sha256 = "dcd8c19f436246f345583eb79fa8f56ca0d79801619ec832bb1965cfa4ef7505"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -16948,7 +17501,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure V1 lifecycle transitions, capacity conservation, replay, configuration gates, credential ordering, provider entry, and reconciliation."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -16969,7 +17522,7 @@ claim_text = "Lean proves the reservation/execution lifecycle property: additive
claim_status = "proved"
lean_declaration = "Auths.Lifecycle.additive_capacity_success_conserves"
lean_statement_sha256 = "4c9f497e6dde81872f4262a4d1ef41176e7fed5cbbfdf5ddef76cd6d2f8720a3"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -17031,7 +17584,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure V1 lifecycle transitions, capacity conservation, replay, configuration gates, credential ordering, provider entry, and reconciliation."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -17051,7 +17604,7 @@ claim_text = "Lean proves the reservation/execution lifecycle property: additive
claim_status = "proved"
lean_declaration = "Auths.Lifecycle.additive_capacity_success_never_overflows_u64"
lean_statement_sha256 = "1505323a6e0d5ea769a5f3ce908b2e63df64a61ee675bf46ad235e4934158e55"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -17113,7 +17666,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure V1 lifecycle transitions, capacity conservation, replay, configuration gates, credential ordering, provider entry, and reconciliation."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -17133,7 +17686,7 @@ claim_text = "Lean proves the reservation/execution lifecycle property: reserve
claim_status = "proved"
lean_declaration = "Auths.Lifecycle.reserve_preserves_capacity"
lean_statement_sha256 = "0215021eae68f54b9d07f77d465d2d5d0f574d82c956050d6a42d65f76360f1e"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -17195,7 +17748,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure V1 lifecycle transitions, capacity conservation, replay, configuration gates, credential ordering, provider entry, and reconciliation."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -17215,7 +17768,7 @@ claim_text = "Lean proves the reservation/execution lifecycle property: commit p
claim_status = "proved"
lean_declaration = "Auths.Lifecycle.commit_preserves_capacity"
lean_statement_sha256 = "6e340919635140c32e62bbb5eaae23126fd0f53586c46407c7e16b970f4fd4ca"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -17277,7 +17830,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure V1 lifecycle transitions, capacity conservation, replay, configuration gates, credential ordering, provider entry, and reconciliation."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -17298,7 +17851,7 @@ claim_text = "Lean proves the reservation/execution lifecycle property: release
claim_status = "proved"
lean_declaration = "Auths.Lifecycle.release_preserves_capacity"
lean_statement_sha256 = "a554833f3a765873264be9c667d0293f8b5a7cf6f3a79d9b1fa03131b6414b94"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -17360,7 +17913,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure V1 lifecycle transitions, capacity conservation, replay, configuration gates, credential ordering, provider entry, and reconciliation."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -17381,7 +17934,7 @@ claim_text = "Lean proves the reservation/execution lifecycle property: exact re
claim_status = "proved"
lean_declaration = "Auths.Lifecycle.exact_replay_is_stable"
lean_statement_sha256 = "ddbb29e971f6f0c416d4ae2bbac689cc221c305745d7a2d90b0b36224f63320c"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -17443,7 +17996,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure V1 lifecycle transitions, capacity conservation, replay, configuration gates, credential ordering, provider entry, and reconciliation."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -17460,7 +18013,7 @@ claim_text = "Lean proves the reservation/execution lifecycle property: conflict
claim_status = "proved"
lean_declaration = "Auths.Lifecycle.conflicting_replay_is_not_exact"
lean_statement_sha256 = "16040fad2ac968a316520f6e522063d1e6aeeb5000305573e650296cec80290f"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -17522,7 +18075,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure V1 lifecycle transitions, capacity conservation, replay, configuration gates, credential ordering, provider entry, and reconciliation."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -17539,7 +18092,7 @@ claim_text = "Lean proves the reservation/execution lifecycle property: absent r
claim_status = "proved"
lean_declaration = "Auths.Lifecycle.absent_replay_never_claims_effect"
lean_statement_sha256 = "52a298bd35ffa236d8370ab74d78a674e9e9243a686c52e611f92be3c5cfcf64"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -17601,7 +18154,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure V1 lifecycle transitions, capacity conservation, replay, configuration gates, credential ordering, provider entry, and reconciliation."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -17618,7 +18171,7 @@ claim_text = "Lean proves the reservation/execution lifecycle property: start at
claim_status = "proved"
lean_declaration = "Auths.Lifecycle.start_attempt_requires_credential"
lean_statement_sha256 = "4dcfb31b2e2662aef2c336d0c4f55f922b3c6ac226776bed4daacca4424d1904"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -17680,7 +18233,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure V1 lifecycle transitions, capacity conservation, replay, configuration gates, credential ordering, provider entry, and reconciliation."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -17697,7 +18250,7 @@ claim_text = "Lean proves the reservation/execution lifecycle property: provider
claim_status = "proved"
lean_declaration = "Auths.Lifecycle.provider_call_requires_attempt"
lean_statement_sha256 = "215e750dce32f2f49fa22b96b5c44699bb15604760223b9f26ace3becbac3bdb"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -17759,7 +18312,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure V1 lifecycle transitions, capacity conservation, replay, configuration gates, credential ordering, provider entry, and reconciliation."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -17776,7 +18329,7 @@ claim_text = "Lean proves the reservation/execution lifecycle property: commit r
claim_status = "proved"
lean_declaration = "Auths.Lifecycle.commit_requires_provider_entry_and_effect"
lean_statement_sha256 = "de9baa696f2dbc23d1dfe965a72bc95f211cda0525f3a212af4197864aab36f2"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -17838,7 +18391,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure V1 lifecycle transitions, capacity conservation, replay, configuration gates, credential ordering, provider entry, and reconciliation."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -17855,7 +18408,7 @@ claim_text = "Lean proves the reservation/execution lifecycle property: outcome
claim_status = "proved"
lean_declaration = "Auths.Lifecycle.outcome_unknown_cannot_release"
lean_statement_sha256 = "3d3102111fb51a188aeffe93d7dc6ad13b25a4db54b871f8d4a4b27f98e88c9b"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -17917,7 +18470,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure V1 lifecycle transitions, capacity conservation, replay, configuration gates, credential ordering, provider entry, and reconciliation."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -17934,7 +18487,7 @@ claim_text = "Lean proves the reservation/execution lifecycle property: outcome
claim_status = "proved"
lean_declaration = "Auths.Lifecycle.outcome_unknown_only_reconciliation_can_terminate"
lean_statement_sha256 = "98af24c5f5fa0f00c4cbc0ef76b168f8fbc0811d3a3069e587d844a9406c5def"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -17996,7 +18549,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure V1 lifecycle transitions, capacity conservation, replay, configuration gates, credential ordering, provider entry, and reconciliation."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -18013,7 +18566,7 @@ claim_text = "Lean proves the reservation/execution lifecycle property: configur
claim_status = "proved"
lean_declaration = "Auths.Lifecycle.configuration_mismatch_stops_reservation"
lean_statement_sha256 = "9eb20bb520a94f7bfd23c41793df96630035b109d3bd77ae5520e384d80f87a7"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -18075,7 +18628,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure V1 lifecycle transitions, capacity conservation, replay, configuration gates, credential ordering, provider entry, and reconciliation."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -18092,7 +18645,7 @@ claim_text = "Lean proves the reservation/execution lifecycle property: configur
claim_status = "proved"
lean_declaration = "Auths.Lifecycle.configuration_mismatch_stops_credential"
lean_statement_sha256 = "6f4e8f2d72115f41f5d9881ece6c3d8bff1fcca74bf34415a14475b316085abb"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -18154,7 +18707,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure V1 lifecycle transitions, capacity conservation, replay, configuration gates, credential ordering, provider entry, and reconciliation."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -18171,7 +18724,7 @@ claim_text = "Lean proves the reservation/execution lifecycle property: terminal
claim_status = "proved"
lean_declaration = "Auths.Lifecycle.terminal_states_never_transition"
lean_statement_sha256 = "ac59e2c5beb8dcaf74cafc08c46f48d6d2345ab773dcf2dae55a337bd1c30acc"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -18233,7 +18786,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Pure V1 lifecycle transitions, capacity conservation, replay, configuration gates, credential ordering, provider entry, and reconciliation."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -18250,7 +18803,7 @@ claim_text = "Lean proves the mechanically translated shipping Rust lifecycle ke
claim_status = "proved"
lean_declaration = "Auths.Lifecycle.Refinement.translated_terminal_refines_rich"
lean_statement_sha256 = "8963dd672ee5d30584b6a8a131ca2e28a824c51ce5126805eea83ebbe55d71c9"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_lifecycle::model::LifecycleState::is_terminal"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -18312,7 +18865,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The pinned Charon/Aeneas translation of `auths_lifecycle::model::LifecycleState::is_terminal` is extensionally equivalent to the corresponding rich Lean V1 lifecycle semantics."
residual_assumptions = ["Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the qualified translation boundary, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -18329,7 +18882,7 @@ claim_text = "Lean proves the mechanically translated shipping Rust lifecycle ke
claim_status = "proved"
lean_declaration = "Auths.Lifecycle.Refinement.translated_transition_refines_rich"
lean_statement_sha256 = "ed38f14b22ae254a97a391b4a7ea4064665d323a26e4ec556769f1451731c6cf"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_lifecycle::kernel::transition_code"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -18391,7 +18944,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The pinned Charon/Aeneas translation of `auths_lifecycle::kernel::transition_code` is extensionally equivalent to the corresponding rich Lean V1 lifecycle semantics."
residual_assumptions = ["Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the qualified translation boundary, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -18411,7 +18964,7 @@ claim_text = "Lean proves the mechanically translated shipping Rust lifecycle ke
claim_status = "proved"
lean_declaration = "Auths.Lifecycle.Refinement.translated_exclusive_capacity_refines_rich"
lean_statement_sha256 = "76c468b3a6a2c9a0231675e987fabf48df92073dc24aadce05f6552165c9e8cf"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_lifecycle::kernel::exclusive_capacity_available"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -18473,7 +19026,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The pinned Charon/Aeneas translation of `auths_lifecycle::kernel::exclusive_capacity_available` is extensionally equivalent to the corresponding rich Lean V1 lifecycle semantics."
residual_assumptions = ["Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the qualified translation boundary, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -18490,7 +19043,7 @@ claim_text = "Lean proves the mechanically translated shipping Rust lifecycle ke
claim_status = "proved"
lean_declaration = "Auths.Lifecycle.Refinement.translated_additive_capacity_refines_rich"
lean_statement_sha256 = "530d89045350d0aa6be9ad09121e16e2a91d8ed3652b8e9508e5ef2a655e9ada"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_lifecycle::kernel::additive_capacity_available"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -18552,7 +19105,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The pinned Charon/Aeneas translation of `auths_lifecycle::kernel::additive_capacity_available` is extensionally equivalent to the corresponding rich Lean V1 lifecycle semantics."
residual_assumptions = ["Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the qualified translation boundary, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -18573,7 +19126,7 @@ claim_text = "Lean proves the mechanically translated shipping Rust lifecycle ke
claim_status = "proved"
lean_declaration = "Auths.Lifecycle.Refinement.translated_replay_refines_rich"
lean_statement_sha256 = "ede2a75a3447bb934a2454580f45e9610dcf521b669bca3fbe697f567764d7f4"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_lifecycle::kernel::replay_code"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -18635,7 +19188,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The pinned Charon/Aeneas translation of `auths_lifecycle::kernel::replay_code` is extensionally equivalent to the corresponding rich Lean V1 lifecycle semantics."
residual_assumptions = ["Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the qualified translation boundary, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -18652,7 +19205,7 @@ claim_text = "Binary all composition is commutative."
claim_status = "proved"
lean_declaration = "Auths.all_commutative"
lean_statement_sha256 = "fd3226a3df5fb4730eeaeb50b3190d9f0ae29aeb961d395825787945fdcdd445"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -18714,7 +19267,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The closed three-valued binary all operator."
residual_assumptions = ["Lean's kernel, the pinned toolchain, and the listed foundational axioms are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -18731,7 +19284,7 @@ claim_text = "Binary all composition is associative."
claim_status = "proved"
lean_declaration = "Auths.all_associative"
lean_statement_sha256 = "5749607a3f2752bdd73ae4559af264763d8a36ec3d1258ed4f60b11467d7e12d"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -18793,7 +19346,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The closed three-valued binary all operator."
residual_assumptions = ["Lean's kernel, the pinned toolchain, and the listed foundational axioms are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -18810,7 +19363,7 @@ claim_text = "Binary all composition is idempotent."
claim_status = "proved"
lean_declaration = "Auths.all_idempotent"
lean_statement_sha256 = "99993635cb0eadeacab1d64ee7f674245481d4160a770deaed82022c0851a3f6"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -18872,7 +19425,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The closed three-valued binary all operator."
residual_assumptions = ["Lean's kernel, the pinned toolchain, and the listed foundational axioms are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -18889,7 +19442,7 @@ claim_text = "Binary any composition is commutative."
claim_status = "proved"
lean_declaration = "Auths.any_commutative"
lean_statement_sha256 = "a2680b963a19fb4f377fefaad4102a4469895954bbea2b7541894902e979115d"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -18951,7 +19504,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The closed three-valued binary any operator."
residual_assumptions = ["Lean's kernel, the pinned toolchain, and the listed foundational axioms are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -18968,7 +19521,7 @@ claim_text = "Binary any composition is associative."
claim_status = "proved"
lean_declaration = "Auths.any_associative"
lean_statement_sha256 = "3ceaf22f6776423ee96cbdf5fa9bc5c13fca94bdbb3355b016d6b8e4c275afa8"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -19030,7 +19583,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The closed three-valued binary any operator."
residual_assumptions = ["Lean's kernel, the pinned toolchain, and the listed foundational axioms are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -19047,7 +19600,7 @@ claim_text = "Binary any composition is idempotent."
claim_status = "proved"
lean_declaration = "Auths.any_idempotent"
lean_statement_sha256 = "8f1cb9d3d130a67d026a96f278678b9070badc44f40f7c538ee68a3f9af90d82"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -19109,7 +19662,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The closed three-valued binary any operator."
residual_assumptions = ["Lean's kernel, the pinned toolchain, and the listed foundational axioms are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -19126,7 +19679,7 @@ claim_text = "A threshold of one over two truth values equals binary any."
claim_status = "proved"
lean_declaration = "Auths.threshold_one_eq_any"
lean_statement_sha256 = "639a98d2dcab0db13ec9b221b75605fa48ec169a263f76feb2034f912d7a69c7"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_algebra_kernel::threshold_counts"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -19188,7 +19741,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The hand-written two-input threshold helper, linked to the generated threshold-count function by thresholdTwo_eq_thresholdCounts."
residual_assumptions = ["Lean's kernel, the pinned toolchain, generator correctness, and the Rust vector harness are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -19210,7 +19763,7 @@ claim_text = "A threshold of two over two truth values equals binary all."
claim_status = "proved"
lean_declaration = "Auths.threshold_n_eq_all"
lean_statement_sha256 = "7bfb0c0d44d1f31949d7039be3c2052cab9112fcb12d2abef005e6c52979b815"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_algebra_kernel::threshold_counts"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -19272,7 +19825,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The hand-written two-input threshold helper, linked to the generated threshold-count function by thresholdTwo_eq_thresholdCounts."
residual_assumptions = ["Lean's kernel, the pinned toolchain, generator correctness, and the Rust vector harness are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -19294,7 +19847,7 @@ claim_text = "Raising the two-input threshold from one to two cannot raise its a
claim_status = "proved"
lean_declaration = "Auths.threshold_monotone_k"
lean_statement_sha256 = "34c77c69b14f1b08db8fba8da86a0a9eb65b95147b85d8d3f48811b47cabcb79"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_algebra_kernel::threshold_counts"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -19356,7 +19909,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Exactly the two-input threshold helper and the declared three-valued truth order."
residual_assumptions = ["Lean's kernel, the pinned toolchain, listed foundational axioms, generator correctness, and the Rust vector harness are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -19381,7 +19934,7 @@ claim_text = "A threshold of zero over two truth values is authorized, the thres
claim_status = "proved"
lean_declaration = "Auths.threshold_zero_authorized"
lean_statement_sha256 = "bc42184fb1f441c5e5ee699d44e0a80c38161401ae018e678d5b7a7300463cbf"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_algebra_kernel::threshold_counts"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -19443,7 +19996,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The hand-written two-input threshold helper, the generated threshold-count function, and the declared three-valued truth order."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -19460,7 +20013,7 @@ claim_text = "A threshold of three or more over two truth values is denied, the
claim_status = "proved"
lean_declaration = "Auths.threshold_above_inputs_denied"
lean_statement_sha256 = "d080b789affedcde693adeadcd36a21e772e0c5d21f40002ccc521f48a1c3748"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_algebra_kernel::threshold_counts"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -19522,7 +20075,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The hand-written two-input threshold helper, the generated threshold-count function, and the declared three-valued truth order."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -19539,7 +20092,7 @@ claim_text = "For every k, raising the two-input threshold cannot raise its auth
claim_status = "proved"
lean_declaration = "Auths.threshold_antitone_k"
lean_statement_sha256 = "efa1ab70eaceb5322dd690f37fd6372ab1092f49da878f391158840b4afed292"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_algebra_kernel::threshold_counts"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -19601,7 +20154,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The hand-written two-input threshold helper, the generated threshold-count function, and the declared three-valued truth order."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -19622,7 +20175,7 @@ claim_text = "For every count of authorized and indeterminate members, raising t
claim_status = "proved"
lean_declaration = "Auths.thresholdCounts_antitone_required"
lean_statement_sha256 = "5f1fda6eb247d5ed916296eccfef581d05ed4ecbed69455445bc8ce5ff12fcd6"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_algebra_kernel::threshold_counts"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -19684,7 +20237,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The generated threshold-count function, whose Rust twin the verifier runs, and the declared three-valued truth order."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -19704,7 +20257,7 @@ claim_text = "For every k, the two-input threshold helper equals the generated t
claim_status = "proved"
lean_declaration = "Auths.thresholdTwo_eq_thresholdCounts"
lean_statement_sha256 = "e36ef8702dd4af7d133609adacf0b481ac3302e59f9b0d64563976531f8a017e"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_algebra_kernel::threshold_counts"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -19766,7 +20319,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The hand-written two-input threshold helper, the generated threshold-count function, and the declared three-valued truth order."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -19786,7 +20339,7 @@ claim_text = "Swapping the two inputs preserves binary all and binary any result
claim_status = "proved"
lean_declaration = "Auths.binary_composition_swap_invariant"
lean_statement_sha256 = "5a2f18a8b889fca7e2d4f239a97a12490193b19c6d3e8703164e04cc556986c7"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -19848,7 +20401,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "A two-value swap only; arbitrary authorization-plan permutations are not claimed."
residual_assumptions = ["Lean's kernel, the pinned toolchain, and the listed foundational axioms are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -19865,7 +20418,7 @@ claim_text = "The two-value minimum diagnostic code is invariant under swapping
claim_status = "proved"
lean_declaration = "Auths.canonical_diagnostic_permutation_invariant"
lean_statement_sha256 = "c0eae1c5c0004d836da0c28c5d5395ecb7e534d05e9816a573d27b7c0f912806"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -19927,7 +20480,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The two-value canonicalCode helper; arbitrary plan diagnostic permutations are not claimed."
residual_assumptions = ["Lean's kernel, the pinned toolchain, and the listed foundational axioms are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -19944,7 +20497,7 @@ claim_text = "The current plan visit helper is definitionally the plan leaf enum
claim_status = "proved"
lean_declaration = "Auths.plan_visit_is_leaf_enumeration"
lean_statement_sha256 = "a9d2d3fd6140d7535c9cf3567cf92a5891e8e77fda5924a34a032bb1d9cab0cb"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -20006,7 +20559,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "A definitional model equality; it does not prove that the shipping evaluator visits leaves once."
residual_assumptions = ["Lean's kernel, the pinned toolchain, and the listed foundational axioms are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -20026,7 +20579,7 @@ claim_text = "The current plan cost helper is definitionally the recursive node
claim_status = "proved"
lean_declaration = "Auths.plan_cost_is_node_count"
lean_statement_sha256 = "d0e9f71500f5da6108662057f11150fa61188b6b5059562b6db0079bd4e5820e"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -20088,7 +20641,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "A definitional model equality; no asymptotic or shipping evaluator cost bound is claimed."
residual_assumptions = ["Lean's kernel, the pinned toolchain, and the listed foundational axioms are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -20108,7 +20661,7 @@ claim_text = "An authorized generated threshold result implies the authorized co
claim_status = "proved"
lean_declaration = "Auths.authorized_implies_threshold_met"
lean_statement_sha256 = "367522026ebf77cc4e8ee694920ff72384ab58cb573bb7cefb8406117955d2d1"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_algebra_kernel::threshold_counts"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -20170,7 +20723,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The generated count classifier and its exhaustive target-V1 bounded Rust vectors."
residual_assumptions = ["Lean's kernel, the pinned toolchain, generator correctness, and the Rust vector harness are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -20192,7 +20745,7 @@ claim_text = "A denied generated threshold result implies authorized plus indete
claim_status = "proved"
lean_declaration = "Auths.denied_implies_threshold_impossible"
lean_statement_sha256 = "50005cbb684ef7566e1b6f71fc766e9960de178f0a86f03e37f4eb63cf7e46d5"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_algebra_kernel::threshold_counts"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -20254,7 +20807,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The generated count classifier and its exhaustive target-V1 bounded Rust vectors."
residual_assumptions = ["Lean's kernel, the pinned toolchain, listed foundational axioms, generator correctness, and the Rust vector harness are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -20279,7 +20832,7 @@ claim_text = "An indeterminate generated threshold result means authorization is
claim_status = "proved"
lean_declaration = "Auths.indeterminate_implies_threshold_reachable"
lean_statement_sha256 = "dd970bbab7dd61f7b1274f9e829358d9627b3c822d4119628191018722f31039"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_algebra_kernel::threshold_counts"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -20341,7 +20894,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The generated count classifier and its exhaustive target-V1 bounded Rust vectors."
residual_assumptions = ["Lean's kernel, the pinned toolchain, listed foundational axioms, generator correctness, and the Rust vector harness are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -20367,7 +20920,7 @@ claim_text = "Lean proves adding an observation requirement never enlarges the s
claim_status = "proved"
lean_declaration = "Auths.Observation.requirements_monotone"
lean_statement_sha256 = "148ccf94301235d93a8436967797e60faca90b144df545bd381377b4a50a6a6a"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -20429,7 +20982,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Abstract evidence-conditioned authority: closed conjunctive observation conditions, freshness, observer and subject binding, over already verified observations and an abstract subject namespace matcher."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -20449,7 +21002,7 @@ claim_text = "Lean proves adding a condition to an observation requirement never
claim_status = "proved"
lean_declaration = "Auths.Observation.conditions_monotone"
lean_statement_sha256 = "580cb89f54be773f8f5b9477dcfa105314d26628e415c3d9e43bc0abbaf13f07"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -20511,7 +21064,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Abstract evidence-conditioned authority: closed conjunctive observation conditions, freshness, observer and subject binding, over already verified observations and an abstract subject namespace matcher."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -20531,7 +21084,7 @@ claim_text = "Lean proves a child grant whose base authority attenuates and whos
claim_status = "proved"
lean_declaration = "Auths.Observation.child_requirements_superset_attenuates"
lean_statement_sha256 = "569899c6f3632631a784c25e616bcf272690422a491a5f6227f51270e442fbe5"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -20593,7 +21146,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Abstract evidence-conditioned authority: closed conjunctive observation conditions, freshness, observer and subject binding, over already verified observations and an abstract subject namespace matcher."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -20613,7 +21166,7 @@ claim_text = "Lean proves attaching more observations never turns a satisfied ob
claim_status = "proved"
lean_declaration = "Auths.Observation.satisfied_monotone_observations"
lean_statement_sha256 = "ba0d15b96cedec283e46c0829abd94525ab9d378848e5f17d737b529e786aff5"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -20675,7 +21228,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Abstract evidence-conditioned authority: closed conjunctive observation conditions, freshness, observer and subject binding, over already verified observations and an abstract subject namespace matcher."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -20695,7 +21248,7 @@ claim_text = "Lean proves an empty observation set satisfies no observation requ
claim_status = "proved"
lean_declaration = "Auths.Observation.no_observation_satisfies_nothing"
lean_statement_sha256 = "2de8b8b35b722baed67f9208a44ed87aad1918842eac9d30b1a0af9248cf6b91"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -20757,7 +21310,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Abstract evidence-conditioned authority: closed conjunctive observation conditions, freshness, observer and subject binding, over already verified observations and an abstract subject namespace matcher."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -20774,7 +21327,7 @@ claim_text = "Lean proves an observation dated after the evaluation time is neve
claim_status = "proved"
lean_declaration = "Auths.Observation.future_observation_never_fresh"
lean_statement_sha256 = "25e1df429190fb5954c738d846264281c1f699bf5f3b2dd09e153458f8f7d3da"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -20836,7 +21389,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Abstract evidence-conditioned authority: closed conjunctive observation conditions, freshness, observer and subject binding, over already verified observations and an abstract subject namespace matcher."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -20857,7 +21410,7 @@ claim_text = "Lean proves an observation older than the requirement's maximum ag
claim_status = "proved"
lean_declaration = "Auths.Observation.stale_observation_never_fresh"
lean_statement_sha256 = "bc723fae1eda0ccab4ab50fe9439378198157d48db4c7cb20c50b99501d4a705"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -20919,7 +21472,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Abstract evidence-conditioned authority: closed conjunctive observation conditions, freshness, observer and subject binding, over already verified observations and an abstract subject namespace matcher."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -20940,7 +21493,7 @@ claim_text = "Lean proves a satisfied observation requirement had every action f
claim_status = "proved"
lean_declaration = "Auths.Observation.satisfied_requires_action_facts"
lean_statement_sha256 = "868823975fffe7737c3f778322256228fee15e6e7aaa6d4ec7f6141da0b11ea1"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -21002,7 +21555,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Abstract evidence-conditioned authority: closed conjunctive observation conditions, freshness, observer and subject binding, over already verified observations and an abstract subject namespace matcher."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -21022,7 +21575,7 @@ claim_text = "Lean proves the three-valued observation verdict is satisfied exac
claim_status = "proved"
lean_declaration = "Auths.Observation.decision_satisfied_iff"
lean_statement_sha256 = "e7e42f0b73ac6aaacf07a646de0481da0942a790607af2896777f242e450da61"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -21084,7 +21637,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Abstract evidence-conditioned authority: closed conjunctive observation conditions, freshness, observer and subject binding, over already verified observations and an abstract subject namespace matcher."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -21104,7 +21657,7 @@ claim_text = "Lean proves a denied observation verdict has an eligible observati
claim_status = "proved"
lean_declaration = "Auths.Observation.decision_denied_has_falsifying_observation"
lean_statement_sha256 = "220d9a14270b12a16d4c386a08706c81032a17990421d6712ca365f94bcacde6"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -21166,7 +21719,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Abstract evidence-conditioned authority: closed conjunctive observation conditions, freshness, observer and subject binding, over already verified observations and an abstract subject namespace matcher."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -21186,7 +21739,7 @@ claim_text = "Lean proves an observation fresh under a smaller maximum age is fr
claim_status = "proved"
lean_declaration = "Auths.Observation.fresh_monotone_max_age"
lean_statement_sha256 = "54093bcffc875104caccb21819e59c367362132d9d2446952efd7a408f65f059"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -21248,7 +21801,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Abstract evidence-conditioned authority: closed conjunctive observation conditions, freshness, observer and subject binding, over already verified observations and an abstract subject namespace matcher."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -21269,7 +21822,7 @@ claim_text = "Lean proves a strictly narrowed observation requirement is satisfi
claim_status = "proved"
lean_declaration = "Auths.Observation.narrows_monotone"
lean_statement_sha256 = "9fba78f0577d3888e9f544a917fe1a44fa53cfbaeb9b600a27cab1a56859cb05"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -21331,7 +21884,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Abstract evidence-conditioned authority: closed conjunctive observation conditions, freshness, observer and subject binding, over already verified observations and an abstract subject namespace matcher."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -21352,7 +21905,7 @@ claim_text = "Lean proves a covering observation requirement is satisfied only w
claim_status = "proved"
lean_declaration = "Auths.Observation.covers_monotone"
lean_statement_sha256 = "aafa19c473cfc60aa626d058fcb1285f5b515a435dbbf9d2d250c9935fd65bbb"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -21414,7 +21967,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Abstract evidence-conditioned authority: closed conjunctive observation conditions, freshness, observer and subject binding, over already verified observations and an abstract subject namespace matcher."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -21435,7 +21988,7 @@ claim_text = "Lean proves observation-requirement covering is transitive."
claim_status = "proved"
lean_declaration = "Auths.Observation.covers_trans"
lean_statement_sha256 = "8be67281de157aa1ce9f37e536382462d33cebfc8b6ad06ebacc950b100cb241"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -21497,7 +22050,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Abstract evidence-conditioned authority: closed conjunctive observation conditions, freshness, observer and subject binding, over already verified observations and an abstract subject namespace matcher."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -21517,7 +22070,7 @@ claim_text = "Lean proves observation-requirement list attenuation is reflexive.
claim_status = "proved"
lean_declaration = "Auths.Observation.requirements_attenuate_refl"
lean_statement_sha256 = "3ddf75aca2a28b5c1f927d5adc747a45ab309a933646b17f963484e5d5128c98"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -21579,7 +22132,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Abstract evidence-conditioned authority: closed conjunctive observation conditions, freshness, observer and subject binding, over already verified observations and an abstract subject namespace matcher."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -21596,7 +22149,7 @@ claim_text = "Lean proves observation-requirement list attenuation is transitive
claim_status = "proved"
lean_declaration = "Auths.Observation.requirements_attenuate_trans"
lean_statement_sha256 = "3d3f4d07e0abf3bb05962c8d2dd052613dcb5be878594ae858290c2d0fa620d3"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -21658,7 +22211,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Abstract evidence-conditioned authority: closed conjunctive observation conditions, freshness, observer and subject binding, over already verified observations and an abstract subject namespace matcher."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -21678,7 +22231,7 @@ claim_text = "Lean proves a child requirement list that attenuates its parent's
claim_status = "proved"
lean_declaration = "Auths.Observation.requirements_attenuate_monotone"
lean_statement_sha256 = "ee97dc30d382ed4737017fb8325d2ed3777d80c69ad989231e81b02b64686073"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -21740,7 +22293,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Abstract evidence-conditioned authority: closed conjunctive observation conditions, freshness, observer and subject binding, over already verified observations and an abstract subject namespace matcher."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -21761,7 +22314,7 @@ claim_text = "Lean proves the observation-requirement-v1 attenuation law is a pr
claim_status = "proved"
lean_declaration = "Auths.Observation.observation_requirement_law_lawful"
lean_statement_sha256 = "617d5169222466363d1b356846ca50e92b48c170f114cd7cfec4bebbbbb64f7d"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -21823,7 +22376,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Critical-extension attenuation laws over decoded payloads: each registered handler law is a preorder that narrows the worlds its payload admits."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -21844,7 +22397,7 @@ claim_text = "Lean proves the observation-requirement-v1 law accepts adding the
claim_status = "proved"
lean_declaration = "Auths.Observation.observation_requirement_law_accepts_addition"
lean_statement_sha256 = "e20a8c08c602235bc1aabfeff4bf5e9eebd473a410cd568b26c88c497b8f4ef0"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -21906,7 +22459,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Critical-extension attenuation laws over decoded payloads: each registered handler law is a preorder that narrows the worlds its payload admits."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -21923,7 +22476,7 @@ claim_text = "Lean proves the K-of-N approval property: approval decide eq thres
claim_status = "proved"
lean_declaration = "Auths.Approval.approval_decide_eq_threshold_counts"
lean_statement_sha256 = "edf74a0cb0c581e211fb2e53b7042562c38c18635c33161d238967b8be2f247e"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -21985,7 +22538,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Abstract K-of-N approval model: per-approval verdicts over the approver, bound action, bound requirement identifier, validity window, and verification truth; distinct counting per listed approver; and the generated thresholdCounts decision."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -22002,7 +22555,7 @@ claim_text = "Lean proves the K-of-N approval property: approval authorized iff
claim_status = "proved"
lean_declaration = "Auths.Approval.approval_authorized_iff_distinct_quorum"
lean_statement_sha256 = "ed8b9c7105938a6c41ee1321aaea4472efc6d75b6a9ee2a6791ed2fa139f316c"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -22064,7 +22617,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Abstract K-of-N approval model: per-approval verdicts over the approver, bound action, bound requirement identifier, validity window, and verification truth; distinct counting per listed approver; and the generated thresholdCounts decision."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -22084,7 +22637,7 @@ claim_text = "Lean proves the K-of-N approval property: approval duplicate irrel
claim_status = "proved"
lean_declaration = "Auths.Approval.approval_duplicate_irrelevant"
lean_statement_sha256 = "091b7dc31c1282d496858bae18b2a24238427aa9e44822bf0b87e902f969f83e"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -22146,7 +22699,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Abstract K-of-N approval model: per-approval verdicts over the approver, bound action, bound requirement identifier, validity window, and verification truth; distinct counting per listed approver; and the generated thresholdCounts decision."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -22166,7 +22719,7 @@ claim_text = "Lean proves the K-of-N approval property: approval outsider irrele
claim_status = "proved"
lean_declaration = "Auths.Approval.approval_outsider_irrelevant"
lean_statement_sha256 = "7919b2b9e3f9af2cc09100de0d401b6481c08132ddc5cce127a4e4bc64ae0e77"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -22228,7 +22781,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Abstract K-of-N approval model: per-approval verdicts over the approver, bound action, bound requirement identifier, validity window, and verification truth; distinct counting per listed approver; and the generated thresholdCounts decision."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -22248,7 +22801,7 @@ claim_text = "Lean proves the K-of-N approval property: approval other action ir
claim_status = "proved"
lean_declaration = "Auths.Approval.approval_other_action_irrelevant"
lean_statement_sha256 = "ab11d19f7f7a07159511b63cf269779db785763da037cfdaac4a6504ac844df4"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -22310,7 +22863,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Abstract K-of-N approval model: per-approval verdicts over the approver, bound action, bound requirement identifier, validity window, and verification truth; distinct counting per listed approver; and the generated thresholdCounts decision."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -22330,7 +22883,7 @@ claim_text = "Lean proves the K-of-N approval property: approval other requireme
claim_status = "proved"
lean_declaration = "Auths.Approval.approval_other_requirement_irrelevant"
lean_statement_sha256 = "6ae502a88055c3daca41ca03e8146337bb020150395d6f05f185e209320c9e78"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -22392,7 +22945,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Abstract K-of-N approval model: per-approval verdicts over the approver, bound action, bound requirement identifier, validity window, and verification truth; distinct counting per listed approver; and the generated thresholdCounts decision."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -22412,7 +22965,7 @@ claim_text = "Lean proves the K-of-N approval property: approval self approval i
claim_status = "proved"
lean_declaration = "Auths.Approval.approval_self_approval_irrelevant"
lean_statement_sha256 = "83b91c2de34c32a06176d103c17b455d7e5c1977f18c3b7d9fb7e758f99a6da4"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -22474,7 +23027,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Abstract K-of-N approval model: per-approval verdicts over the approver, bound action, bound requirement identifier, validity window, and verification truth; distinct counting per listed approver; and the generated thresholdCounts decision."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -22494,7 +23047,7 @@ claim_text = "Lean proves the K-of-N approval property: approval permutation inv
claim_status = "proved"
lean_declaration = "Auths.Approval.approval_permutation_invariant"
lean_statement_sha256 = "e5167218ec0c197a06f1eae05406ec507b75f3b2272a409c3f431f236070be33"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -22556,7 +23109,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Abstract K-of-N approval model: per-approval verdicts over the approver, bound action, bound requirement identifier, validity window, and verification truth; distinct counting per listed approver; and the generated thresholdCounts decision."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -22576,7 +23129,7 @@ claim_text = "Lean proves the K-of-N approval property: approval monotone in app
claim_status = "proved"
lean_declaration = "Auths.Approval.approval_monotone_in_approvals"
lean_statement_sha256 = "eb3583553f5f013a3a8f76bc84808f167a852b199d4528faf1beaa6f9aa4b98e"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -22638,7 +23191,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Abstract K-of-N approval model: per-approval verdicts over the approver, bound action, bound requirement identifier, validity window, and verification truth; distinct counting per listed approver; and the generated thresholdCounts decision."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -22659,7 +23212,7 @@ claim_text = "Lean proves the K-of-N approval property: approval antitone in thr
claim_status = "proved"
lean_declaration = "Auths.Approval.approval_antitone_in_threshold"
lean_statement_sha256 = "b1e0b00684d659884ad65523e4f4b42ceec355686cebdc950ddd6f70c57b67a1"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -22721,7 +23274,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Abstract K-of-N approval model: per-approval verdicts over the approver, bound action, bound requirement identifier, validity window, and verification truth; distinct counting per listed approver; and the generated thresholdCounts decision."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -22741,7 +23294,7 @@ claim_text = "Lean proves the K-of-N approval property: approval antitone in app
claim_status = "proved"
lean_declaration = "Auths.Approval.approval_antitone_in_approvers"
lean_statement_sha256 = "91f04d22b450064899e426f5a02c45536f8e0a550cfd2209f3032c8a9cd0bc9d"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -22803,7 +23356,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Abstract K-of-N approval model: per-approval verdicts over the approver, bound action, bound requirement identifier, validity window, and verification truth; distinct counting per listed approver; and the generated thresholdCounts decision."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -22824,7 +23377,7 @@ claim_text = "Lean proves the K-of-N approval property: approval authorized meet
claim_status = "proved"
lean_declaration = "Auths.Approval.approval_authorized_meets"
lean_statement_sha256 = "7543f43fad4398cf6f5514a0f628581d569ddb1b0190d4d341ed4ad762cb0174"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -22886,7 +23439,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Abstract K-of-N approval model: per-approval verdicts over the approver, bound action, bound requirement identifier, validity window, and verification truth; distinct counting per listed approver; and the generated thresholdCounts decision."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -22907,7 +23460,7 @@ claim_text = "Lean proves the K-of-N approval property: covers monotone."
claim_status = "proved"
lean_declaration = "Auths.Approval.covers_monotone"
lean_statement_sha256 = "af566b39841cd9fb7769ebeb7004d3b64edfbcb35a7d779a72d9c017459a7d35"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -22969,7 +23522,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Abstract approval-requirement-v1 attenuation law over decoded requirement lists (approver sets and thresholds) and the sets of approving principals each admits."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -22990,7 +23543,7 @@ claim_text = "Lean proves the K-of-N approval property: covers trans."
claim_status = "proved"
lean_declaration = "Auths.Approval.covers_trans"
lean_statement_sha256 = "71bb145d77cd3db75a110f32d90269a90ada8f43a25e5569a379a87ec68a2d9f"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -23052,7 +23605,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Abstract approval-requirement-v1 attenuation law over decoded requirement lists (approver sets and thresholds) and the sets of approving principals each admits."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -23069,7 +23622,7 @@ claim_text = "Lean proves the K-of-N approval property: approval requirements at
claim_status = "proved"
lean_declaration = "Auths.Approval.approval_requirements_attenuate_refl"
lean_statement_sha256 = "5001fe3083c023f9cc8260ff9d89ca869fbdf3b85cdf4bfc267ce78e9bae9358"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -23131,7 +23684,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Abstract approval-requirement-v1 attenuation law over decoded requirement lists (approver sets and thresholds) and the sets of approving principals each admits."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -23148,7 +23701,7 @@ claim_text = "Lean proves the K-of-N approval property: approval requirements at
claim_status = "proved"
lean_declaration = "Auths.Approval.approval_requirements_attenuate_trans"
lean_statement_sha256 = "e3cdeabbbb19bff2db617206edf53cbbc5758f6276b7144cb9174aa52dd430ad"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -23210,7 +23763,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Abstract approval-requirement-v1 attenuation law over decoded requirement lists (approver sets and thresholds) and the sets of approving principals each admits."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -23227,7 +23780,7 @@ claim_text = "Lean proves the K-of-N approval property: approval requirements at
claim_status = "proved"
lean_declaration = "Auths.Approval.approval_requirements_attenuate_monotone"
lean_statement_sha256 = "bbfc53b74d226bc67466f477529aa6cdbf67df24408ff3eeae10773d92e5728d"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -23289,7 +23842,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Abstract approval-requirement-v1 attenuation law over decoded requirement lists (approver sets and thresholds) and the sets of approving principals each admits."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -23310,7 +23863,7 @@ claim_text = "Lean proves the K-of-N approval property: approval requirement law
claim_status = "proved"
lean_declaration = "Auths.Approval.approval_requirement_law_lawful"
lean_statement_sha256 = "53044016738fba16175dea47ebacba4ada4e9286616ea79a8cac69f94cad5b03"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -23372,7 +23925,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Abstract approval-requirement-v1 attenuation law over decoded requirement lists (approver sets and thresholds) and the sets of approving principals each admits."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -23393,7 +23946,7 @@ claim_text = "Lean proves the K-of-N approval property: approval requirement law
claim_status = "proved"
lean_declaration = "Auths.Approval.approval_requirement_law_accepts_addition"
lean_statement_sha256 = "f8224b7c2a7f5b10c9fe3d1b368b240888e88c62ace4db0ef100ce24faea3661"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -23455,7 +24008,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Abstract approval-requirement-v1 attenuation law over decoded requirement lists (approver sets and thresholds) and the sets of approving principals each admits."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -23472,7 +24025,7 @@ claim_text = "Lean proves the K-of-N approval property: approval requirement law
claim_status = "proved"
lean_declaration = "Auths.Approval.approval_requirement_law_refuses_drop"
lean_statement_sha256 = "fe7f7ba5f0892d4b6f9077950bbc333541a83d79fafe5d8ceee3dbed044e242e"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -23534,7 +24087,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Abstract approval-requirement-v1 attenuation law over decoded requirement lists (approver sets and thresholds) and the sets of approving principals each admits."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -23551,7 +24104,7 @@ claim_text = "Lean proves the exact-marker-v1 attenuation law is a preorder that
claim_status = "proved"
lean_declaration = "Auths.exact_marker_law_lawful"
lean_statement_sha256 = "6f9ff0b2e1f580f8cfc27dcaeaefeee2a19b7cda02879335bf9763287a89eb05"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -23613,7 +24166,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Critical-extension attenuation laws over decoded payloads: each registered handler law is a preorder that narrows the worlds its payload admits."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -23630,7 +24183,7 @@ claim_text = "Lean proves the exact-marker-v1 law refuses adding the marker."
claim_status = "proved"
lean_declaration = "Auths.exact_marker_law_refuses_addition"
lean_statement_sha256 = "1019a69a34a92a24de8ebdb88a1a5b5ee4707f044a355fbb62a097b913a9c7f5"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -23692,7 +24245,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Critical-extension attenuation laws over decoded payloads: each registered handler law is a preorder that narrows the worlds its payload admits."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -23709,7 +24262,7 @@ claim_text = "Lean proves the narrowing premise holds when each registered ident
claim_status = "proved"
lean_declaration = "Auths.Rich.ExtensionLawsNarrow.of_lawful"
lean_statement_sha256 = "131876b8431f7e185e96f5249b6329b058bd7a3ad1feac47caa1d72c5921c8a4"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = []
semantic_source_closure = [
"formal/Auths.lean",
@@ -23771,7 +24324,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Rich target-V1 authority semantics over opaque identity carriers and extensional finite sets."
residual_assumptions = ["Lean's kernel, the pinned toolchain, the listed foundational axioms, and the theorem premises are trusted."]
toolchain_lock_sha256 = "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
@@ -23788,7 +24341,7 @@ claim_text = "The mechanically translated fact-name comparison returns exactly t
claim_status = "qualified"
lean_declaration = "Auths.Refinement.Observation.translated_fact_name_equal_refines_model"
lean_statement_sha256 = "6b15e1125bd960ce5bf8c3447e0587ceebd672e592a891b26bc2d5e68a9add5c"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_model::observation::fact_name_equal"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -23850,7 +24403,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "UTF-8 byte comparison of two fact names, bridged to String equality by UTF-8 injectivity."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, the listed foundational axioms, and the theorem's explicit representation-validity premises (every compared text fits the u32 UTF-8 byte bound) are trusted.",
@@ -23876,7 +24429,7 @@ sha256 = "d4d36b72718a4076b7c0561abfa2edc006a1f9a4491762979f86c3dd99a71fdc"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-189"
@@ -23884,7 +24437,7 @@ claim_text = "The mechanically translated freshness predicate returns exactly th
claim_status = "qualified"
lean_declaration = "Auths.Refinement.Observation.translated_observation_fresh_refines_model"
lean_statement_sha256 = "3db4557c386de7ea8812abdf9a330bc6a96432ce30daf16df94c060958d2673a"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_model::observation::observation_fresh"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -23946,7 +24499,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The guarded u64 age subtraction, maximum age, and anchor validity window, for every u64 input."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, the listed foundational axioms, and the theorem's explicit representation-validity premises (every compared text fits the u32 UTF-8 byte bound) are trusted.",
@@ -23972,7 +24525,7 @@ sha256 = "d4d36b72718a4076b7c0561abfa2edc006a1f9a4491762979f86c3dd99a71fdc"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-190"
@@ -23980,7 +24533,7 @@ claim_text = "The mechanically translated exact-subject comparison returns exact
claim_status = "qualified"
lean_declaration = "Auths.Refinement.Observation.translated_observation_subject_equal_refines_model"
lean_statement_sha256 = "9b0aec23b54259d75ed186dbc7ef48d9e63bd0552e0a669478e2c95434501da4"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_model::observation::observation_subject_equal"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -24042,7 +24595,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "UTF-8 byte comparison of the expected and observed resource, bridged to String equality."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, the listed foundational axioms, and the theorem's explicit representation-validity premises (every compared text fits the u32 UTF-8 byte bound) are trusted.",
@@ -24068,7 +24621,7 @@ sha256 = "d4d36b72718a4076b7c0561abfa2edc006a1f9a4491762979f86c3dd99a71fdc"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-191"
@@ -24076,7 +24629,7 @@ claim_text = "The mechanically translated fact-value comparison returns exactly
claim_status = "qualified"
lean_declaration = "Auths.Refinement.Observation.translated_fact_value_equal_refines_model"
lean_statement_sha256 = "c2e5d55a276fda45f44996688fa2ce37eba6a02c4af677e93f0438495cee4112"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_model::observation::fact_value_equal"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -24138,7 +24691,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Unsigned, byte-string, and text fact values, including every cross-type pair."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, the listed foundational axioms, and the theorem's explicit representation-validity premises (every compared text fits the u32 UTF-8 byte bound) are trusted.",
@@ -24164,7 +24717,7 @@ sha256 = "d4d36b72718a4076b7c0561abfa2edc006a1f9a4491762979f86c3dd99a71fdc"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-192"
@@ -24172,7 +24725,7 @@ claim_text = "The mechanically translated inclusive range test returns exactly t
claim_status = "qualified"
lean_declaration = "Auths.Refinement.Observation.translated_uint_range_contains_refines_model"
lean_statement_sha256 = "a62d865cc0aed16965c3f48786e28f4eb056e75a8af68ea1b68a97d40dfcd81a"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_model::observation::uint_range_contains"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -24234,7 +24787,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The inclusive u64 range test on an observed unsigned value."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, the listed foundational axioms, and the theorem's explicit representation-validity premises (every compared text fits the u32 UTF-8 byte bound) are trusted.",
@@ -24259,7 +24812,7 @@ sha256 = "d4d36b72718a4076b7c0561abfa2edc006a1f9a4491762979f86c3dd99a71fdc"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-193"
@@ -24267,7 +24820,7 @@ claim_text = "The mechanically translated membership loop returns exactly the mo
claim_status = "qualified"
lean_declaration = "Auths.Refinement.Observation.translated_member_values_contain_refines_model"
lean_statement_sha256 = "1a823612cdb0ab8b1dfa2561f4563a452c907691b7412a258ffee9c2538693e6"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = [
"auths_model::observation::member_values_contain",
"auths_model::observation::fact_value_equal",
@@ -24332,7 +24885,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The terminating scan of a finite literal list for an observed fact value."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, the listed foundational axioms, and the theorem's explicit representation-validity premises (every compared text fits the u32 UTF-8 byte bound) are trusted.",
@@ -24358,7 +24911,7 @@ sha256 = "d4d36b72718a4076b7c0561abfa2edc006a1f9a4491762979f86c3dd99a71fdc"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-194"
@@ -24366,7 +24919,7 @@ claim_text = "The mechanically translated fact lookup returns a value whose abst
claim_status = "qualified"
lean_declaration = "Auths.Refinement.Observation.translated_observation_fact_refines_model"
lean_statement_sha256 = "5f566de2f5bf7018c2f9a683d466d8bcc99d1bc88bd19f0467efb34fe40b1bfa"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = [
"auths_model::observation::observation_fact",
"auths_model::observation::fact_name_equal",
@@ -24431,7 +24984,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The terminating first-match scan of an observation's facts by name."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, the listed foundational axioms, and the theorem's explicit representation-validity premises (every compared text fits the u32 UTF-8 byte bound) are trusted.",
@@ -24457,7 +25010,7 @@ sha256 = "d4d36b72718a4076b7c0561abfa2edc006a1f9a4491762979f86c3dd99a71fdc"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-195"
@@ -24465,7 +25018,7 @@ claim_text = "The mechanically translated condition-atom evaluator returns exact
claim_status = "qualified"
lean_declaration = "Auths.Refinement.Observation.translated_condition_value_holds_refines_model"
lean_statement_sha256 = "87c518e6cd6ddd3ab66a1acf9cb01dd6a8ed729895310f6dd6b0f36c871be3ed"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = [
"auths_model::observation::condition_value_holds",
"auths_model::observation::fact_value_equal",
@@ -24532,7 +25085,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "All four condition atoms against a present or absent observed value and a present or absent action value."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, the listed foundational axioms, and the theorem's explicit representation-validity premises (every compared text fits the u32 UTF-8 byte bound) are trusted.",
@@ -24558,7 +25111,7 @@ sha256 = "d4d36b72718a4076b7c0561abfa2edc006a1f9a4491762979f86c3dd99a71fdc"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-196"
@@ -24566,7 +25119,7 @@ claim_text = "The mechanically translated named-condition evaluator returns exac
claim_status = "qualified"
lean_declaration = "Auths.Refinement.Observation.translated_observation_condition_holds_refines_model"
lean_statement_sha256 = "d5df152ab006c6eccea9e9e1d19a4b2bf4643f4a6629d2f0a89a5b9a3759ea72"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = [
"auths_model::observation::observation_condition_holds",
"auths_model::observation::observation_fact",
@@ -24636,7 +25189,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "One named condition over an observation's facts with its supplied action value."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, the listed foundational axioms, and the theorem's explicit representation-validity premises (every compared text fits the u32 UTF-8 byte bound) are trusted.",
@@ -24662,7 +25215,7 @@ sha256 = "d4d36b72718a4076b7c0561abfa2edc006a1f9a4491762979f86c3dd99a71fdc"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-197"
@@ -24670,7 +25223,7 @@ claim_text = "The mechanically translated condition conjunction returns exactly
claim_status = "qualified"
lean_declaration = "Auths.Refinement.Observation.translated_observation_conditions_hold_refines_model"
lean_statement_sha256 = "1693b5b82975dde2606b046c58afddf01ef58c6c6c4f973a1b0528dc99132810"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = [
"auths_model::observation::observation_conditions_hold",
"auths_model::observation::observation_condition_holds",
@@ -24741,7 +25294,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The terminating conjunction loop over a condition slice and a parallel action-value slice."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, the listed foundational axioms, and the theorem's explicit representation-validity premises (every compared text fits the u32 UTF-8 byte bound) are trusted.",
@@ -24767,7 +25320,7 @@ sha256 = "d4d36b72718a4076b7c0561abfa2edc006a1f9a4491762979f86c3dd99a71fdc"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-198"
@@ -24775,7 +25328,7 @@ claim_text = "The mechanically translated condition conjunction rejects a condit
claim_status = "qualified"
lean_declaration = "Auths.Refinement.Observation.translated_observation_conditions_length_mismatch_fails_closed"
lean_statement_sha256 = "4ff3b4c45f16ea6a3db0d4e48e54202e20f11ce220a3fee37d426a5f55616eba"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_model::observation::observation_conditions_hold"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -24837,7 +25390,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The length guard that precedes every condition evaluation."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, the listed foundational axioms, and the theorem's explicit representation-validity premises (every compared text fits the u32 UTF-8 byte bound) are trusted.",
@@ -24863,7 +25416,7 @@ sha256 = "d4d36b72718a4076b7c0561abfa2edc006a1f9a4491762979f86c3dd99a71fdc"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-199"
@@ -24871,7 +25424,7 @@ claim_text = "With action values drawn from the environment, the mechanically tr
claim_status = "qualified"
lean_declaration = "Auths.Refinement.Observation.translated_observation_conditions_hold_refines_conditions_hold"
lean_statement_sha256 = "4d030ad8523e9b4e7031109a2d67e482fbb681b9424036d53496178523be7493"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = [
"auths_model::observation::observation_conditions_hold",
"auths_model::observation::observation_condition_holds",
@@ -24942,7 +25495,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The conjunction loop when each eqAction condition's supplied action value is the environment's action fact for its reference."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, the listed foundational axioms, and the theorem's explicit representation-validity premises (every compared text fits the u32 UTF-8 byte bound) are trusted.",
@@ -24968,7 +25521,7 @@ sha256 = "d4d36b72718a4076b7c0561abfa2edc006a1f9a4491762979f86c3dd99a71fdc"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-200"
@@ -24976,7 +25529,7 @@ claim_text = "The mechanically translated verdict combinator returns exactly the
claim_status = "qualified"
lean_declaration = "Auths.Refinement.Observation.translated_requirement_verdict_refines_model"
lean_statement_sha256 = "3ca419beb07857d75b92c8fb57b48431a8db672f27ec1ac81c8a112a920b40d0"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_model::observation::requirement_verdict"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -25038,7 +25591,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Satisfied before denied before indeterminate, for every pair of eligibility and satisfaction inputs."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, the listed foundational axioms, and the theorem's explicit representation-validity premises (every compared text fits the u32 UTF-8 byte bound) are trusted.",
@@ -25063,7 +25616,7 @@ sha256 = "d4d36b72718a4076b7c0561abfa2edc006a1f9a4491762979f86c3dd99a71fdc"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-201"
@@ -25071,7 +25624,7 @@ claim_text = "When every referenced action fact is available, the mechanically t
claim_status = "qualified"
lean_declaration = "Auths.Refinement.Observation.translated_requirement_verdict_refines_requirement_decision"
lean_statement_sha256 = "52a0068811cc19be5605093531410d061f617ca4f3b9a5b1f056d6949077f118"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_model::observation::requirement_verdict"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -25133,7 +25686,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The per-requirement verdict when the requirement's action facts are available; unavailable action facts are indeterminate in the model."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, the listed foundational axioms, and the theorem's explicit representation-validity premises (every compared text fits the u32 UTF-8 byte bound) are trusted.",
@@ -25158,7 +25711,7 @@ sha256 = "d4d36b72718a4076b7c0561abfa2edc006a1f9a4491762979f86c3dd99a71fdc"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-271"
@@ -25166,7 +25719,7 @@ claim_text = "The mechanically translated observer-anchor comparison returns exa
claim_status = "qualified"
lean_declaration = "Auths.Refinement.Observation.translated_observer_anchor_id_equal_refines_model"
lean_statement_sha256 = "564453fe589aab144b7c4696318d798cb9489197156d1ee8d88cc68ca4efca6d"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_model::observation::observer_anchor_id_equal"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -25228,7 +25781,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "UTF-8 byte comparison of two observer-anchor identifiers, bridged to String equality."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, the listed foundational axioms, and the theorem's explicit representation-validity premises (every compared text fits the u32 UTF-8 byte bound) are trusted.",
@@ -25254,7 +25807,7 @@ sha256 = "d4d36b72718a4076b7c0561abfa2edc006a1f9a4491762979f86c3dd99a71fdc"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-272"
@@ -25262,7 +25815,7 @@ claim_text = "The mechanically translated schema comparison returns exactly the
claim_status = "qualified"
lean_declaration = "Auths.Refinement.Observation.translated_observation_schema_equal_refines_model"
lean_statement_sha256 = "4b027d678991524961d7103204315cda5f7af4cd88a06f875db770fbfc94bd83"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_model::observation::observation_schema_equal"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -25324,7 +25877,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "UTF-8 byte comparison of two observation-schema identifiers, bridged to String equality."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, the listed foundational axioms, and the theorem's explicit representation-validity premises (every compared text fits the u32 UTF-8 byte bound) are trusted.",
@@ -25350,7 +25903,7 @@ sha256 = "d4d36b72718a4076b7c0561abfa2edc006a1f9a4491762979f86c3dd99a71fdc"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-273"
@@ -25358,7 +25911,7 @@ claim_text = "The mechanically translated subject comparison returns exactly the
claim_status = "qualified"
lean_declaration = "Auths.Refinement.Observation.translated_requirement_subject_equal_refines_model"
lean_statement_sha256 = "efbd6e09d1ed52b992044847b3a47c6f7f7bcea46410b284e140355071a4fb2a"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = [
"auths_model::observation::requirement_subject_equal",
"auths_model::observation::fact_name_equal",
@@ -25423,7 +25976,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Literal-resource and action-fact subjects, including every cross-kind pair."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, the listed foundational axioms, and the theorem's explicit representation-validity premises (every compared text fits the u32 UTF-8 byte bound) are trusted.",
@@ -25449,7 +26002,7 @@ sha256 = "d4d36b72718a4076b7c0561abfa2edc006a1f9a4491762979f86c3dd99a71fdc"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-274"
@@ -25457,7 +26010,7 @@ claim_text = "The mechanically translated membership-list comparison returns exa
claim_status = "qualified"
lean_declaration = "Auths.Refinement.Observation.translated_member_values_equal_refines_model"
lean_statement_sha256 = "5c5ad163115e51cbe1d70ccb962047e73e69ff7fa94111d0187ce2429a83d0bb"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = [
"auths_model::observation::member_values_equal",
"auths_model::observation::fact_value_equal",
@@ -25522,7 +26075,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The length guard and the terminating positional loop over two finite literal lists."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, the listed foundational axioms, and the theorem's explicit representation-validity premises (every compared text fits the u32 UTF-8 byte bound) are trusted.",
@@ -25548,7 +26101,7 @@ sha256 = "d4d36b72718a4076b7c0561abfa2edc006a1f9a4491762979f86c3dd99a71fdc"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-275"
@@ -25556,7 +26109,7 @@ claim_text = "The mechanically translated condition-atom comparison returns exac
claim_status = "qualified"
lean_declaration = "Auths.Refinement.Observation.translated_condition_test_equal_refines_model"
lean_statement_sha256 = "e3c6bf03b6853dbedee9fb11e2a4c93c599b64f24f59ae7a13cf742af8c07df3"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = [
"auths_model::observation::condition_test_equal",
"auths_model::observation::fact_value_equal",
@@ -25623,7 +26176,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "All four condition atoms, including every cross-atom pair."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, the listed foundational axioms, and the theorem's explicit representation-validity premises (every compared text fits the u32 UTF-8 byte bound) are trusted.",
@@ -25649,7 +26202,7 @@ sha256 = "d4d36b72718a4076b7c0561abfa2edc006a1f9a4491762979f86c3dd99a71fdc"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-276"
@@ -25657,7 +26210,7 @@ claim_text = "The mechanically translated condition comparison returns exactly t
claim_status = "qualified"
lean_declaration = "Auths.Refinement.Observation.translated_observation_condition_equal_refines_model"
lean_statement_sha256 = "44b638fa8228688cc4f4d77ad87bd9fbaf9a9b2eb037cb846e62d8caa6611447"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = [
"auths_model::observation::observation_condition_equal",
"auths_model::observation::fact_name_equal",
@@ -25723,7 +26276,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The fact name and the condition atom of two named conditions."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, the listed foundational axioms, and the theorem's explicit representation-validity premises (every compared text fits the u32 UTF-8 byte bound) are trusted.",
@@ -25749,7 +26302,7 @@ sha256 = "d4d36b72718a4076b7c0561abfa2edc006a1f9a4491762979f86c3dd99a71fdc"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-277"
@@ -25757,7 +26310,7 @@ claim_text = "The mechanically translated containment loop returns exactly the m
claim_status = "qualified"
lean_declaration = "Auths.Refinement.Observation.translated_observation_conditions_contain_refines_model"
lean_statement_sha256 = "688fb6edaf3b5e37be4d7823f1fc273e24a61e7299d2ef3eb0ef9b877159c9b2"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = [
"auths_model::observation::observation_conditions_contain",
"auths_model::observation::observation_condition_equal",
@@ -25822,7 +26375,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The terminating scan of a condition slice for one condition."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, the listed foundational axioms, and the theorem's explicit representation-validity premises (every compared text fits the u32 UTF-8 byte bound) are trusted.",
@@ -25848,7 +26401,7 @@ sha256 = "d4d36b72718a4076b7c0561abfa2edc006a1f9a4491762979f86c3dd99a71fdc"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-278"
@@ -25856,7 +26409,7 @@ claim_text = "The mechanically translated inclusion loop returns exactly the mod
claim_status = "qualified"
lean_declaration = "Auths.Refinement.Observation.translated_observation_conditions_include_refines_model"
lean_statement_sha256 = "96280dce9a82d09c459e502cdd7dc12f413aedb86d6846eeee90627795910722"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = [
"auths_model::observation::observation_conditions_include",
"auths_model::observation::observation_conditions_contain",
@@ -25922,7 +26475,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Every atom of the wider slice occurring in the narrower slice, compared as a set of canonical atoms."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, the listed foundational axioms, and the theorem's explicit representation-validity premises (every compared text fits the u32 UTF-8 byte bound) are trusted.",
@@ -25948,7 +26501,7 @@ sha256 = "d4d36b72718a4076b7c0561abfa2edc006a1f9a4491762979f86c3dd99a71fdc"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-279"
@@ -25956,7 +26509,7 @@ claim_text = "The mechanically translated condition-list comparison returns exac
claim_status = "qualified"
lean_declaration = "Auths.Refinement.Observation.translated_observation_conditions_equal_refines_model"
lean_statement_sha256 = "a235b885de3a11bd2db81a9b91c30f0a03082254b03dbb20cc561af3fb304eff"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = [
"auths_model::observation::observation_conditions_equal",
"auths_model::observation::observation_condition_equal",
@@ -26021,7 +26574,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The length guard and the terminating positional loop over two condition slices."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, the listed foundational axioms, and the theorem's explicit representation-validity premises (every compared text fits the u32 UTF-8 byte bound) are trusted.",
@@ -26047,7 +26600,7 @@ sha256 = "d4d36b72718a4076b7c0561abfa2edc006a1f9a4491762979f86c3dd99a71fdc"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-280"
@@ -26055,7 +26608,7 @@ claim_text = "The mechanically translated same-target check returns exactly the
claim_status = "qualified"
lean_declaration = "Auths.Refinement.Observation.translated_observation_requirement_same_target_refines_model"
lean_statement_sha256 = "b79e3eaa4bb4270a1d44d9c62d457e4b3e396250ac36588a79068cd2d513e84b"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = [
"auths_model::observation::observation_requirement_same_target",
"auths_model::observation::observer_anchor_id_equal",
@@ -26122,7 +26675,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The observer anchor, schema, and subject of two requirements."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, the listed foundational axioms, and the theorem's explicit representation-validity premises (every compared text fits the u32 UTF-8 byte bound) are trusted.",
@@ -26148,7 +26701,7 @@ sha256 = "d4d36b72718a4076b7c0561abfa2edc006a1f9a4491762979f86c3dd99a71fdc"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-281"
@@ -26156,7 +26709,7 @@ claim_text = "The mechanically translated requirement comparison returns exactly
claim_status = "qualified"
lean_declaration = "Auths.Refinement.Observation.translated_observation_requirement_equal_refines_model"
lean_statement_sha256 = "f78c2bb3bae9cdc9b1fb76d183c691737066bcc4ee4781fac6b82b335bba184c"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = [
"auths_model::observation::observation_requirement_equal",
"auths_model::observation::observation_requirement_same_target",
@@ -26222,7 +26775,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The target, the u64 maximum age, and the positional condition list of two requirements."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, the listed foundational axioms, and the theorem's explicit representation-validity premises (every compared text fits the u32 UTF-8 byte bound) are trusted.",
@@ -26248,7 +26801,7 @@ sha256 = "d4d36b72718a4076b7c0561abfa2edc006a1f9a4491762979f86c3dd99a71fdc"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-282"
@@ -26256,7 +26809,7 @@ claim_text = "The mechanically translated narrowing check returns exactly the mo
claim_status = "qualified"
lean_declaration = "Auths.Refinement.Observation.translated_observation_requirement_narrows_refines_model"
lean_statement_sha256 = "46ad4c556c162406b17661c87f1e1d246d960a3be471a4441a81012ba12711c1"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = [
"auths_model::observation::observation_requirement_narrows",
"auths_model::observation::observation_requirement_same_target",
@@ -26322,7 +26875,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The same target, a maximum age no larger, a superset of condition atoms, and a strictly narrower age or atom set."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, the listed foundational axioms, and the theorem's explicit representation-validity premises (every compared text fits the u32 UTF-8 byte bound) are trusted.",
@@ -26348,7 +26901,7 @@ sha256 = "d4d36b72718a4076b7c0561abfa2edc006a1f9a4491762979f86c3dd99a71fdc"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-283"
@@ -26356,7 +26909,7 @@ claim_text = "The mechanically translated cover check returns exactly the model'
claim_status = "qualified"
lean_declaration = "Auths.Refinement.Observation.translated_observation_requirement_covers_refines_model"
lean_statement_sha256 = "9cbb882a6b66b35880cbb5fbb594262a8d665894b107bc8681332665bc3efc21"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = [
"auths_model::observation::observation_requirement_covers",
"auths_model::observation::observation_requirement_equal",
@@ -26422,7 +26975,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "An identical requirement or a strict narrowing."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, the listed foundational axioms, and the theorem's explicit representation-validity premises (every compared text fits the u32 UTF-8 byte bound) are trusted.",
@@ -26448,7 +27001,7 @@ sha256 = "d4d36b72718a4076b7c0561abfa2edc006a1f9a4491762979f86c3dd99a71fdc"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-284"
@@ -26456,7 +27009,7 @@ claim_text = "The mechanically translated retention loop returns exactly the mod
claim_status = "qualified"
lean_declaration = "Auths.Refinement.Observation.translated_observation_requirement_retained_refines_model"
lean_statement_sha256 = "db23559560127b12376f178ce5a707aead02785eb5451e8aea0616a02d40e4a2"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = [
"auths_model::observation::observation_requirement_retained",
"auths_model::observation::observation_requirement_covers",
@@ -26521,7 +27074,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The terminating scan of a child requirement list for one parent requirement."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, the listed foundational axioms, and the theorem's explicit representation-validity premises (every compared text fits the u32 UTF-8 byte bound) are trusted.",
@@ -26547,7 +27100,7 @@ sha256 = "d4d36b72718a4076b7c0561abfa2edc006a1f9a4491762979f86c3dd99a71fdc"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-285"
@@ -26555,7 +27108,7 @@ claim_text = "The mechanically translated observation-requirement attenuation la
claim_status = "qualified"
lean_declaration = "Auths.Refinement.Observation.translated_observation_requirements_attenuate_refines_model"
lean_statement_sha256 = "e1c90caefc8cea86b905fce1bdaaaf489a1f64bcb46e2d9645a41268a45dd683"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = [
"auths_model::observation::observation_requirements_attenuate",
"auths_model::observation::observation_requirement_retained",
@@ -26623,7 +27176,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Every parent requirement covered by some child requirement, over requirement lists of any length."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, the listed foundational axioms, and the theorem's explicit representation-validity premises (every compared text fits the u32 UTF-8 byte bound) are trusted.",
@@ -26649,7 +27202,7 @@ sha256 = "d4d36b72718a4076b7c0561abfa2edc006a1f9a4491762979f86c3dd99a71fdc"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-306"
@@ -26657,7 +27210,7 @@ claim_text = "The mechanically translated approver verdict returns exactly the m
claim_status = "qualified"
lean_declaration = "Auths.Refinement.Approval.translated_approver_verdict_refines_model"
lean_statement_sha256 = "96a93cb521f7ceee6550abe06fd102d42204467d4c97e1c0b2d19a4175f035c9"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_model::approval::approver_verdict"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -26719,7 +27272,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The terminating scan of per-approval outcomes for one approver: counted when some outcome is counted, pending when none is and some is pending, ignored otherwise."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, the listed foundational axioms, and the theorem's explicit representation-validity premises (every compared principal fits the u32 UTF-8 byte bound) are trusted.",
@@ -26745,7 +27298,7 @@ sha256 = "d4d36b72718a4076b7c0561abfa2edc006a1f9a4491762979f86c3dd99a71fdc"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-307"
@@ -26753,7 +27306,7 @@ claim_text = "The mechanically translated distinct counting returns exactly the
claim_status = "qualified"
lean_declaration = "Auths.Refinement.Approval.translated_approval_counts_refines_model"
lean_statement_sha256 = "36686ea58c4b0815ecfc6c85715c734d24d3273b4956ce66d58aaffa2abc5f54"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = [
"auths_model::approval::approval_counts",
"auths_model::approval::approver_verdict",
@@ -26818,7 +27371,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The terminating loop over a requirement's listed approvers, each counted once, for every outcome list and every u16 threshold."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, the listed foundational axioms, and the theorem's explicit representation-validity premises (every compared principal fits the u32 UTF-8 byte bound) are trusted.",
@@ -26844,7 +27397,7 @@ sha256 = "d4d36b72718a4076b7c0561abfa2edc006a1f9a4491762979f86c3dd99a71fdc"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-308"
@@ -26852,7 +27405,7 @@ claim_text = "The mechanically translated principal-list membership returns exac
claim_status = "qualified"
lean_declaration = "Auths.Refinement.Approval.translated_principal_ids_contain_refines_model"
lean_statement_sha256 = "9c377995783f05529ff05f6655022584e0f6669d7acdf3ef1e57014e25ee1ec0"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = ["auths_model::approval::principal_ids_contain"]
semantic_source_closure = [
"formal/Auths.lean",
@@ -26914,7 +27467,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The terminating scan of a principal slice for one principal, compared by UTF-8 bytes bridged to String equality."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, the listed foundational axioms, and the theorem's explicit representation-validity premises (every compared principal fits the u32 UTF-8 byte bound) are trusted.",
@@ -26940,7 +27493,7 @@ sha256 = "d4d36b72718a4076b7c0561abfa2edc006a1f9a4491762979f86c3dd99a71fdc"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-309"
@@ -26948,7 +27501,7 @@ claim_text = "The mechanically translated principal-list inclusion returns exact
claim_status = "qualified"
lean_declaration = "Auths.Refinement.Approval.translated_principal_ids_subset_refines_model"
lean_statement_sha256 = "77c6b790409d4c2e7277d80217281bba70f67e6f21a3a23841c3110dd625f36f"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = [
"auths_model::approval::principal_ids_subset",
"auths_model::approval::principal_ids_contain",
@@ -27013,7 +27566,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Every principal of the narrower slice occurring in the wider slice."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, the listed foundational axioms, and the theorem's explicit representation-validity premises (every compared principal fits the u32 UTF-8 byte bound) are trusted.",
@@ -27039,7 +27592,7 @@ sha256 = "d4d36b72718a4076b7c0561abfa2edc006a1f9a4491762979f86c3dd99a71fdc"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-310"
@@ -27047,7 +27600,7 @@ claim_text = "The mechanically translated cover check returns exactly the model'
claim_status = "qualified"
lean_declaration = "Auths.Refinement.Approval.translated_approval_requirement_covers_refines_model"
lean_statement_sha256 = "f36b9195ab1f24fad69a6deccc01d1f9895b7803fdf243dd948595f4744f73f7"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = [
"auths_model::approval::approval_requirement_covers",
"auths_model::approval::principal_ids_subset",
@@ -27113,7 +27666,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "One child and one parent approval requirement."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, the listed foundational axioms, and the theorem's explicit representation-validity premises (every compared principal fits the u32 UTF-8 byte bound) are trusted.",
@@ -27139,7 +27692,7 @@ sha256 = "d4d36b72718a4076b7c0561abfa2edc006a1f9a4491762979f86c3dd99a71fdc"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-311"
@@ -27147,7 +27700,7 @@ claim_text = "The mechanically translated retention loop returns exactly the mod
claim_status = "qualified"
lean_declaration = "Auths.Refinement.Approval.translated_approval_requirement_retained_refines_model"
lean_statement_sha256 = "3b4e9f2ae9beff66fef142c2ebefc11dd86bd65202950c55914458bff46d6c8a"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = [
"auths_model::approval::approval_requirement_retained",
"auths_model::approval::approval_requirement_covers",
@@ -27212,7 +27765,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "The terminating scan of a child requirement list for one parent requirement."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, the listed foundational axioms, and the theorem's explicit representation-validity premises (every compared principal fits the u32 UTF-8 byte bound) are trusted.",
@@ -27238,7 +27791,7 @@ sha256 = "d4d36b72718a4076b7c0561abfa2edc006a1f9a4491762979f86c3dd99a71fdc"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
[[claims]]
claim_id = "AP-FORMAL-RICH-312"
@@ -27246,7 +27799,7 @@ claim_text = "The mechanically translated approval-requirement-v1 attenuation la
claim_status = "qualified"
lean_declaration = "Auths.Refinement.Approval.translated_approval_requirements_attenuate_refines_model"
lean_statement_sha256 = "4eab09e98c4ff9d49768010b51dad1f0e7f5807c9899733a5ead580f7dc797c1"
-formal_review = "native-k-of-n-approvals-2026-09-30"
+formal_review = "ap074-durable-entry-and-late-evidence-2026-10-29"
rust_symbols = [
"auths_model::approval::approval_requirements_attenuate",
"auths_model::approval::approval_requirement_retained",
@@ -27313,7 +27866,7 @@ semantic_source_closure = [
"formal/qualification/aeneas/generated/model/FunsExternal.lean",
"formal/qualification/aeneas/generated/model/Types.lean",
]
-semantic_source_closure_sha256 = "a3c4451de61c9b0e5a5e01b2db5e9e428150a5da66adcf2ee7a3e22abdefe986"
+semantic_source_closure_sha256 = "d8398eb7a07ecb3faa82380718c7a20af70d26c96cbac7f229b16240fb7f3659"
scope = "Every parent requirement covered by some child requirement, over requirement lists of any length."
residual_assumptions = [
"Lean's kernel, the pinned Rust/Charon/Aeneas/Lean toolchain, the reviewed transparent external bridges, the listed foundational axioms, and the theorem's explicit representation-validity premises (every compared principal fits the u32 UTF-8 byte bound) are trusted.",
@@ -27339,4 +27892,4 @@ sha256 = "d4d36b72718a4076b7c0561abfa2edc006a1f9a4491762979f86c3dd99a71fdc"
[[claims.evidence]]
kind = "source-closure"
artifact = "formal/qualification/aeneas/source-closure.json"
-sha256 = "f2fdfa9bfe3a92173af2ac0da3a376f104c7d87398f9d1806dad13193e578362"
+sha256 = "229b327c87cecc1405b12dcf0d2b36e733629c8bb5f96b1762fa1fe6d075b769"
diff --git a/formal/lean-assurance-audit-v1.json b/formal/lean-assurance-audit-v1.json
index 0c14778c6..0fcbf11e6 100644
--- a/formal/lean-assurance-audit-v1.json
+++ b/formal/lean-assurance-audit-v1.json
@@ -1596,7 +1596,43 @@
"kind": "theorem",
"module": "Auths.Product.Recovery",
"name": "Auths.Product.Recovery.unknown_resolves_only_linked",
- "statement": "(∀ (old new : Auths.Product.Recovery.AttemptView),\n Auths.Product.Recovery.validTransition old new = true →\n old.stage = Auths.Product.Recovery.Stage.unknown →\n new.stage = Auths.Product.Recovery.Stage.observedByProvider ∧\n old.link = Auths.Product.Recovery.Link.verified ∧\n new.link = Auths.Product.Recovery.Link.verified ∧ new.evidence = true) ∧\n (∀ (views : List Auths.Product.Recovery.AttemptView),\n Auths.Product.Recovery.validChain views →\n ∀ (index : ℕ) (bound : index + 1 < views.length),\n views[index].stage = Auths.Product.Recovery.Stage.unknown →\n views[index + 1].stage = Auths.Product.Recovery.Stage.observedByProvider ∧\n views[index + 1].link = Auths.Product.Recovery.Link.verified ∧ views[index + 1].evidence = true) ∧\n ∀ (declarations : Auths.Product.Recovery.Declarations),\n Auths.Product.Recovery.linkOf (Auths.Product.Recovery.capability declarations).providerLink =\n Auths.Product.Recovery.Link.verified ↔\n (Auths.Product.Recovery.capability declarations).recoveryClass = Auths.Product.Recovery.RecoveryClass.linked"
+ "statement": "(∀ (old new : Auths.Product.Recovery.AttemptView),\n Auths.Product.Recovery.validTransition old new = true →\n old.stage = Auths.Product.Recovery.Stage.unknown →\n new.stage = Auths.Product.Recovery.Stage.observedByProvider →\n old.link = Auths.Product.Recovery.Link.verified ∧\n new.link = Auths.Product.Recovery.Link.verified ∧ new.evidence = true) ∧\n (∀ (views : List Auths.Product.Recovery.AttemptView),\n Auths.Product.Recovery.validChain views →\n ∀ (index : ℕ) (bound : index + 1 < views.length),\n views[index].stage = Auths.Product.Recovery.Stage.unknown →\n views[index + 1].stage = Auths.Product.Recovery.Stage.observedByProvider →\n views[index + 1].link = Auths.Product.Recovery.Link.verified ∧ views[index + 1].evidence = true) ∧\n ∀ (declarations : Auths.Product.Recovery.Declarations),\n Auths.Product.Recovery.linkOf (Auths.Product.Recovery.capability declarations).providerLink =\n Auths.Product.Recovery.Link.verified ↔\n (Auths.Product.Recovery.capability declarations).recoveryClass = Auths.Product.Recovery.RecoveryClass.linked"
+ },
+ {
+ "axioms": [
+ "propext"
+ ],
+ "kind": "theorem",
+ "module": "Auths.Product.Recovery",
+ "name": "Auths.Product.Recovery.unknown_exits_with_bound_evidence",
+ "statement": "∀ (old new : Auths.Product.Recovery.AttemptView),\n Auths.Product.Recovery.validTransition old new = true →\n old.stage = Auths.Product.Recovery.Stage.unknown →\n new.stage = Auths.Product.Recovery.Stage.responseRecorded ∨\n new.stage = Auths.Product.Recovery.Stage.acknowledged ∨\n new.stage = Auths.Product.Recovery.Stage.observedByProvider ∧ old.link = Auths.Product.Recovery.Link.verified"
+ },
+ {
+ "axioms": [
+ "propext"
+ ],
+ "kind": "theorem",
+ "module": "Auths.Product.Recovery",
+ "name": "Auths.Product.Recovery.transition_ownership",
+ "statement": "∀ {old new : Auths.Product.Recovery.AttemptView},\n Auths.Product.Recovery.validTransition old new = true →\n new.protocol = old.protocol ∧\n new.revision = old.revision + 1 ∧ new.revision ≤ 9007199254740991 ∧ (old.entry ≠ [] → new.entry = old.entry)"
+ },
+ {
+ "axioms": [
+ "propext"
+ ],
+ "kind": "theorem",
+ "module": "Auths.Product.Recovery",
+ "name": "Auths.Product.Recovery.late_response_retains_source",
+ "statement": "∀ (old new : Auths.Product.Recovery.AttemptView),\n Auths.Product.Recovery.validTransition old new = true →\n old.stage = Auths.Product.Recovery.Stage.unknown →\n new.stage = Auths.Product.Recovery.Stage.responseRecorded →\n new.protocol = old.protocol ∧\n new.revision = old.revision + 1 ∧\n new.entry = old.entry ∧ old.entry ≠ [] ∧ new.response ≠ [] ∧ new.evidence = false ∧ new.acknowledgement = []"
+ },
+ {
+ "axioms": [
+ "propext"
+ ],
+ "kind": "theorem",
+ "module": "Auths.Product.Recovery",
+ "name": "Auths.Product.Recovery.late_acknowledgement_retains_source",
+ "statement": "∀ (old new : Auths.Product.Recovery.AttemptView),\n Auths.Product.Recovery.validTransition old new = true →\n old.stage = Auths.Product.Recovery.Stage.unknown →\n new.stage = Auths.Product.Recovery.Stage.acknowledged →\n old.protocol = Auths.Product.Recovery.Protocol.disclosure ∧\n new.protocol = Auths.Product.Recovery.Protocol.disclosure ∧\n new.revision = old.revision + 1 ∧\n new.entry = old.entry ∧\n old.entry ≠ [] ∧ new.acknowledgement ≠ [] ∧ new.response = [] ∧ new.evidence = false"
},
{
"axioms": [
@@ -1607,7 +1643,7 @@
"kind": "theorem",
"module": "Auths.Product.Recovery",
"name": "Auths.Product.Recovery.provider_claim_requires_evidence",
- "statement": "(∀ (old new : Auths.Product.Recovery.AttemptView),\n Auths.Product.Recovery.validTransition old new = true →\n new.stage = Auths.Product.Recovery.Stage.observedByProvider →\n new.evidence = true ∧\n Auths.Product.Recovery.hasLink new.link = true ∧\n new.link = old.link ∧\n (old.stage = Auths.Product.Recovery.Stage.responseRecorded ∨\n (old.stage = Auths.Product.Recovery.Stage.attempting ∨\n old.stage = Auths.Product.Recovery.Stage.unknown) ∧\n old.link = Auths.Product.Recovery.Link.verified)) ∧\n ∀ (first : Auths.Product.Recovery.AttemptView) (rest : List Auths.Product.Recovery.AttemptView),\n first.stage = Auths.Product.Recovery.Stage.attempting →\n Auths.Product.Recovery.consistent first = true →\n Auths.Product.Recovery.validChain (first :: rest) →\n ∀ (index : ℕ) (bound : index < rest.length),\n rest[index].stage = Auths.Product.Recovery.Stage.observedByProvider →\n rest[index].evidence = true ∧\n rest[index].link = first.link ∧\n Auths.Product.Recovery.hasLink first.link = true ∧\n ((first :: rest)[index].stage = Auths.Product.Recovery.Stage.responseRecorded ∨\n ((first :: rest)[index].stage = Auths.Product.Recovery.Stage.attempting ∨\n (first :: rest)[index].stage = Auths.Product.Recovery.Stage.unknown) ∧\n first.link = Auths.Product.Recovery.Link.verified)"
+ "statement": "(∀ (old new : Auths.Product.Recovery.AttemptView),\n Auths.Product.Recovery.validTransition old new = true →\n new.stage = Auths.Product.Recovery.Stage.observedByProvider →\n new.evidence = true ∧\n Auths.Product.Recovery.hasLink new.link = true ∧\n new.link = old.link ∧\n (old.stage = Auths.Product.Recovery.Stage.responseRecorded ∨\n (old.stage = Auths.Product.Recovery.Stage.entryGranted ∨\n old.stage = Auths.Product.Recovery.Stage.unknown) ∧\n old.link = Auths.Product.Recovery.Link.verified)) ∧\n ∀ (first : Auths.Product.Recovery.AttemptView) (rest : List Auths.Product.Recovery.AttemptView),\n first.stage = Auths.Product.Recovery.Stage.reserved →\n Auths.Product.Recovery.consistent first = true →\n Auths.Product.Recovery.validChain (first :: rest) →\n ∀ (index : ℕ) (bound : index < rest.length),\n rest[index].stage = Auths.Product.Recovery.Stage.observedByProvider →\n rest[index].evidence = true ∧\n rest[index].link = first.link ∧\n Auths.Product.Recovery.hasLink first.link = true ∧\n ((first :: rest)[index].stage = Auths.Product.Recovery.Stage.responseRecorded ∨\n ((first :: rest)[index].stage = Auths.Product.Recovery.Stage.entryGranted ∨\n (first :: rest)[index].stage = Auths.Product.Recovery.Stage.unknown) ∧\n first.link = Auths.Product.Recovery.Link.verified)"
},
{
"axioms": [
@@ -1617,7 +1653,7 @@
"kind": "theorem",
"module": "Auths.Product.Recovery",
"name": "Auths.Product.Recovery.terminal_stages_final",
- "statement": "(∀ (stage : Auths.Product.Recovery.Stage),\n Auths.Product.Recovery.terminalStage stage = true ↔\n stage = Auths.Product.Recovery.Stage.notEntered ∨\n stage = Auths.Product.Recovery.Stage.observed ∨ stage = Auths.Product.Recovery.Stage.observedByProvider) ∧\n (∀ (old new : Auths.Product.Recovery.AttemptView),\n Auths.Product.Recovery.terminalStage old.stage = true → Auths.Product.Recovery.validTransition old new = false) ∧\n ∀ (views : List Auths.Product.Recovery.AttemptView),\n Auths.Product.Recovery.validChain views →\n ∀ (index : ℕ) (bound : index < views.length),\n Auths.Product.Recovery.terminalStage views[index].stage = true → index + 1 = views.length"
+ "statement": "(∀ (stage : Auths.Product.Recovery.Stage),\n Auths.Product.Recovery.terminalStage stage = true ↔\n stage = Auths.Product.Recovery.Stage.notEntered ∨\n stage = Auths.Product.Recovery.Stage.observed ∨\n stage = Auths.Product.Recovery.Stage.observedByProvider ∨ stage = Auths.Product.Recovery.Stage.acknowledged) ∧\n (∀ (old new : Auths.Product.Recovery.AttemptView),\n Auths.Product.Recovery.terminalStage old.stage = true → Auths.Product.Recovery.validTransition old new = false) ∧\n ∀ (views : List Auths.Product.Recovery.AttemptView),\n Auths.Product.Recovery.validChain views →\n ∀ (index : ℕ) (bound : index < views.length),\n Auths.Product.Recovery.terminalStage views[index].stage = true → index + 1 = views.length"
},
{
"axioms": [
@@ -1638,7 +1674,7 @@
"kind": "theorem",
"module": "Auths.Product.SubmitOrder",
"name": "Auths.Product.SubmitOrder.lease_requires_verified_claim",
- "statement": "∀ (plan : Auths.Product.SubmitOrder.Plan) (pre : List Auths.Product.SubmitOrder.Event)\n (event : Auths.Product.SubmitOrder.Event) (mode : Auths.Product.SubmitOrder.Mode),\n (Auths.Product.SubmitOrder.nextStep (Auths.Product.SubmitOrder.stateAfter plan pre) event).action =\n Auths.Product.SubmitOrder.Action.lease mode →\n event = Auths.Product.SubmitOrder.Event.reload true ∧\n (Auths.Product.SubmitOrder.stateAfter plan pre).phase = Auths.Product.SubmitOrder.Phase.reload mode ∧\n (∃ argument, Auths.Product.SubmitOrder.Verified (Auths.Product.SubmitOrder.steps plan pre) argument) ∧\n (mode = Auths.Product.SubmitOrder.Mode.reobserve →\n Auths.Product.SubmitOrder.Answered (Auths.Product.SubmitOrder.steps plan pre)\n Auths.Product.SubmitOrder.Phase.claim\n (Auths.Product.SubmitOrder.Event.claim Auths.Product.SubmitOrder.ClaimResult.replay)) ∧\n (mode ≠ Auths.Product.SubmitOrder.Mode.reobserve →\n Auths.Product.SubmitOrder.Answered (Auths.Product.SubmitOrder.steps plan pre)\n Auths.Product.SubmitOrder.Phase.claim\n (Auths.Product.SubmitOrder.Event.claim Auths.Product.SubmitOrder.ClaimResult.inserted))"
+ "statement": "∀ (plan : Auths.Product.SubmitOrder.Plan) (pre : List Auths.Product.SubmitOrder.Event)\n (event : Auths.Product.SubmitOrder.Event),\n (Auths.Product.SubmitOrder.nextStep (Auths.Product.SubmitOrder.stateAfter plan pre) event).action =\n Auths.Product.SubmitOrder.Action.lease →\n event = Auths.Product.SubmitOrder.Event.reload true ∧\n (Auths.Product.SubmitOrder.stateAfter plan pre).phase = Auths.Product.SubmitOrder.Phase.reload ∧\n (∃ argument, Auths.Product.SubmitOrder.Verified (Auths.Product.SubmitOrder.steps plan pre) argument) ∧\n Auths.Product.SubmitOrder.Answered (Auths.Product.SubmitOrder.steps plan pre)\n Auths.Product.SubmitOrder.Phase.claim\n (Auths.Product.SubmitOrder.Event.claim Auths.Product.SubmitOrder.ClaimResult.reserved)"
},
{
"axioms": [
@@ -1648,7 +1684,7 @@
"kind": "theorem",
"module": "Auths.Product.SubmitOrder",
"name": "Auths.Product.SubmitOrder.send_requires_lease_and_deadline",
- "statement": "∀ (plan : Auths.Product.SubmitOrder.Plan) (pre : List Auths.Product.SubmitOrder.Event)\n (event : Auths.Product.SubmitOrder.Event),\n (Auths.Product.SubmitOrder.nextStep (Auths.Product.SubmitOrder.stateAfter plan pre) event).action =\n Auths.Product.SubmitOrder.Action.send →\n event = Auths.Product.SubmitOrder.Event.deadline true ∧\n (Auths.Product.SubmitOrder.stateAfter plan pre).phase = Auths.Product.SubmitOrder.Phase.deadline ∧\n (∃ argument, Auths.Product.SubmitOrder.Verified (Auths.Product.SubmitOrder.steps plan pre) argument) ∧\n Auths.Product.SubmitOrder.Answered (Auths.Product.SubmitOrder.steps plan pre)\n Auths.Product.SubmitOrder.Phase.claim\n (Auths.Product.SubmitOrder.Event.claim Auths.Product.SubmitOrder.ClaimResult.inserted) ∧\n Auths.Product.SubmitOrder.Answered (Auths.Product.SubmitOrder.steps plan pre)\n (Auths.Product.SubmitOrder.Phase.lease Auths.Product.SubmitOrder.Mode.entry)\n (Auths.Product.SubmitOrder.Event.lease true) ∧\n (plan.preEntry = true →\n Auths.Product.SubmitOrder.Answered (Auths.Product.SubmitOrder.steps plan pre)\n Auths.Product.SubmitOrder.Phase.preEntry\n (Auths.Product.SubmitOrder.Event.preEntry Auths.Product.SubmitOrder.PreEntryResult.satisfied)) ∧\n Auths.Product.SubmitOrder.Answered (Auths.Product.SubmitOrder.steps plan pre)\n Auths.Product.SubmitOrder.Phase.entryReload (Auths.Product.SubmitOrder.Event.reload true)"
+ "statement": "∀ (plan : Auths.Product.SubmitOrder.Plan) (pre : List Auths.Product.SubmitOrder.Event)\n (event : Auths.Product.SubmitOrder.Event),\n (Auths.Product.SubmitOrder.nextStep (Auths.Product.SubmitOrder.stateAfter plan pre) event).action =\n Auths.Product.SubmitOrder.Action.send →\n event = Auths.Product.SubmitOrder.Event.granted true ∧\n (Auths.Product.SubmitOrder.stateAfter plan pre).phase = Auths.Product.SubmitOrder.Phase.grant ∧\n (∃ argument, Auths.Product.SubmitOrder.Verified (Auths.Product.SubmitOrder.steps plan pre) argument) ∧\n Auths.Product.SubmitOrder.Answered (Auths.Product.SubmitOrder.steps plan pre)\n Auths.Product.SubmitOrder.Phase.claim\n (Auths.Product.SubmitOrder.Event.claim Auths.Product.SubmitOrder.ClaimResult.reserved) ∧\n Auths.Product.SubmitOrder.Answered (Auths.Product.SubmitOrder.steps plan pre)\n Auths.Product.SubmitOrder.Phase.lease (Auths.Product.SubmitOrder.Event.lease true) ∧\n (plan.preEntry = true →\n Auths.Product.SubmitOrder.Answered (Auths.Product.SubmitOrder.steps plan pre)\n Auths.Product.SubmitOrder.Phase.preEntry\n (Auths.Product.SubmitOrder.Event.preEntry Auths.Product.SubmitOrder.PreEntryResult.satisfied)) ∧\n Auths.Product.SubmitOrder.Answered (Auths.Product.SubmitOrder.steps plan pre)\n Auths.Product.SubmitOrder.Phase.entryReload (Auths.Product.SubmitOrder.Event.reload true) ∧\n Auths.Product.SubmitOrder.Answered (Auths.Product.SubmitOrder.steps plan pre)\n Auths.Product.SubmitOrder.Phase.deadline (Auths.Product.SubmitOrder.Event.deadline true)"
},
{
"axioms": [
@@ -1658,7 +1694,7 @@
"kind": "theorem",
"module": "Auths.Product.SubmitOrder",
"name": "Auths.Product.SubmitOrder.send_requires_credential_checks",
- "statement": "∀ (plan : Auths.Product.SubmitOrder.Plan) (pre : List Auths.Product.SubmitOrder.Event)\n (event : Auths.Product.SubmitOrder.Event),\n (Auths.Product.SubmitOrder.nextStep (Auths.Product.SubmitOrder.stateAfter plan pre) event).action =\n Auths.Product.SubmitOrder.Action.send →\n Auths.Product.SubmitOrder.Answered (Auths.Product.SubmitOrder.steps plan pre)\n (Auths.Product.SubmitOrder.Phase.prefixGuard Auths.Product.SubmitOrder.Mode.entry)\n (Auths.Product.SubmitOrder.Event.prefixGuard true) ∧\n (plan.accountRead = true →\n Auths.Product.SubmitOrder.Answered (Auths.Product.SubmitOrder.steps plan pre)\n (Auths.Product.SubmitOrder.Phase.account Auths.Product.SubmitOrder.Mode.entry)\n (Auths.Product.SubmitOrder.Event.account Auths.Product.SubmitOrder.AccountResult.equal)) ∧\n ∀ index < plan.deniedReads,\n Auths.Product.SubmitOrder.Answered (Auths.Product.SubmitOrder.steps plan pre)\n (Auths.Product.SubmitOrder.Phase.denied Auths.Product.SubmitOrder.Mode.entry index)\n (Auths.Product.SubmitOrder.Event.denied Auths.Product.SubmitOrder.DeniedResult.refused)"
+ "statement": "∀ (plan : Auths.Product.SubmitOrder.Plan) (pre : List Auths.Product.SubmitOrder.Event)\n (event : Auths.Product.SubmitOrder.Event),\n (Auths.Product.SubmitOrder.nextStep (Auths.Product.SubmitOrder.stateAfter plan pre) event).action =\n Auths.Product.SubmitOrder.Action.send →\n Auths.Product.SubmitOrder.Answered (Auths.Product.SubmitOrder.steps plan pre)\n Auths.Product.SubmitOrder.Phase.prefixGuard (Auths.Product.SubmitOrder.Event.prefixGuard true) ∧\n (plan.accountRead = true →\n Auths.Product.SubmitOrder.Answered (Auths.Product.SubmitOrder.steps plan pre)\n Auths.Product.SubmitOrder.Phase.account\n (Auths.Product.SubmitOrder.Event.account Auths.Product.SubmitOrder.AccountResult.equal)) ∧\n ∀ index < plan.deniedReads,\n Auths.Product.SubmitOrder.Answered (Auths.Product.SubmitOrder.steps plan pre)\n (Auths.Product.SubmitOrder.Phase.denied index)\n (Auths.Product.SubmitOrder.Event.denied Auths.Product.SubmitOrder.DeniedResult.refused)"
},
{
"axioms": [
@@ -1690,6 +1726,33 @@
"name": "Auths.Product.SubmitOrder.at_most_one_send",
"statement": "∀ (plan : Auths.Product.SubmitOrder.Plan) (events : List Auths.Product.SubmitOrder.Event),\n List.count Auths.Product.SubmitOrder.Action.send\n (List.map Auths.Product.SubmitOrder.Step.action (Auths.Product.SubmitOrder.steps plan events)) ≤\n 1 ∧\n (Auths.Product.SubmitOrder.Answered (Auths.Product.SubmitOrder.steps plan events)\n Auths.Product.SubmitOrder.Phase.claim\n (Auths.Product.SubmitOrder.Event.claim Auths.Product.SubmitOrder.ClaimResult.replay) →\n ¬Auths.Product.SubmitOrder.Emitted (Auths.Product.SubmitOrder.steps plan events)\n Auths.Product.SubmitOrder.Action.send)"
},
+ {
+ "axioms": [
+ "propext"
+ ],
+ "kind": "theorem",
+ "module": "Auths.Product.SubmitOrder",
+ "name": "Auths.Product.SubmitOrder.response_stops_without_read",
+ "statement": "∀ (state : Auths.Product.SubmitOrder.State),\n state.phase = Auths.Product.SubmitOrder.Phase.recordResponse →\n Auths.Product.SubmitOrder.nextStep state\n (Auths.Product.SubmitOrder.Event.response Auths.Product.SubmitOrder.ResponseRecord.recorded) =\n Auths.Product.SubmitOrder.stop state Auths.Product.SubmitOrder.Stop.response"
+ },
+ {
+ "axioms": [
+ "propext"
+ ],
+ "kind": "theorem",
+ "module": "Auths.Product.SubmitOrder",
+ "name": "Auths.Product.SubmitOrder.replay_returns_stored",
+ "statement": "∀ (state : Auths.Product.SubmitOrder.State),\n state.phase = Auths.Product.SubmitOrder.Phase.claim →\n Auths.Product.SubmitOrder.nextStep state\n (Auths.Product.SubmitOrder.Event.claim Auths.Product.SubmitOrder.ClaimResult.replay) =\n Auths.Product.SubmitOrder.stop state Auths.Product.SubmitOrder.Stop.stored"
+ },
+ {
+ "axioms": [
+ "propext"
+ ],
+ "kind": "theorem",
+ "module": "Auths.Product.SubmitOrder",
+ "name": "Auths.Product.SubmitOrder.refused_grant_stops_without_send",
+ "statement": "∀ (state : Auths.Product.SubmitOrder.State),\n state.phase = Auths.Product.SubmitOrder.Phase.grant →\n Auths.Product.SubmitOrder.nextStep state (Auths.Product.SubmitOrder.Event.granted false) =\n Auths.Product.SubmitOrder.stop state Auths.Product.SubmitOrder.Stop.grantUnavailable"
+ },
{
"axioms": [
"propext",
@@ -1718,7 +1781,7 @@
"kind": "theorem",
"module": "Auths.Product.SubmitOrder",
"name": "Auths.Product.SubmitOrder.post_claim_refusal_records_not_entered",
- "statement": "∀ (plan : Auths.Product.SubmitOrder.Plan) (pre : List Auths.Product.SubmitOrder.Event)\n (event : Auths.Product.SubmitOrder.Event) (refusal : Auths.Product.SubmitOrder.Refusal),\n ((Auths.Product.SubmitOrder.nextStep (Auths.Product.SubmitOrder.stateAfter plan pre) event).action =\n Auths.Product.SubmitOrder.Action.recordNotEntered refusal ↔\n Auths.Product.SubmitOrder.Answered (Auths.Product.SubmitOrder.steps plan pre)\n Auths.Product.SubmitOrder.Phase.claim\n (Auths.Product.SubmitOrder.Event.claim Auths.Product.SubmitOrder.ClaimResult.inserted) ∧\n Auths.Product.SubmitOrder.refusalOf (Auths.Product.SubmitOrder.stateAfter plan pre) event = some refusal) ∧\n ((Auths.Product.SubmitOrder.nextStep (Auths.Product.SubmitOrder.stateAfter plan pre) event).action =\n Auths.Product.SubmitOrder.Action.recordNotEntered refusal →\n (Auths.Product.SubmitOrder.nextStep (Auths.Product.SubmitOrder.stateAfter plan pre) event).state.phase =\n Auths.Product.SubmitOrder.Phase.recordNotEntered ∧\n (Auths.Product.SubmitOrder.Emitted (Auths.Product.SubmitOrder.steps plan pre)\n Auths.Product.SubmitOrder.Action.send →\n refusal = Auths.Product.SubmitOrder.Refusal.transportNotEntered))"
+ "statement": "∀ (plan : Auths.Product.SubmitOrder.Plan) (pre : List Auths.Product.SubmitOrder.Event)\n (event : Auths.Product.SubmitOrder.Event) (refusal : Auths.Product.SubmitOrder.Refusal),\n ((Auths.Product.SubmitOrder.nextStep (Auths.Product.SubmitOrder.stateAfter plan pre) event).action =\n Auths.Product.SubmitOrder.Action.recordNotEntered refusal ↔\n Auths.Product.SubmitOrder.Answered (Auths.Product.SubmitOrder.steps plan pre)\n Auths.Product.SubmitOrder.Phase.claim\n (Auths.Product.SubmitOrder.Event.claim Auths.Product.SubmitOrder.ClaimResult.reserved) ∧\n Auths.Product.SubmitOrder.refusalOf (Auths.Product.SubmitOrder.stateAfter plan pre) event = some refusal) ∧\n ((Auths.Product.SubmitOrder.nextStep (Auths.Product.SubmitOrder.stateAfter plan pre) event).action =\n Auths.Product.SubmitOrder.Action.recordNotEntered refusal →\n (Auths.Product.SubmitOrder.nextStep (Auths.Product.SubmitOrder.stateAfter plan pre) event).state.phase =\n Auths.Product.SubmitOrder.Phase.recordNotEntered ∧\n ¬Auths.Product.SubmitOrder.Emitted (Auths.Product.SubmitOrder.steps plan pre)\n Auths.Product.SubmitOrder.Action.send)"
},
{
"axioms": [
diff --git a/formal/proof-coverage-functions-v1.tsv b/formal/proof-coverage-functions-v1.tsv
index 9e3656230..b91ad42c2 100644
--- a/formal/proof-coverage-functions-v1.tsv
+++ b/formal/proof-coverage-functions-v1.tsv
@@ -1205,49 +1205,53 @@ product/runtime/auths-connections/src/credential.rs 145 147 3 StoredSecretLease:
product/runtime/auths-connections/src/credential.rs 182 187 6 CredentialStoreKind::parse product/runtime/auths-connections decision 0 0 0 0 0 0
product/runtime/auths-connections/src/credential.rs 191 196 6 CredentialStoreKind::as_str product/runtime/auths-connections decision 0 0 0 0 0 0
product/runtime/auths-connections/src/credential.rs 201 206 6 CredentialStoreKind::is_production product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 320 325 6 StoredSecret::holds product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 334 343 10 InMemoryCredentialStore::new product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 345 347 3 InMemoryCredentialStore::total_bytes product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 352 382 31 InMemoryCredentialStore::install product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 384 398 15 InMemoryCredentialStore::lease_secret product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 400 441 42 InMemoryCredentialStore::replace product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 443 455 13 InMemoryCredentialStore::revoke product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 457 463 7 InMemoryCredentialStore::holds product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 465 503 39 InMemoryCredentialStore::confirm product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 505 520 16 InMemoryCredentialStore::retire_superseded product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 522 534 13 InMemoryCredentialStore::delete_connection product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 560 566 7 PersistentCredentialStore::open product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 574 603 30 PersistentCredentialStore::open_with_limits product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 605 620 16 PersistentCredentialStore::mutate product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 622 624 3 PersistentCredentialStore::total_bytes product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 639 651 13 PersistentCredentialStore::retained_commitment product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 660 671 12 PersistentCredentialStore::stored_generations product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 687 701 15 PersistentCredentialStore::lease_for_record product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 709 718 10 PersistentCredentialStore::holds_record_credential product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 738 774 37 PersistentCredentialStore::store_confirmed product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 790 795 6 PersistentCredentialStore::revoke_connection product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 812 837 26 PersistentCredentialStore::retain_generations product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 839 860 22 PersistentCredentialStore::delete_generations product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 865 875 11 retained_entry product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 883 903 21 binding_entry product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 907 930 24 record_entry product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 932 940 9 connection_generations product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 944 972 29 PersistentCredentialStore::install product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 974 988 15 PersistentCredentialStore::lease_secret product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 990 1034 45 PersistentCredentialStore::replace product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 1036 1047 11 PersistentCredentialStore::revoke product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 1049 1055 7 PersistentCredentialStore::holds product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 1057 1069 13 PersistentCredentialStore::confirm product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 1071 1084 14 PersistentCredentialStore::retire_superseded product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 1086 1092 7 PersistentCredentialStore::delete_connection product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 1095 1108 14 validate_parent product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 1110 1125 16 validate_secret_file product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 1127 1158 32 persist_entries product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 1163 1174 12 encode_persistent_entries product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 1176 1198 23 write_persistent_entries product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 1206 1209 4 EncodedLength::write_all product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 1212 1306 95 decode_persistent_entries product/runtime/auths-connections decision 0 0 0 0 0 0
-product/runtime/auths-connections/src/credential.rs 1337 1348 12 credential_commitment product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 322 327 6 StoredSecret::holds product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 336 345 10 InMemoryCredentialStore::new product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 347 349 3 InMemoryCredentialStore::total_bytes product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 354 384 31 InMemoryCredentialStore::install product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 386 400 15 InMemoryCredentialStore::lease_secret product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 402 443 42 InMemoryCredentialStore::replace product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 445 457 13 InMemoryCredentialStore::revoke product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 459 465 7 InMemoryCredentialStore::holds product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 467 505 39 InMemoryCredentialStore::confirm product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 507 522 16 InMemoryCredentialStore::retire_superseded product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 524 536 13 InMemoryCredentialStore::delete_connection product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 565 571 7 PersistentCredentialStore::open product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 579 605 27 PersistentCredentialStore::open_with_limits product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 607 624 18 PersistentCredentialStore::mutate product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 629 673 45 PersistentCredentialStore::writer_lock product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 675 688 14 PersistentCredentialStore::refresh product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 690 698 9 PersistentCredentialStore::current_entries product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 700 702 3 PersistentCredentialStore::total_bytes product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 717 726 10 PersistentCredentialStore::retained_commitment product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 735 743 9 PersistentCredentialStore::stored_generations product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 759 770 12 PersistentCredentialStore::lease_for_record product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 778 784 7 PersistentCredentialStore::holds_record_credential product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 804 840 37 PersistentCredentialStore::store_confirmed product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 856 861 6 PersistentCredentialStore::revoke_connection product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 878 903 26 PersistentCredentialStore::retain_generations product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 905 926 22 PersistentCredentialStore::delete_generations product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 931 941 11 retained_entry product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 949 969 21 binding_entry product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 973 996 24 record_entry product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 998 1006 9 connection_generations product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 1010 1038 29 PersistentCredentialStore::install product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 1040 1051 12 PersistentCredentialStore::lease_secret product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 1053 1097 45 PersistentCredentialStore::replace product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 1099 1110 11 PersistentCredentialStore::revoke product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 1112 1115 4 PersistentCredentialStore::holds product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 1117 1129 13 PersistentCredentialStore::confirm product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 1131 1144 14 PersistentCredentialStore::retire_superseded product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 1146 1152 7 PersistentCredentialStore::delete_connection product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 1155 1168 14 validate_parent product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 1170 1185 16 validate_secret_file product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 1187 1216 29 read_persistent_entries product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 1218 1249 32 persist_entries product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 1254 1265 12 encode_persistent_entries product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 1267 1289 23 write_persistent_entries product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 1297 1300 4 EncodedLength::write_all product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 1303 1397 95 decode_persistent_entries product/runtime/auths-connections decision 0 0 0 0 0 0
+product/runtime/auths-connections/src/credential.rs 1428 1439 12 credential_commitment product/runtime/auths-connections decision 0 0 0 0 0 0
product/runtime/auths-connections/src/kernel.rs 29 31 3 next_generation product/runtime/auths-connections decision 0 0 1 1 1 1
product/runtime/auths-connections/src/kernel.rs 36 44 9 state_change product/runtime/auths-connections decision 0 0 1 1 1 1
product/runtime/auths-connections/src/kernel.rs 49 57 9 rotation product/runtime/auths-connections decision 0 0 1 1 1 1
@@ -1385,62 +1389,66 @@ product/runtime/auths-gateway-kernel/src/construct.rs 780 804 25 build_body prod
product/runtime/auths-gateway-kernel/src/construct.rs 812 833 22 construct_write product/runtime/auths-gateway-kernel decision 0 0 1 1 1 1
product/runtime/auths-gateway-kernel/src/construct.rs 840 852 13 construct_action_read product/runtime/auths-gateway-kernel decision 0 0 1 1 1 1
product/runtime/auths-gateway-kernel/src/construct.rs 858 865 8 construct_credential_read product/runtime/auths-gateway-kernel decision 0 0 1 1 1 1
-product/runtime/auths-gateway-kernel/src/order.rs 387 393 7 start product/runtime/auths-gateway-kernel decision 0 0 1 1 1 1
-product/runtime/auths-gateway-kernel/src/order.rs 397 406 10 go product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/order.rs 410 412 3 stop product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/order.rs 416 418 3 halt product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/order.rs 422 428 7 not_entered product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/order.rs 433 439 7 lease_failed product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/order.rs 445 451 7 after_reads product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/order.rs 455 461 7 ceiling_from product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/order.rs 465 471 7 pre_entry_from product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/order.rs 476 483 8 after_credentials product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/order.rs 487 497 11 denied_from product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/order.rs 501 507 7 after_prefix product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/order.rs 511 517 7 after_admission product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/order.rs 521 527 7 pre_claim product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/order.rs 531 542 12 after_claim product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/order.rs 546 556 11 after_account product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/order.rs 560 577 18 after_denied product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/order.rs 581 587 7 after_pre_entry product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/order.rs 592 605 14 after_ceiling product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/order.rs 609 615 7 after_write product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/order.rs 619 634 16 after_response product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/order.rs 638 648 11 after_read_back product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/order.rs 653 705 53 pre_claim_step product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/order.rs 710 766 57 lease_step product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/order.rs 772 840 69 entry_step product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/order.rs 846 873 28 next_step product/runtime/auths-gateway-kernel decision 0 0 1 1 1 1
-product/runtime/auths-gateway-kernel/src/outcome.rs 48 57 10 outcome_stage product/runtime/auths-gateway-kernel decision 0 0 0 0 0 0
-product/runtime/auths-gateway-kernel/src/outcome.rs 67 76 10 outcome_facts_present product/runtime/auths-gateway-kernel decision 0 0 0 0 0 0
+product/runtime/auths-gateway-kernel/src/order.rs 367 373 7 start product/runtime/auths-gateway-kernel decision 0 0 1 1 1 1
+product/runtime/auths-gateway-kernel/src/order.rs 377 386 10 go product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/order.rs 390 392 3 stop product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/order.rs 396 398 3 halt product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/order.rs 402 408 7 not_entered product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/order.rs 414 420 7 after_reads product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/order.rs 424 430 7 ceiling_from product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/order.rs 434 440 7 pre_entry_from product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/order.rs 444 450 7 denied_from product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/order.rs 454 460 7 after_prefix product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/order.rs 464 470 7 after_admission product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/order.rs 474 480 7 pre_claim product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/order.rs 484 491 8 after_claim product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/order.rs 495 501 7 after_account product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/order.rs 505 517 13 after_denied product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/order.rs 521 527 7 after_pre_entry product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/order.rs 532 545 14 after_ceiling product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/order.rs 549 554 6 after_write product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/order.rs 558 563 6 after_response product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/order.rs 568 620 53 pre_claim_step product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/order.rs 625 667 43 lease_step product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/order.rs 673 742 70 entry_step product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/order.rs 748 772 25 next_step product/runtime/auths-gateway-kernel decision 0 0 1 1 1 1
+product/runtime/auths-gateway-kernel/src/outcome.rs 48 58 11 outcome_stage product/runtime/auths-gateway-kernel decision 0 0 0 0 0 0
+product/runtime/auths-gateway-kernel/src/outcome.rs 68 77 10 outcome_facts_present product/runtime/auths-gateway-kernel decision 0 0 0 0 0 0
product/runtime/auths-gateway-kernel/src/ratio.rs 16 27 12 relative_basis product/runtime/auths-gateway-kernel decision 0 0 1 1 1 1
product/runtime/auths-gateway-kernel/src/ratio.rs 35 39 5 relative_ceiling_admits product/runtime/auths-gateway-kernel decision 0 0 1 1 1 1
product/runtime/auths-gateway-kernel/src/recovery.rs 98 104 7 provider_link product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
product/runtime/auths-gateway-kernel/src/recovery.rs 108 122 15 recovery_class product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
product/runtime/auths-gateway-kernel/src/recovery.rs 126 133 8 unknown_resolution product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
product/runtime/auths-gateway-kernel/src/recovery.rs 139 151 13 recovery_capability product/runtime/auths-gateway-kernel decision 0 0 1 1 1 1
-product/runtime/auths-gateway-kernel/src/transition.rs 89 91 3 absent product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/transition.rs 95 104 10 same_bytes_from product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/transition.rs 108 114 7 same_bytes product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/transition.rs 119 124 6 resolves_unknown product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/transition.rs 128 133 6 has_link product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/transition.rs 137 146 10 refusal_stage product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/transition.rs 150 155 6 requires_response product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/transition.rs 159 164 6 permits_response product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/transition.rs 168 177 10 observed_stage product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/transition.rs 181 190 10 provider_stage product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/transition.rs 194 199 6 terminal_stage product/runtime/auths-gateway-kernel decision 0 0 0 0 0 0
-product/runtime/auths-gateway-kernel/src/transition.rs 203 208 6 has_reading product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/transition.rs 217 228 12 consistent product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/transition.rs 235 260 26 stage_transition_allowed product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/transition.rs 265 280 16 adds_pre_entry product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/transition.rs 284 300 17 adds_response product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/transition.rs 305 321 17 checkpoint_valid product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/transition.rs 326 332 7 pre_entry_kept product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/transition.rs 337 340 4 response_kept product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/transition.rs 344 346 3 plan_kept product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/transition.rs 350 365 16 link_equal product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
-product/runtime/auths-gateway-kernel/src/transition.rs 373 381 9 valid_transition product/runtime/auths-gateway-kernel decision 0 0 1 1 1 1
+product/runtime/auths-gateway-kernel/src/transition.rs 113 115 3 absent product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/transition.rs 119 128 10 same_bytes_from product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/transition.rs 132 138 7 same_bytes product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/transition.rs 143 148 6 resolves_unknown product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/transition.rs 152 157 6 has_link product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/transition.rs 161 172 12 refusal_stage product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/transition.rs 176 186 11 requires_response product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/transition.rs 190 199 10 permits_response product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/transition.rs 203 214 12 observed_stage product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/transition.rs 218 229 12 provider_stage product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/transition.rs 233 240 8 terminal_stage product/runtime/auths-gateway-kernel decision 0 0 0 0 0 0
+product/runtime/auths-gateway-kernel/src/transition.rs 244 249 6 has_reading product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/transition.rs 258 274 17 consistent product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/transition.rs 278 280 3 acknowledgement_stage product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/transition.rs 285 321 37 protocol_consistent product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/transition.rs 325 332 8 same_protocol product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/transition.rs 336 346 11 requires_entry product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/transition.rs 352 373 22 stage_transition_allowed product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/transition.rs 377 379 3 adds_pre_entry product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/transition.rs 383 385 3 adds_response product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/transition.rs 389 395 7 checkpoint_valid product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/transition.rs 399 405 7 entry_kept product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/transition.rs 410 416 7 pre_entry_kept product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/transition.rs 421 424 4 response_kept product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/transition.rs 429 436 8 acknowledgement_kept product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/transition.rs 444 449 6 exact_acknowledgement_replay product/runtime/auths-gateway-kernel decision 0 0 0 0 0 0
+product/runtime/auths-gateway-kernel/src/transition.rs 453 455 3 plan_kept product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/transition.rs 459 474 16 link_equal product/runtime/auths-gateway-kernel decision 0 0 1 0 1 0
+product/runtime/auths-gateway-kernel/src/transition.rs 482 496 15 valid_transition product/runtime/auths-gateway-kernel decision 0 0 1 1 1 1
product/runtime/auths-lifecycle/src/codec.rs 53 63 11 encode_record product/runtime/auths-lifecycle decoding 0 0 0 0 0 0
product/runtime/auths-lifecycle/src/codec.rs 71 83 13 decode_record product/runtime/auths-lifecycle decoding 0 0 0 0 0 0
product/runtime/auths-lifecycle/src/codec.rs 85 95 11 encode_unchecked product/runtime/auths-lifecycle decoding 0 0 0 0 0 0
diff --git a/formal/proof-coverage-v1.json b/formal/proof-coverage-v1.json
index 49af70735..6c6ddb669 100644
--- a/formal/proof-coverage-v1.json
+++ b/formal/proof-coverage-v1.json
@@ -16,9 +16,9 @@
"translation_toolchain_lock_sha256": "0c45a8a08ac06e313d775e669749097bd3a0abfb8e3395c69416ba0238770d53"
},
"scope_definition_sha256": "053f0be216b0003768d422d9ccc92a692726488874284fa80615657b30a495b5",
- "inputs_sha256": "56f5f8df74c31a2554725ec967451d36bfa6e7c4a49004b3c83bb5cd32c35f52",
+ "inputs_sha256": "ac052be07bf5d42e8b136ee8c1e0cf94f0cc3253f24c787165acc5b8607e3781",
"inputs": {
- "Cargo.toml": "sha256:6e0240aa432bcdb5b4f855618994582d2e428afe5ee4e4f6a3f8d21a518f8d24",
+ "Cargo.toml": "sha256:ccc17deac72da945940a9ab4be2ce27628c056b977d19429bc1679b19fae30f3",
"core/crates/auths-algebra-kernel/Cargo.toml": "sha256:0cb078915287b8c47d20fb97b6fa14135ed1c61c4985c4de2753855563c27bba",
"core/crates/auths-algebra-kernel/src/generated.rs": "sha256:ae4bcbef42a408c7ffa0c6a74a42f7de177a6f14479e51d205acf830bda05960",
"core/crates/auths-algebra-kernel/src/lib.rs": "sha256:6654667eeb9b353d6648954c9fd2e8838462a82ce6b2803eb49e1366892c68c6",
@@ -59,7 +59,7 @@
"core/crates/auths-verifier/src/trace.rs": "sha256:2213d72023a1d9e0413a2cf2c51c3dad9e04775d86033a0caaa9d29a8f0036f8",
"formal/check-site-functions-v1.toml": "sha256:a4eebcb806f2be41aa39fb18700c70170cfe5792162feca0501d6e3fb69325ea",
"formal/coverage-scope-v1.toml": "sha256:053f0be216b0003768d422d9ccc92a692726488874284fa80615657b30a495b5",
- "formal/lean-assurance-audit-v1.json": "sha256:adbc3882efd275e821ae5fb5da68107d44915cb9205dbdc57681d4493e8b85ee",
+ "formal/lean-assurance-audit-v1.json": "sha256:9f975367344d2d1d9e84ba115a9115540b822e84218bab0eaed90024bd3a64b9",
"formal/qualification/aeneas/generated/algebra/Funs.lean": "sha256:f78659289b73410828dceb4557cd4fbf2fa801b06ca083e8268be33074f68960",
"formal/qualification/aeneas/generated/algebra/translation.json": "sha256:7b486aa626850f509f6c1038dbc5c74c753d53707d0085260e30321e570fab9d",
"formal/qualification/aeneas/generated/authority/Funs.lean": "sha256:2bb401ffca0136622cd79bb14a5a38852061649a06cc46100503870c22d300c7",
@@ -68,13 +68,13 @@
"formal/qualification/aeneas/generated/bounded_policy/translation.json": "sha256:3be48bd557146aef4d7452b17af0793fd61ed1e390c2274d7c75908a80fd79f5",
"formal/qualification/aeneas/generated/connections/Funs.lean": "sha256:dfa1998e3506f06dce142222dc383116d0952155517edc399f5d94e688e22912",
"formal/qualification/aeneas/generated/connections/translation.json": "sha256:c43f8a044ad0c19c676d43e3a6f7e24a492864c94f50b8fa4afbd08dd01f82ba",
- "formal/qualification/aeneas/generated/gateway/Funs.lean": "sha256:c8c86db058f313ab46ee86370d79ee70130c1ead3860dd7dabd49f2a89af1636",
- "formal/qualification/aeneas/generated/gateway/translation.json": "sha256:4b9e3e1158b6751a458ae6581e7e2230a1b242913d9a4842710d3bea3e08ee45",
+ "formal/qualification/aeneas/generated/gateway/Funs.lean": "sha256:8c6c2ab1a63280e94f4084566888d727b13a5890b9de4c75aa90322406c5c601",
+ "formal/qualification/aeneas/generated/gateway/translation.json": "sha256:8b6b36d15312382cbed785d6a196056bad9da515e0c21a9b28ca5eee495a1ade",
"formal/qualification/aeneas/generated/lifecycle/Funs.lean": "sha256:817dcbb854a599fb14c89854d8023809a390f7a00e1cc70dbc7151d7b254649f",
"formal/qualification/aeneas/generated/lifecycle/translation.json": "sha256:b729089853215acf4d019aae5f04da6d902b75ac4910feeeb8d801e9991cdb4a",
"formal/qualification/aeneas/generated/model/Funs.lean": "sha256:d4d36b72718a4076b7c0561abfa2edc006a1f9a4491762979f86c3dd99a71fdc",
"formal/qualification/aeneas/generated/model/translation.json": "sha256:66e62345ea87b3150722b2d8b37473269c1cbfdaa09c875a3546ae381b7f5cd0",
- "formal/qualification/aeneas/qualification.toml": "sha256:38aa0babb5be3ac958c7e91f0640ad0ae3b6e4f265257642976dfe8bd3330850",
+ "formal/qualification/aeneas/qualification.toml": "sha256:1da6c96a8d64598ff8a6999d6834e4f78fc0dee69d8c56ab3f69c37ecf436d58",
"formal/theorem-kinds-v1.toml": "sha256:9714b276af0a45cadb7582e39beaa57dcb3494d6af23ca1b6a8702cc78e59728",
"product/policy/auths-bounded-policy/src/arithmetic.rs": "sha256:75e43c5d4eab8d1f71c1daf53839f5f9fe45a400afc1f63a30c6f55ce3632627",
"product/policy/auths-bounded-policy/src/commitment.rs": "sha256:c5706f2d583454e1a7f570f7ada138d4d864bef0205dfae7c1296774f396bb3f",
@@ -84,18 +84,18 @@
"product/policy/auths-bounded-policy/src/lib.rs": "sha256:76a18638b810e62e594d6e951504679305ce21841388e0c282ad51cd7841b91c",
"product/policy/auths-bounded-policy/src/receipt.rs": "sha256:3b9eae82f349956d3e7f29b57b254bff5ce67cc703e6cf43445ce2827676d1ad",
"product/policy/auths-bounded-policy/src/registry.rs": "sha256:4e251ea5833559841f8ee0bd9b66c0c01c39fa5e3a21720a19c4e8adbfcb99d0",
- "product/runtime/auths-connections/src/credential.rs": "sha256:82921413bb053f0f6c7931fbd4e6a86c8cc1d8c6b51debd9c2cef33a8d5b1c51",
+ "product/runtime/auths-connections/src/credential.rs": "sha256:7e6fbca8911d5c0c1baddf35572236d419566dd6e3fcbbdfda77c8dd114dbcad",
"product/runtime/auths-connections/src/kernel.rs": "sha256:a456af19526ef7b19278551fd8d97606b6e7aa5a195e70bdd85d1e0786090ce8",
"product/runtime/auths-connections/src/lib.rs": "sha256:79209249a0acba0e854d9937255ea0e30a9a6f8e1c2baad77ab83646d392de22",
"product/runtime/auths-connections/src/model.rs": "sha256:de34a089cac9702acb889b7602b3e4732306c9e61029ec892582b7e02b73fc77",
"product/runtime/auths-connections/src/registry.rs": "sha256:e57640e003e72ad3235c567c2bb944a262bd64e673f023b62ea091094cc7e946",
"product/runtime/auths-gateway-kernel/src/construct.rs": "sha256:e9d65744bae273d65face4285e5d2da75c06a92ab48f7b8333d496604e31c63c",
"product/runtime/auths-gateway-kernel/src/lib.rs": "sha256:cfda6f8453426564c085475f183b6f790b438768a19668f1b65d33ba59dc8632",
- "product/runtime/auths-gateway-kernel/src/order.rs": "sha256:e0045f877c4b7fe3860c291039853885064ddcdebba7c9b345ab65e81e646dda",
- "product/runtime/auths-gateway-kernel/src/outcome.rs": "sha256:e9785b1afe70788c89b51b18655bcd5acff7b0c8e3d12fb79c6f05a202e67e6b",
+ "product/runtime/auths-gateway-kernel/src/order.rs": "sha256:efc2f2eef68232fe3082e1607e6f1e87d4a19cbb2f492bdc6e71916a2461b6a8",
+ "product/runtime/auths-gateway-kernel/src/outcome.rs": "sha256:ac0245699457c802a70cdd6f52a01dcb4777c487b58ed7dc9cfc99345465ccaa",
"product/runtime/auths-gateway-kernel/src/ratio.rs": "sha256:7ace22264812e6fc1ef3a62afee402cf43494bb71ae0d3de219ad1b39d690bcb",
"product/runtime/auths-gateway-kernel/src/recovery.rs": "sha256:9f46dcd1685ab6e4d1ba04a3fc286a9f3afee8fb47829be1fbc670a7926a2dc6",
- "product/runtime/auths-gateway-kernel/src/transition.rs": "sha256:cefb8f01d915880f6921f2501ff881ad94b9de61072c34d9cd19220992f97374",
+ "product/runtime/auths-gateway-kernel/src/transition.rs": "sha256:9c9c292397aa640c2f002ffb5befe110f86d0aa246f3ffe99440718e5174af3c",
"product/runtime/auths-lifecycle/src/codec.rs": "sha256:51c4b746749150e8bfcd8ff972b3f6c1ef40bec15583448ec1b471d9ad737dc0",
"product/runtime/auths-lifecycle/src/digest.rs": "sha256:55282cd7d8aba37713c2dcb5357c9877819edb72a30eadefb860d28d7d3a3eb7",
"product/runtime/auths-lifecycle/src/identifier.rs": "sha256:31620e8c16de68d3d2d3b84d6353c1a5b9d4151d8247f91fa389b906a8e660f1",
@@ -245,33 +245,33 @@
"translated-crates": {
"scope": "translated-crates",
"minimum_code_lines": 0,
- "functions": 1138,
- "span_lines": 10681,
- "code_lines": 10662,
+ "functions": 1146,
+ "span_lines": 10758,
+ "code_lines": 10738,
"by_level": {
"translated": {
- "functions": 210,
- "function_share": 0.1845,
- "span_lines": 2519,
- "span_line_share": 0.2358,
- "code_lines": 2508,
- "code_line_share": 0.2352
+ "functions": 213,
+ "function_share": 0.1859,
+ "span_lines": 2508,
+ "span_line_share": 0.2331,
+ "code_lines": 2497,
+ "code_line_share": 0.2325
},
"direct": {
"functions": 70,
- "function_share": 0.0615,
- "span_lines": 1035,
- "span_line_share": 0.0969,
- "code_lines": 1024,
- "code_line_share": 0.096
+ "function_share": 0.0611,
+ "span_lines": 1038,
+ "span_line_share": 0.0965,
+ "code_lines": 1027,
+ "code_line_share": 0.0956
},
"closure": {
- "functions": 210,
- "function_share": 0.1845,
- "span_lines": 2519,
- "span_line_share": 0.2358,
- "code_lines": 2508,
- "code_line_share": 0.2352
+ "functions": 213,
+ "function_share": 0.1859,
+ "span_lines": 2508,
+ "span_line_share": 0.2331,
+ "code_lines": 2497,
+ "code_line_share": 0.2325
}
}
},
@@ -548,18 +548,18 @@
"refined_closure_code_lines": 182
},
"product/runtime/auths-connections": {
- "decision_functions": 153,
- "decision_code_lines": 1887,
+ "decision_functions": 157,
+ "decision_code_lines": 1965,
"translated": 5,
"refined_closure": 5,
"refined_closure_code_lines": 51
},
"product/runtime/auths-gateway-kernel": {
- "decision_functions": 98,
- "decision_code_lines": 1194,
- "translated": 95,
- "refined_closure": 95,
- "refined_closure_code_lines": 1168
+ "decision_functions": 102,
+ "decision_code_lines": 1192,
+ "translated": 98,
+ "refined_closure": 98,
+ "refined_closure_code_lines": 1157
},
"product/runtime/auths-lifecycle": {
"decision_functions": 138,
@@ -2986,19 +2986,19 @@
}
},
"translation_inventory": {
- "aeneas_local_definitions": 318,
+ "aeneas_local_definitions": 321,
"by_kind": {
- "function": 210,
+ "function": 213,
"loop": 92,
"derived": 16
},
"refined_direct": 70,
- "refined_closure": 215,
+ "refined_closure": 218,
"theorems_naming_translated_functions": 75
},
"mutation_adequacy": {
"results": "formal/model-mutations-v1.json",
- "current_formal_inputs_sha256": "f250c011aa84a978f1f65e91ead3ed8aee7e3b6502e7d80868a583381df305db",
+ "current_formal_inputs_sha256": "9f655722b5ba3e4322f25ffd60c37fa9149373099562a28a8faa08f3a9848169",
"fresh": false,
"campaigns": {
"s1": {
diff --git a/formal/qualification/aeneas/cases/Gateway.lean b/formal/qualification/aeneas/cases/Gateway.lean
index e197bf179..2e70a0ccd 100644
--- a/formal/qualification/aeneas/cases/Gateway.lean
+++ b/formal/qualification/aeneas/cases/Gateway.lean
@@ -1,4 +1,5 @@
import qualification.aeneas.generated.gateway.Funs
+import Auths.Product.Refinement.Gateway
open Aeneas Aeneas.Std Result
@@ -94,4 +95,90 @@ example :
action := order.SubmitAction.Stop order.Stop.Halted } := by
rfl
+-- AP074: the deadline names durable grant creation; only its successful
+-- result names the send. Failure stops and replay returns stored state.
+example :
+ order.next_step { plan, phase := .Deadline, argument := 1#u64 } (.Deadline true) =
+ ok { state := { plan, phase := .Grant, argument := 1#u64 }, action := .GrantEntry } := by
+ rfl
+example :
+ order.next_step { plan, phase := .Grant, argument := 1#u64 } (.Granted true) =
+ ok { state := { plan, phase := .Send, argument := 1#u64 }, action := .Send } := by
+ rfl
+example :
+ order.next_step { plan, phase := .Grant, argument := 1#u64 } (.Granted false) =
+ ok { state := { plan, phase := .Done, argument := 1#u64 },
+ action := .Stop .GrantUnavailable } := by
+ rfl
+example :
+ order.next_step { plan, phase := .Claim, argument := 1#u64 } (.Claim .Replay) =
+ ok { state := { plan, phase := .Done, argument := 1#u64 }, action := .Stop .Stored } := by
+ rfl
+example :
+ order.next_step { plan, phase := .RecordResponse, argument := 1#u64 } (.Response .Recorded) =
+ ok { state := { plan, phase := .Done, argument := 1#u64 }, action := .Stop .Response } := by
+ rfl
+
+-- Late source response and recipient ACK are allowed evidence routes, but
+-- an unknown cannot gain a second grant or become refundable not-entered.
+example : transition.stage_transition_allowed .Unknown .ResponseRecorded .None = ok true := by
+ rfl
+example : transition.stage_transition_allowed .Unknown .Acknowledged .None = ok true := by
+ rfl
+example : transition.stage_transition_allowed .Unknown .EntryGranted .Verified = ok false := by
+ rfl
+example : transition.stage_transition_allowed .Unknown .NotEntered .Verified = ok false := by
+ rfl
+example : transition.stage_transition_allowed .Acknowledged .Unknown .None = ok false := by
+ rfl
+
+-- Full projected records demonstrate inhabited evidence routes and reject
+-- changed source grants, protocol substitution, and non-successor revisions.
+private def byteVec (byte : Std.U8) : alloc.vec.Vec Std.U8 :=
+ ⟨[byte], by simp; scalar_tac⟩
+
+private def unknownView (protocol : transition.Protocol) : transition.AttemptView :=
+ { protocol, revision := 2#u64, fixed := byteVec 1#u8, entry := byteVec 2#u8,
+ stage := .Unknown, refusal := false, pre_entry := ⟨[], by simp⟩,
+ response := ⟨[], by simp⟩, acknowledgement := ⟨[], by simp⟩,
+ locator := ⟨[], by simp⟩, reading := .None, evidence := false,
+ observation := false, link := .None }
+
+private def lateResponse : transition.AttemptView :=
+ { unknownView .Mutation with revision := 3#u64, stage := .ResponseRecorded, response := byteVec 3#u8 }
+
+private def lateAcknowledgement : transition.AttemptView :=
+ { unknownView .Disclosure with revision := 3#u64, stage := .Acknowledged, acknowledgement := byteVec 4#u8 }
+
+private theorem projected_transition (old new : transition.AttemptView) :
+ transition.valid_transition old new =
+ ok (Auths.Product.Recovery.validTransition
+ (Auths.Product.Refinement.Gateway.viewOf old)
+ (Auths.Product.Refinement.Gateway.viewOf new)) := by
+ obtain ⟨result, resultEq, refined⟩ := Aeneas.Std.WP.spec_imp_exists
+ (Auths.Product.Refinement.Gateway.translated_valid_transition_refines_model old new)
+ rw [resultEq, refined]
+
+example : transition.valid_transition (unknownView .Mutation) lateResponse = ok true := by
+ rw [projected_transition]
+ rfl
+example : transition.valid_transition (unknownView .Disclosure) lateAcknowledgement = ok true := by
+ rw [projected_transition]
+ rfl
+example : transition.valid_transition (unknownView .Mutation)
+ { lateResponse with entry := byteVec 5#u8 } = ok false := by
+ rw [projected_transition]
+ rfl
+example : transition.valid_transition (unknownView .Mutation)
+ { lateResponse with revision := 4#u64 } = ok false := by
+ rw [projected_transition]
+ rfl
+example : transition.valid_transition (unknownView .Mutation)
+ { lateAcknowledgement with protocol := .Mutation } = ok false := by
+ rw [projected_transition]
+ rfl
+example : transition.valid_transition (unknownView .Mutation) lateAcknowledgement = ok false := by
+ rw [projected_transition]
+ rfl
+
end qualification.aeneas.cases
diff --git a/formal/qualification/aeneas/generated/gateway/Funs.lean b/formal/qualification/aeneas/generated/gateway/Funs.lean
index 7cd1ea322..af1c9c6c6 100644
--- a/formal/qualification/aeneas/generated/gateway/Funs.lean
+++ b/formal/qualification/aeneas/generated/gateway/Funs.lean
@@ -1287,13 +1287,13 @@ def construct.construct_credential_read
}
/-- [auths_gateway_kernel::order::start]:
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 387:0-393:1
+ Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 367:0-373:1
Visibility: public -/
def order.start (plan : order.SubmitPlan) : Result order.SubmitState := do
ok { plan, phase := order.Phase.Start, argument := 0#u64 }
/-- [auths_gateway_kernel::order::go]:
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 397:0-406:1
+ Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 377:0-386:1
Visibility: public -/
def order.go
(state : order.SubmitState) (phase : order.Phase)
@@ -1303,7 +1303,7 @@ def order.go
ok { state := { state with phase }, action }
/-- [auths_gateway_kernel::order::stop]:
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 410:0-412:1
+ Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 390:0-392:1
Visibility: public -/
def order.stop
(state : order.SubmitState) (stop : order.Stop) :
@@ -1312,13 +1312,13 @@ def order.stop
order.go state order.Phase.Done (order.SubmitAction.Stop stop)
/-- [auths_gateway_kernel::order::halt]:
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 416:0-418:1
+ Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 396:0-398:1
Visibility: public -/
def order.halt (state : order.SubmitState) : Result order.SubmitDecision := do
order.stop state order.Stop.Halted
/-- [auths_gateway_kernel::order::not_entered]:
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 422:0-428:1
+ Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 402:0-408:1
Visibility: public -/
def order.not_entered
(state : order.SubmitState) (refusal : order.Refusal) :
@@ -1327,20 +1327,8 @@ def order.not_entered
order.go state order.Phase.RecordNotEntered
(order.SubmitAction.RecordNotEntered refusal)
-/-- [auths_gateway_kernel::order::lease_failed]:
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 433:0-439:1
- Visibility: public -/
-def order.lease_failed
- (state : order.SubmitState) (mode : order.Mode) (refusal : order.Refusal) :
- Result order.SubmitDecision
- := do
- match mode with
- | order.Mode.Entry => order.not_entered state refusal
- | order.Mode.ReadBack => order.stop state order.Stop.Response
- | order.Mode.Reobserve => order.stop state order.Stop.ReplayRefused
-
/-- [auths_gateway_kernel::order::after_reads]:
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 445:0-451:1
+ Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 414:0-420:1
Visibility: public -/
def order.after_reads
(state : order.SubmitState) : Result order.SubmitDecision := do
@@ -1356,7 +1344,7 @@ def order.after_reads
order.SubmitAction.ReloadBeforeEntry
/-- [auths_gateway_kernel::order::ceiling_from]:
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 455:0-461:1
+ Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 424:0-430:1
Visibility: public -/
def order.ceiling_from
(state : order.SubmitState) : Result order.SubmitDecision := do
@@ -1365,7 +1353,7 @@ def order.ceiling_from
else order.after_reads state
/-- [auths_gateway_kernel::order::pre_entry_from]:
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 465:0-471:1
+ Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 434:0-440:1
Visibility: public -/
def order.pre_entry_from
(state : order.SubmitState) : Result order.SubmitDecision := do
@@ -1373,50 +1361,30 @@ def order.pre_entry_from
then order.go state order.Phase.PreEntry order.SubmitAction.PreEntryRead
else order.ceiling_from state
-/-- [auths_gateway_kernel::order::after_credentials]:
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 476:0-483:1
- Visibility: public -/
-def order.after_credentials
- (state : order.SubmitState) (mode : order.Mode) :
- Result order.SubmitDecision
- := do
- match mode with
- | order.Mode.Entry => order.pre_entry_from state
- | order.Mode.ReadBack =>
- order.go state (order.Phase.ReadBack order.Mode.ReadBack)
- (order.SubmitAction.ReadBack order.Mode.ReadBack)
- | order.Mode.Reobserve =>
- order.go state (order.Phase.ReadBack order.Mode.Reobserve)
- (order.SubmitAction.ReadBack order.Mode.Reobserve)
-
/-- [auths_gateway_kernel::order::denied_from]:
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 487:0-497:1
+ Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 444:0-450:1
Visibility: public -/
def order.denied_from
- (state : order.SubmitState) (mode : order.Mode) (index : Std.U8) :
+ (state : order.SubmitState) (index : Std.U8) :
Result order.SubmitDecision
:= do
if index < state.plan.denied_reads
then
- order.go state (order.Phase.Denied mode index)
- (order.SubmitAction.DeniedRead mode index)
- else order.after_credentials state mode
+ order.go state (order.Phase.Denied index) (order.SubmitAction.DeniedRead
+ index)
+ else order.pre_entry_from state
/-- [auths_gateway_kernel::order::after_prefix]:
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 501:0-507:1
+ Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 454:0-460:1
Visibility: public -/
def order.after_prefix
- (state : order.SubmitState) (mode : order.Mode) :
- Result order.SubmitDecision
- := do
+ (state : order.SubmitState) : Result order.SubmitDecision := do
if state.plan.account_read
- then
- order.go state (order.Phase.Account mode) (order.SubmitAction.ReadAccount
- mode)
- else order.denied_from state mode 0#u8
+ then order.go state order.Phase.Account order.SubmitAction.ReadAccount
+ else order.denied_from state 0#u8
/-- [auths_gateway_kernel::order::after_admission]:
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 511:0-517:1
+ Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 464:0-470:1
Visibility: public -/
def order.after_admission
(state : order.SubmitState) : Result order.SubmitDecision := do
@@ -1425,7 +1393,7 @@ def order.after_admission
else order.go state order.Phase.Prepare order.SubmitAction.Prepare
/-- [auths_gateway_kernel::order::pre_claim]:
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 521:0-527:1
+ Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 474:0-480:1
Visibility: public -/
def order.pre_claim
(state : order.SubmitState) (passed : Bool) (next : order.SubmitDecision) :
@@ -1436,58 +1404,54 @@ def order.pre_claim
else order.stop state order.Stop.Refused
/-- [auths_gateway_kernel::order::after_claim]:
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 531:0-542:1
+ Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 484:0-491:1
Visibility: public -/
def order.after_claim
(state : order.SubmitState) (result : order.ClaimResult) :
Result order.SubmitDecision
:= do
match result with
- | order.ClaimResult.Inserted =>
- order.go state (order.Phase.Reload order.Mode.Entry)
- (order.SubmitAction.Reload order.Mode.Entry)
+ | order.ClaimResult.Reserved =>
+ order.go state order.Phase.Reload order.SubmitAction.Reload
| order.ClaimResult.Refused => order.stop state order.Stop.ClaimRefused
- | order.ClaimResult.Replay =>
- order.go state order.Phase.Resume order.SubmitAction.Resume
+ | order.ClaimResult.Replay => order.stop state order.Stop.Stored
| order.ClaimResult.Unavailable =>
order.stop state order.Stop.StoreUnavailable
/-- [auths_gateway_kernel::order::after_account]:
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 546:0-556:1
+ Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 495:0-501:1
Visibility: public -/
def order.after_account
- (state : order.SubmitState) (mode : order.Mode)
- (result : order.AccountResult) :
+ (state : order.SubmitState) (result : order.AccountResult) :
Result order.SubmitDecision
:= do
match result with
- | order.AccountResult.Equal => order.denied_from state mode 0#u8
+ | order.AccountResult.Equal => order.denied_from state 0#u8
| order.AccountResult.Mismatch =>
- order.lease_failed state mode order.Refusal.AccountMismatch
+ order.not_entered state order.Refusal.AccountMismatch
| order.AccountResult.Unavailable =>
- order.lease_failed state mode order.Refusal.AccountUnavailable
+ order.not_entered state order.Refusal.AccountUnavailable
/-- [auths_gateway_kernel::order::after_denied]:
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 560:0-577:1
+ Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 505:0-517:1
Visibility: public -/
def order.after_denied
- (state : order.SubmitState) (mode : order.Mode) (index : Std.U8)
- (result : order.DeniedResult) :
+ (state : order.SubmitState) (index : Std.U8) (result : order.DeniedResult) :
Result order.SubmitDecision
:= do
match result with
| order.DeniedResult.Refused =>
if index < state.plan.denied_reads
then let i ← index + 1#u8
- order.denied_from state mode i
+ order.denied_from state i
else order.halt state
| order.DeniedResult.Answered =>
- order.lease_failed state mode order.Refusal.CapabilityExcess
+ order.not_entered state order.Refusal.CapabilityExcess
| order.DeniedResult.Unavailable =>
- order.lease_failed state mode order.Refusal.CapabilityUnavailable
+ order.not_entered state order.Refusal.CapabilityUnavailable
/-- [auths_gateway_kernel::order::after_pre_entry]:
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 581:0-587:1
+ Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 521:0-527:1
Visibility: public -/
def order.after_pre_entry
(state : order.SubmitState) (result : order.PreEntryResult) :
@@ -1521,7 +1485,7 @@ def ratio.relative_ceiling_admits
ok (scaled_argument <= scaled_basis)
/-- [auths_gateway_kernel::order::after_ceiling]:
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 592:0-605:1
+ Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 532:0-545:1
Visibility: public -/
def order.after_ceiling
(state : order.SubmitState) (read : order.CeilingRead) :
@@ -1542,22 +1506,20 @@ def order.after_ceiling
else order.not_entered state order.Refusal.CeilingBindingMismatch
/-- [auths_gateway_kernel::order::after_write]:
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 609:0-615:1
+ Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 549:0-554:1
Visibility: public -/
def order.after_write
(state : order.SubmitState) (result : order.WriteResult) :
Result order.SubmitDecision
:= do
match result with
- | order.WriteResult.NotEntered =>
- order.not_entered state order.Refusal.TransportNotEntered
| order.WriteResult.Unknown =>
order.go state order.Phase.RecordUnknown order.SubmitAction.RecordUnknown
| order.WriteResult.Response =>
order.go state order.Phase.RecordResponse order.SubmitAction.RecordResponse
/-- [auths_gateway_kernel::order::after_response]:
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 619:0-634:1
+ Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 558:0-563:1
Visibility: public -/
def order.after_response
(state : order.SubmitState) (record1 : order.ResponseRecord) :
@@ -1565,30 +1527,10 @@ def order.after_response
:= do
match record1 with
| order.ResponseRecord.Failed => order.stop state order.Stop.Unknown
- | order.ResponseRecord.Recorded observable =>
- if observable
- then
- order.go state (order.Phase.Reload order.Mode.ReadBack)
- (order.SubmitAction.Reload order.Mode.ReadBack)
- else order.stop state order.Stop.Response
-
-/-- [auths_gateway_kernel::order::after_read_back]:
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 638:0-648:1
- Visibility: public -/
-def order.after_read_back
- (state : order.SubmitState) (mode : order.Mode) (recorded : Bool) :
- Result order.SubmitDecision
- := do
- if recorded
- then order.stop state order.Stop.Observed
- else
- match mode with
- | order.Mode.Entry => order.halt state
- | order.Mode.ReadBack => order.stop state order.Stop.Response
- | order.Mode.Reobserve => order.stop state order.Stop.ReplayRefused
+ | order.ResponseRecord.Recorded => order.stop state order.Stop.Response
/-- [auths_gateway_kernel::order::pre_claim_step]:
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 653:0-705:1
+ Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 568:0-620:1
Visibility: public -/
def order.pre_claim_step
(state : order.SubmitState) (event : order.SubmitEvent) :
@@ -1605,7 +1547,6 @@ def order.pre_claim_step
| order.SubmitEvent.Scope _ => order.halt state
| order.SubmitEvent.Preparation _ => order.halt state
| order.SubmitEvent.Claim _ => order.halt state
- | order.SubmitEvent.Resume _ => order.halt state
| order.SubmitEvent.Reload _ => order.halt state
| order.SubmitEvent.Lease _ => order.halt state
| order.SubmitEvent.Prefix _ => order.halt state
@@ -1615,6 +1556,7 @@ def order.pre_claim_step
| order.SubmitEvent.Ceiling _ => order.halt state
| order.SubmitEvent.Recorded _ => order.halt state
| order.SubmitEvent.Deadline _ => order.halt state
+ | order.SubmitEvent.Granted _ => order.halt state
| order.SubmitEvent.Write _ => order.halt state
| order.SubmitEvent.Response _ => order.halt state
| order.Phase.Clock =>
@@ -1628,7 +1570,6 @@ def order.pre_claim_step
| order.SubmitEvent.Scope _ => order.halt state
| order.SubmitEvent.Preparation _ => order.halt state
| order.SubmitEvent.Claim _ => order.halt state
- | order.SubmitEvent.Resume _ => order.halt state
| order.SubmitEvent.Reload _ => order.halt state
| order.SubmitEvent.Lease _ => order.halt state
| order.SubmitEvent.Prefix _ => order.halt state
@@ -1638,6 +1579,7 @@ def order.pre_claim_step
| order.SubmitEvent.Ceiling _ => order.halt state
| order.SubmitEvent.Recorded _ => order.halt state
| order.SubmitEvent.Deadline _ => order.halt state
+ | order.SubmitEvent.Granted _ => order.halt state
| order.SubmitEvent.Write _ => order.halt state
| order.SubmitEvent.Response _ => order.halt state
| order.Phase.Verify =>
@@ -1657,7 +1599,6 @@ def order.pre_claim_step
| order.SubmitEvent.Scope _ => order.halt state
| order.SubmitEvent.Preparation _ => order.halt state
| order.SubmitEvent.Claim _ => order.halt state
- | order.SubmitEvent.Resume _ => order.halt state
| order.SubmitEvent.Reload _ => order.halt state
| order.SubmitEvent.Lease _ => order.halt state
| order.SubmitEvent.Prefix _ => order.halt state
@@ -1667,6 +1608,7 @@ def order.pre_claim_step
| order.SubmitEvent.Ceiling _ => order.halt state
| order.SubmitEvent.Recorded _ => order.halt state
| order.SubmitEvent.Deadline _ => order.halt state
+ | order.SubmitEvent.Granted _ => order.halt state
| order.SubmitEvent.Write _ => order.halt state
| order.SubmitEvent.Response _ => order.halt state
| order.Phase.Admit =>
@@ -1680,7 +1622,6 @@ def order.pre_claim_step
| order.SubmitEvent.Scope _ => order.halt state
| order.SubmitEvent.Preparation _ => order.halt state
| order.SubmitEvent.Claim _ => order.halt state
- | order.SubmitEvent.Resume _ => order.halt state
| order.SubmitEvent.Reload _ => order.halt state
| order.SubmitEvent.Lease _ => order.halt state
| order.SubmitEvent.Prefix _ => order.halt state
@@ -1690,6 +1631,7 @@ def order.pre_claim_step
| order.SubmitEvent.Ceiling _ => order.halt state
| order.SubmitEvent.Recorded _ => order.halt state
| order.SubmitEvent.Deadline _ => order.halt state
+ | order.SubmitEvent.Granted _ => order.halt state
| order.SubmitEvent.Write _ => order.halt state
| order.SubmitEvent.Response _ => order.halt state
| order.Phase.Scope =>
@@ -1703,7 +1645,6 @@ def order.pre_claim_step
order.pre_claim state bound sd
| order.SubmitEvent.Preparation _ => order.halt state
| order.SubmitEvent.Claim _ => order.halt state
- | order.SubmitEvent.Resume _ => order.halt state
| order.SubmitEvent.Reload _ => order.halt state
| order.SubmitEvent.Lease _ => order.halt state
| order.SubmitEvent.Prefix _ => order.halt state
@@ -1713,6 +1654,7 @@ def order.pre_claim_step
| order.SubmitEvent.Ceiling _ => order.halt state
| order.SubmitEvent.Recorded _ => order.halt state
| order.SubmitEvent.Deadline _ => order.halt state
+ | order.SubmitEvent.Granted _ => order.halt state
| order.SubmitEvent.Write _ => order.halt state
| order.SubmitEvent.Response _ => order.halt state
| order.Phase.Prepare =>
@@ -1726,7 +1668,6 @@ def order.pre_claim_step
let sd ← order.go state order.Phase.Claim order.SubmitAction.Claim
order.pre_claim state prepared sd
| order.SubmitEvent.Claim _ => order.halt state
- | order.SubmitEvent.Resume _ => order.halt state
| order.SubmitEvent.Reload _ => order.halt state
| order.SubmitEvent.Lease _ => order.halt state
| order.SubmitEvent.Prefix _ => order.halt state
@@ -1736,6 +1677,7 @@ def order.pre_claim_step
| order.SubmitEvent.Ceiling _ => order.halt state
| order.SubmitEvent.Recorded _ => order.halt state
| order.SubmitEvent.Deadline _ => order.halt state
+ | order.SubmitEvent.Granted _ => order.halt state
| order.SubmitEvent.Write _ => order.halt state
| order.SubmitEvent.Response _ => order.halt state
| order.Phase.Claim =>
@@ -1747,7 +1689,6 @@ def order.pre_claim_step
| order.SubmitEvent.Scope _ => order.halt state
| order.SubmitEvent.Preparation _ => order.halt state
| order.SubmitEvent.Claim result => order.after_claim state result
- | order.SubmitEvent.Resume _ => order.halt state
| order.SubmitEvent.Reload _ => order.halt state
| order.SubmitEvent.Lease _ => order.halt state
| order.SubmitEvent.Prefix _ => order.halt state
@@ -1757,28 +1698,28 @@ def order.pre_claim_step
| order.SubmitEvent.Ceiling _ => order.halt state
| order.SubmitEvent.Recorded _ => order.halt state
| order.SubmitEvent.Deadline _ => order.halt state
+ | order.SubmitEvent.Granted _ => order.halt state
| order.SubmitEvent.Write _ => order.halt state
| order.SubmitEvent.Response _ => order.halt state
- | order.Phase.Resume => order.halt state
- | order.Phase.Reload _ => order.halt state
- | order.Phase.Lease _ => order.halt state
- | order.Phase.Prefix _ => order.halt state
- | order.Phase.Account _ => order.halt state
- | order.Phase.Denied _ _ => order.halt state
+ | order.Phase.Reload => order.halt state
+ | order.Phase.Lease => order.halt state
+ | order.Phase.Prefix => order.halt state
+ | order.Phase.Account => order.halt state
+ | order.Phase.Denied _ => order.halt state
| order.Phase.PreEntry => order.halt state
| order.Phase.Ceiling => order.halt state
| order.Phase.Checkpoint => order.halt state
| order.Phase.EntryReload => order.halt state
| order.Phase.Deadline => order.halt state
+ | order.Phase.Grant => order.halt state
| order.Phase.Send => order.halt state
| order.Phase.RecordResponse => order.halt state
| order.Phase.RecordUnknown => order.halt state
| order.Phase.RecordNotEntered => order.halt state
- | order.Phase.ReadBack _ => order.halt state
| order.Phase.Done => order.halt state
/-- [auths_gateway_kernel::order::lease_step]:
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 710:0-766:1
+ Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 625:0-667:1
Visibility: public -/
def order.lease_step
(state : order.SubmitState) (event : order.SubmitEvent) :
@@ -1792,33 +1733,7 @@ def order.lease_step
| order.Phase.Scope => order.halt state
| order.Phase.Prepare => order.halt state
| order.Phase.Claim => order.halt state
- | order.Phase.Resume =>
- match event with
- | order.SubmitEvent.Start => order.halt state
- | order.SubmitEvent.Clock _ => order.halt state
- | order.SubmitEvent.Verification _ => order.halt state
- | order.SubmitEvent.Admission _ => order.halt state
- | order.SubmitEvent.Scope _ => order.halt state
- | order.SubmitEvent.Preparation _ => order.halt state
- | order.SubmitEvent.Claim _ => order.halt state
- | order.SubmitEvent.Resume resumable =>
- if resumable
- then
- order.go state (order.Phase.Reload order.Mode.Reobserve)
- (order.SubmitAction.Reload order.Mode.Reobserve)
- else order.stop state order.Stop.ReplayRefused
- | order.SubmitEvent.Reload _ => order.halt state
- | order.SubmitEvent.Lease _ => order.halt state
- | order.SubmitEvent.Prefix _ => order.halt state
- | order.SubmitEvent.Account _ => order.halt state
- | order.SubmitEvent.Denied _ => order.halt state
- | order.SubmitEvent.PreEntry _ => order.halt state
- | order.SubmitEvent.Ceiling _ => order.halt state
- | order.SubmitEvent.Recorded _ => order.halt state
- | order.SubmitEvent.Deadline _ => order.halt state
- | order.SubmitEvent.Write _ => order.halt state
- | order.SubmitEvent.Response _ => order.halt state
- | order.Phase.Reload mode =>
+ | order.Phase.Reload =>
match event with
| order.SubmitEvent.Start => order.halt state
| order.SubmitEvent.Clock _ => order.halt state
@@ -1827,12 +1742,10 @@ def order.lease_step
| order.SubmitEvent.Scope _ => order.halt state
| order.SubmitEvent.Preparation _ => order.halt state
| order.SubmitEvent.Claim _ => order.halt state
- | order.SubmitEvent.Resume _ => order.halt state
| order.SubmitEvent.Reload unchanged =>
if unchanged
- then
- order.go state (order.Phase.Lease mode) (order.SubmitAction.Lease mode)
- else order.lease_failed state mode order.Refusal.ConnectionChanged
+ then order.go state order.Phase.Lease order.SubmitAction.Lease
+ else order.not_entered state order.Refusal.ConnectionChanged
| order.SubmitEvent.Lease _ => order.halt state
| order.SubmitEvent.Prefix _ => order.halt state
| order.SubmitEvent.Account _ => order.halt state
@@ -1841,9 +1754,10 @@ def order.lease_step
| order.SubmitEvent.Ceiling _ => order.halt state
| order.SubmitEvent.Recorded _ => order.halt state
| order.SubmitEvent.Deadline _ => order.halt state
+ | order.SubmitEvent.Granted _ => order.halt state
| order.SubmitEvent.Write _ => order.halt state
| order.SubmitEvent.Response _ => order.halt state
- | order.Phase.Lease mode =>
+ | order.Phase.Lease =>
match event with
| order.SubmitEvent.Start => order.halt state
| order.SubmitEvent.Clock _ => order.halt state
@@ -1852,14 +1766,11 @@ def order.lease_step
| order.SubmitEvent.Scope _ => order.halt state
| order.SubmitEvent.Preparation _ => order.halt state
| order.SubmitEvent.Claim _ => order.halt state
- | order.SubmitEvent.Resume _ => order.halt state
| order.SubmitEvent.Reload _ => order.halt state
| order.SubmitEvent.Lease leased =>
if leased
- then
- order.go state (order.Phase.Prefix mode)
- (order.SubmitAction.CheckPrefix mode)
- else order.lease_failed state mode order.Refusal.CredentialUnavailable
+ then order.go state order.Phase.Prefix order.SubmitAction.CheckPrefix
+ else order.not_entered state order.Refusal.CredentialUnavailable
| order.SubmitEvent.Prefix _ => order.halt state
| order.SubmitEvent.Account _ => order.halt state
| order.SubmitEvent.Denied _ => order.halt state
@@ -1867,9 +1778,10 @@ def order.lease_step
| order.SubmitEvent.Ceiling _ => order.halt state
| order.SubmitEvent.Recorded _ => order.halt state
| order.SubmitEvent.Deadline _ => order.halt state
+ | order.SubmitEvent.Granted _ => order.halt state
| order.SubmitEvent.Write _ => order.halt state
| order.SubmitEvent.Response _ => order.halt state
- | order.Phase.Prefix mode =>
+ | order.Phase.Prefix =>
match event with
| order.SubmitEvent.Start => order.halt state
| order.SubmitEvent.Clock _ => order.halt state
@@ -1878,22 +1790,22 @@ def order.lease_step
| order.SubmitEvent.Scope _ => order.halt state
| order.SubmitEvent.Preparation _ => order.halt state
| order.SubmitEvent.Claim _ => order.halt state
- | order.SubmitEvent.Resume _ => order.halt state
| order.SubmitEvent.Reload _ => order.halt state
| order.SubmitEvent.Lease _ => order.halt state
| order.SubmitEvent.Prefix allowed =>
if allowed
- then order.after_prefix state mode
- else order.lease_failed state mode order.Refusal.ModeGuard
+ then order.after_prefix state
+ else order.not_entered state order.Refusal.ModeGuard
| order.SubmitEvent.Account _ => order.halt state
| order.SubmitEvent.Denied _ => order.halt state
| order.SubmitEvent.PreEntry _ => order.halt state
| order.SubmitEvent.Ceiling _ => order.halt state
| order.SubmitEvent.Recorded _ => order.halt state
| order.SubmitEvent.Deadline _ => order.halt state
+ | order.SubmitEvent.Granted _ => order.halt state
| order.SubmitEvent.Write _ => order.halt state
| order.SubmitEvent.Response _ => order.halt state
- | order.Phase.Account mode =>
+ | order.Phase.Account =>
match event with
| order.SubmitEvent.Start => order.halt state
| order.SubmitEvent.Clock _ => order.halt state
@@ -1902,19 +1814,19 @@ def order.lease_step
| order.SubmitEvent.Scope _ => order.halt state
| order.SubmitEvent.Preparation _ => order.halt state
| order.SubmitEvent.Claim _ => order.halt state
- | order.SubmitEvent.Resume _ => order.halt state
| order.SubmitEvent.Reload _ => order.halt state
| order.SubmitEvent.Lease _ => order.halt state
| order.SubmitEvent.Prefix _ => order.halt state
- | order.SubmitEvent.Account result => order.after_account state mode result
+ | order.SubmitEvent.Account result => order.after_account state result
| order.SubmitEvent.Denied _ => order.halt state
| order.SubmitEvent.PreEntry _ => order.halt state
| order.SubmitEvent.Ceiling _ => order.halt state
| order.SubmitEvent.Recorded _ => order.halt state
| order.SubmitEvent.Deadline _ => order.halt state
+ | order.SubmitEvent.Granted _ => order.halt state
| order.SubmitEvent.Write _ => order.halt state
| order.SubmitEvent.Response _ => order.halt state
- | order.Phase.Denied mode index =>
+ | order.Phase.Denied index =>
match event with
| order.SubmitEvent.Start => order.halt state
| order.SubmitEvent.Clock _ => order.halt state
@@ -1923,17 +1835,16 @@ def order.lease_step
| order.SubmitEvent.Scope _ => order.halt state
| order.SubmitEvent.Preparation _ => order.halt state
| order.SubmitEvent.Claim _ => order.halt state
- | order.SubmitEvent.Resume _ => order.halt state
| order.SubmitEvent.Reload _ => order.halt state
| order.SubmitEvent.Lease _ => order.halt state
| order.SubmitEvent.Prefix _ => order.halt state
| order.SubmitEvent.Account _ => order.halt state
- | order.SubmitEvent.Denied result =>
- order.after_denied state mode index result
+ | order.SubmitEvent.Denied result => order.after_denied state index result
| order.SubmitEvent.PreEntry _ => order.halt state
| order.SubmitEvent.Ceiling _ => order.halt state
| order.SubmitEvent.Recorded _ => order.halt state
| order.SubmitEvent.Deadline _ => order.halt state
+ | order.SubmitEvent.Granted _ => order.halt state
| order.SubmitEvent.Write _ => order.halt state
| order.SubmitEvent.Response _ => order.halt state
| order.Phase.PreEntry => order.halt state
@@ -1941,15 +1852,15 @@ def order.lease_step
| order.Phase.Checkpoint => order.halt state
| order.Phase.EntryReload => order.halt state
| order.Phase.Deadline => order.halt state
+ | order.Phase.Grant => order.halt state
| order.Phase.Send => order.halt state
| order.Phase.RecordResponse => order.halt state
| order.Phase.RecordUnknown => order.halt state
| order.Phase.RecordNotEntered => order.halt state
- | order.Phase.ReadBack _ => order.halt state
| order.Phase.Done => order.halt state
/-- [auths_gateway_kernel::order::entry_step]:
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 772:0-840:1
+ Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 673:0-742:1
Visibility: public -/
def order.entry_step
(state : order.SubmitState) (event : order.SubmitEvent) :
@@ -1963,12 +1874,11 @@ def order.entry_step
| order.Phase.Scope => order.halt state
| order.Phase.Prepare => order.halt state
| order.Phase.Claim => order.halt state
- | order.Phase.Resume => order.halt state
- | order.Phase.Reload _ => order.halt state
- | order.Phase.Lease _ => order.halt state
- | order.Phase.Prefix _ => order.halt state
- | order.Phase.Account _ => order.halt state
- | order.Phase.Denied _ _ => order.halt state
+ | order.Phase.Reload => order.halt state
+ | order.Phase.Lease => order.halt state
+ | order.Phase.Prefix => order.halt state
+ | order.Phase.Account => order.halt state
+ | order.Phase.Denied _ => order.halt state
| order.Phase.PreEntry =>
match event with
| order.SubmitEvent.Start => order.halt state
@@ -1978,7 +1888,6 @@ def order.entry_step
| order.SubmitEvent.Scope _ => order.halt state
| order.SubmitEvent.Preparation _ => order.halt state
| order.SubmitEvent.Claim _ => order.halt state
- | order.SubmitEvent.Resume _ => order.halt state
| order.SubmitEvent.Reload _ => order.halt state
| order.SubmitEvent.Lease _ => order.halt state
| order.SubmitEvent.Prefix _ => order.halt state
@@ -1988,6 +1897,7 @@ def order.entry_step
| order.SubmitEvent.Ceiling _ => order.halt state
| order.SubmitEvent.Recorded _ => order.halt state
| order.SubmitEvent.Deadline _ => order.halt state
+ | order.SubmitEvent.Granted _ => order.halt state
| order.SubmitEvent.Write _ => order.halt state
| order.SubmitEvent.Response _ => order.halt state
| order.Phase.Ceiling =>
@@ -1999,7 +1909,6 @@ def order.entry_step
| order.SubmitEvent.Scope _ => order.halt state
| order.SubmitEvent.Preparation _ => order.halt state
| order.SubmitEvent.Claim _ => order.halt state
- | order.SubmitEvent.Resume _ => order.halt state
| order.SubmitEvent.Reload _ => order.halt state
| order.SubmitEvent.Lease _ => order.halt state
| order.SubmitEvent.Prefix _ => order.halt state
@@ -2009,6 +1918,7 @@ def order.entry_step
| order.SubmitEvent.Ceiling read => order.after_ceiling state read
| order.SubmitEvent.Recorded _ => order.halt state
| order.SubmitEvent.Deadline _ => order.halt state
+ | order.SubmitEvent.Granted _ => order.halt state
| order.SubmitEvent.Write _ => order.halt state
| order.SubmitEvent.Response _ => order.halt state
| order.Phase.Checkpoint =>
@@ -2020,7 +1930,6 @@ def order.entry_step
| order.SubmitEvent.Scope _ => order.halt state
| order.SubmitEvent.Preparation _ => order.halt state
| order.SubmitEvent.Claim _ => order.halt state
- | order.SubmitEvent.Resume _ => order.halt state
| order.SubmitEvent.Reload _ => order.halt state
| order.SubmitEvent.Lease _ => order.halt state
| order.SubmitEvent.Prefix _ => order.halt state
@@ -2035,6 +1944,7 @@ def order.entry_step
order.SubmitAction.ReloadBeforeEntry
else order.stop state order.Stop.Unknown
| order.SubmitEvent.Deadline _ => order.halt state
+ | order.SubmitEvent.Granted _ => order.halt state
| order.SubmitEvent.Write _ => order.halt state
| order.SubmitEvent.Response _ => order.halt state
| order.Phase.EntryReload =>
@@ -2046,7 +1956,6 @@ def order.entry_step
| order.SubmitEvent.Scope _ => order.halt state
| order.SubmitEvent.Preparation _ => order.halt state
| order.SubmitEvent.Claim _ => order.halt state
- | order.SubmitEvent.Resume _ => order.halt state
| order.SubmitEvent.Reload unchanged =>
if unchanged
then order.go state order.Phase.Deadline order.SubmitAction.CheckDeadline
@@ -2059,6 +1968,7 @@ def order.entry_step
| order.SubmitEvent.Ceiling _ => order.halt state
| order.SubmitEvent.Recorded _ => order.halt state
| order.SubmitEvent.Deadline _ => order.halt state
+ | order.SubmitEvent.Granted _ => order.halt state
| order.SubmitEvent.Write _ => order.halt state
| order.SubmitEvent.Response _ => order.halt state
| order.Phase.Deadline =>
@@ -2070,7 +1980,6 @@ def order.entry_step
| order.SubmitEvent.Scope _ => order.halt state
| order.SubmitEvent.Preparation _ => order.halt state
| order.SubmitEvent.Claim _ => order.halt state
- | order.SubmitEvent.Resume _ => order.halt state
| order.SubmitEvent.Reload _ => order.halt state
| order.SubmitEvent.Lease _ => order.halt state
| order.SubmitEvent.Prefix _ => order.halt state
@@ -2081,11 +1990,12 @@ def order.entry_step
| order.SubmitEvent.Recorded _ => order.halt state
| order.SubmitEvent.Deadline within =>
if within
- then order.go state order.Phase.Send order.SubmitAction.Send
+ then order.go state order.Phase.Grant order.SubmitAction.GrantEntry
else order.not_entered state order.Refusal.EntryDeadline
+ | order.SubmitEvent.Granted _ => order.halt state
| order.SubmitEvent.Write _ => order.halt state
| order.SubmitEvent.Response _ => order.halt state
- | order.Phase.Send =>
+ | order.Phase.Grant =>
match event with
| order.SubmitEvent.Start => order.halt state
| order.SubmitEvent.Clock _ => order.halt state
@@ -2094,7 +2004,6 @@ def order.entry_step
| order.SubmitEvent.Scope _ => order.halt state
| order.SubmitEvent.Preparation _ => order.halt state
| order.SubmitEvent.Claim _ => order.halt state
- | order.SubmitEvent.Resume _ => order.halt state
| order.SubmitEvent.Reload _ => order.halt state
| order.SubmitEvent.Lease _ => order.halt state
| order.SubmitEvent.Prefix _ => order.halt state
@@ -2104,9 +2013,13 @@ def order.entry_step
| order.SubmitEvent.Ceiling _ => order.halt state
| order.SubmitEvent.Recorded _ => order.halt state
| order.SubmitEvent.Deadline _ => order.halt state
- | order.SubmitEvent.Write result => order.after_write state result
+ | order.SubmitEvent.Granted b =>
+ if b
+ then order.go state order.Phase.Send order.SubmitAction.Send
+ else order.stop state order.Stop.GrantUnavailable
+ | order.SubmitEvent.Write _ => order.halt state
| order.SubmitEvent.Response _ => order.halt state
- | order.Phase.RecordResponse =>
+ | order.Phase.Send =>
match event with
| order.SubmitEvent.Start => order.halt state
| order.SubmitEvent.Clock _ => order.halt state
@@ -2115,7 +2028,6 @@ def order.entry_step
| order.SubmitEvent.Scope _ => order.halt state
| order.SubmitEvent.Preparation _ => order.halt state
| order.SubmitEvent.Claim _ => order.halt state
- | order.SubmitEvent.Resume _ => order.halt state
| order.SubmitEvent.Reload _ => order.halt state
| order.SubmitEvent.Lease _ => order.halt state
| order.SubmitEvent.Prefix _ => order.halt state
@@ -2125,9 +2037,10 @@ def order.entry_step
| order.SubmitEvent.Ceiling _ => order.halt state
| order.SubmitEvent.Recorded _ => order.halt state
| order.SubmitEvent.Deadline _ => order.halt state
- | order.SubmitEvent.Write _ => order.halt state
- | order.SubmitEvent.Response record1 => order.after_response state record1
- | order.Phase.RecordUnknown =>
+ | order.SubmitEvent.Granted _ => order.halt state
+ | order.SubmitEvent.Write result => order.after_write state result
+ | order.SubmitEvent.Response _ => order.halt state
+ | order.Phase.RecordResponse =>
match event with
| order.SubmitEvent.Start => order.halt state
| order.SubmitEvent.Clock _ => order.halt state
@@ -2136,7 +2049,6 @@ def order.entry_step
| order.SubmitEvent.Scope _ => order.halt state
| order.SubmitEvent.Preparation _ => order.halt state
| order.SubmitEvent.Claim _ => order.halt state
- | order.SubmitEvent.Resume _ => order.halt state
| order.SubmitEvent.Reload _ => order.halt state
| order.SubmitEvent.Lease _ => order.halt state
| order.SubmitEvent.Prefix _ => order.halt state
@@ -2144,11 +2056,12 @@ def order.entry_step
| order.SubmitEvent.Denied _ => order.halt state
| order.SubmitEvent.PreEntry _ => order.halt state
| order.SubmitEvent.Ceiling _ => order.halt state
- | order.SubmitEvent.Recorded _ => order.stop state order.Stop.Unknown
+ | order.SubmitEvent.Recorded _ => order.halt state
| order.SubmitEvent.Deadline _ => order.halt state
+ | order.SubmitEvent.Granted _ => order.halt state
| order.SubmitEvent.Write _ => order.halt state
- | order.SubmitEvent.Response _ => order.halt state
- | order.Phase.RecordNotEntered =>
+ | order.SubmitEvent.Response record1 => order.after_response state record1
+ | order.Phase.RecordUnknown =>
match event with
| order.SubmitEvent.Start => order.halt state
| order.SubmitEvent.Clock _ => order.halt state
@@ -2157,7 +2070,6 @@ def order.entry_step
| order.SubmitEvent.Scope _ => order.halt state
| order.SubmitEvent.Preparation _ => order.halt state
| order.SubmitEvent.Claim _ => order.halt state
- | order.SubmitEvent.Resume _ => order.halt state
| order.SubmitEvent.Reload _ => order.halt state
| order.SubmitEvent.Lease _ => order.halt state
| order.SubmitEvent.Prefix _ => order.halt state
@@ -2165,14 +2077,12 @@ def order.entry_step
| order.SubmitEvent.Denied _ => order.halt state
| order.SubmitEvent.PreEntry _ => order.halt state
| order.SubmitEvent.Ceiling _ => order.halt state
- | order.SubmitEvent.Recorded recorded =>
- if recorded
- then order.stop state order.Stop.NotEntered
- else order.stop state order.Stop.Unknown
+ | order.SubmitEvent.Recorded _ => order.stop state order.Stop.Unknown
| order.SubmitEvent.Deadline _ => order.halt state
+ | order.SubmitEvent.Granted _ => order.halt state
| order.SubmitEvent.Write _ => order.halt state
| order.SubmitEvent.Response _ => order.halt state
- | order.Phase.ReadBack mode =>
+ | order.Phase.RecordNotEntered =>
match event with
| order.SubmitEvent.Start => order.halt state
| order.SubmitEvent.Clock _ => order.halt state
@@ -2181,7 +2091,6 @@ def order.entry_step
| order.SubmitEvent.Scope _ => order.halt state
| order.SubmitEvent.Preparation _ => order.halt state
| order.SubmitEvent.Claim _ => order.halt state
- | order.SubmitEvent.Resume _ => order.halt state
| order.SubmitEvent.Reload _ => order.halt state
| order.SubmitEvent.Lease _ => order.halt state
| order.SubmitEvent.Prefix _ => order.halt state
@@ -2190,14 +2099,17 @@ def order.entry_step
| order.SubmitEvent.PreEntry _ => order.halt state
| order.SubmitEvent.Ceiling _ => order.halt state
| order.SubmitEvent.Recorded recorded =>
- order.after_read_back state mode recorded
+ if recorded
+ then order.stop state order.Stop.NotEntered
+ else order.stop state order.Stop.Unknown
| order.SubmitEvent.Deadline _ => order.halt state
+ | order.SubmitEvent.Granted _ => order.halt state
| order.SubmitEvent.Write _ => order.halt state
| order.SubmitEvent.Response _ => order.halt state
| order.Phase.Done => order.halt state
/-- [auths_gateway_kernel::order::next_step]:
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 846:0-873:1
+ Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 748:0-772:1
Visibility: public -/
def order.next_step
(state : order.SubmitState) (event : order.SubmitEvent) :
@@ -2211,22 +2123,21 @@ def order.next_step
| order.Phase.Scope => order.pre_claim_step state event
| order.Phase.Prepare => order.pre_claim_step state event
| order.Phase.Claim => order.pre_claim_step state event
- | order.Phase.Resume => order.lease_step state event
- | order.Phase.Reload _ => order.lease_step state event
- | order.Phase.Lease _ => order.lease_step state event
- | order.Phase.Prefix _ => order.lease_step state event
- | order.Phase.Account _ => order.lease_step state event
- | order.Phase.Denied _ _ => order.lease_step state event
+ | order.Phase.Reload => order.lease_step state event
+ | order.Phase.Lease => order.lease_step state event
+ | order.Phase.Prefix => order.lease_step state event
+ | order.Phase.Account => order.lease_step state event
+ | order.Phase.Denied _ => order.lease_step state event
| order.Phase.PreEntry => order.entry_step state event
| order.Phase.Ceiling => order.entry_step state event
| order.Phase.Checkpoint => order.entry_step state event
| order.Phase.EntryReload => order.entry_step state event
| order.Phase.Deadline => order.entry_step state event
+ | order.Phase.Grant => order.entry_step state event
| order.Phase.Send => order.entry_step state event
| order.Phase.RecordResponse => order.entry_step state event
| order.Phase.RecordUnknown => order.entry_step state event
| order.Phase.RecordNotEntered => order.entry_step state event
- | order.Phase.ReadBack _ => order.entry_step state event
| order.Phase.Done => order.halt state
/-- [auths_gateway_kernel::ratio::relative_basis]:
@@ -2312,14 +2223,14 @@ def recovery.recovery_capability
}
/-- [auths_gateway_kernel::transition::absent]:
- Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 89:0-91:1
+ Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 113:0-115:1
Visibility: public -/
def transition.absent (bytes : Slice Std.U8) : Result Bool := do
let i := Slice.len bytes
ok (i = 0#usize)
/-- [auths_gateway_kernel::transition::same_bytes_from]: loop body 0:
- Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 97:4-104:1
+ Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 121:4-128:1
Visibility: public -/
@[rust_loop_body]
def transition.same_bytes_from_loop.body
@@ -2342,7 +2253,7 @@ def transition.same_bytes_from_loop.body
else ok (done true)
/-- [auths_gateway_kernel::transition::same_bytes_from]: loop 0:
- Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 97:4-104:1
+ Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 121:4-128:1
Visibility: public -/
@[rust_loop]
def transition.same_bytes_from_loop
@@ -2354,7 +2265,7 @@ def transition.same_bytes_from_loop
index
/-- [auths_gateway_kernel::transition::same_bytes_from]:
- Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 95:0-104:1
+ Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 119:0-128:1
Visibility: public -/
@[reducible]
def transition.same_bytes_from
@@ -2362,7 +2273,7 @@ def transition.same_bytes_from
transition.same_bytes_from_loop left right 0#usize
/-- [auths_gateway_kernel::transition::same_bytes]:
- Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 108:0-114:1
+ Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 132:0-138:1
Visibility: public -/
def transition.same_bytes
(left : Slice Std.U8) (right : Slice Std.U8) : Result Bool := do
@@ -2373,7 +2284,7 @@ def transition.same_bytes
else ok false
/-- [auths_gateway_kernel::transition::resolves_unknown]:
- Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 119:0-124:1
+ Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 143:0-148:1
Visibility: public -/
def transition.resolves_unknown (link : transition.Link) : Result Bool := do
match link with
@@ -2382,7 +2293,7 @@ def transition.resolves_unknown (link : transition.Link) : Result Bool := do
| transition.Link.AfterResponse => ok false
/-- [auths_gateway_kernel::transition::has_link]:
- Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 128:0-133:1
+ Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 152:0-157:1
Visibility: public -/
def transition.has_link (link : transition.Link) : Result Bool := do
match link with
@@ -2391,67 +2302,77 @@ def transition.has_link (link : transition.Link) : Result Bool := do
| transition.Link.AfterResponse => ok true
/-- [auths_gateway_kernel::transition::refusal_stage]:
- Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 137:0-146:1
+ Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 161:0-172:1
Visibility: public -/
def transition.refusal_stage (stage : transition.Stage) : Result Bool := do
match stage with
- | transition.Stage.Attempting => ok false
+ | transition.Stage.Reserved => ok false
+ | transition.Stage.EntryGranted => ok false
| transition.Stage.NotEntered => ok true
| transition.Stage.Unknown => ok false
| transition.Stage.ResponseRecorded => ok false
| transition.Stage.Observed => ok false
| transition.Stage.ObservedByProvider => ok false
+ | transition.Stage.Acknowledged => ok false
/-- [auths_gateway_kernel::transition::requires_response]:
- Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 150:0-155:1
+ Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 176:0-186:1
Visibility: public -/
def transition.requires_response (stage : transition.Stage) : Result Bool := do
match stage with
- | transition.Stage.Attempting => ok false
+ | transition.Stage.Reserved => ok false
+ | transition.Stage.EntryGranted => ok false
| transition.Stage.NotEntered => ok false
| transition.Stage.Unknown => ok false
| transition.Stage.ResponseRecorded => ok true
| transition.Stage.Observed => ok true
| transition.Stage.ObservedByProvider => ok false
+ | transition.Stage.Acknowledged => ok false
/-- [auths_gateway_kernel::transition::permits_response]:
- Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 159:0-164:1
+ Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 190:0-199:1
Visibility: public -/
def transition.permits_response (stage : transition.Stage) : Result Bool := do
match stage with
- | transition.Stage.Attempting => ok false
+ | transition.Stage.Reserved => ok false
+ | transition.Stage.EntryGranted => ok false
| transition.Stage.NotEntered => ok false
| transition.Stage.Unknown => ok false
| transition.Stage.ResponseRecorded => ok true
| transition.Stage.Observed => ok true
| transition.Stage.ObservedByProvider => ok true
+ | transition.Stage.Acknowledged => ok false
/-- [auths_gateway_kernel::transition::observed_stage]:
- Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 168:0-177:1
+ Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 203:0-214:1
Visibility: public -/
def transition.observed_stage (stage : transition.Stage) : Result Bool := do
match stage with
- | transition.Stage.Attempting => ok false
+ | transition.Stage.Reserved => ok false
+ | transition.Stage.EntryGranted => ok false
| transition.Stage.NotEntered => ok false
| transition.Stage.Unknown => ok false
| transition.Stage.ResponseRecorded => ok false
| transition.Stage.Observed => ok true
| transition.Stage.ObservedByProvider => ok false
+ | transition.Stage.Acknowledged => ok false
/-- [auths_gateway_kernel::transition::provider_stage]:
- Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 181:0-190:1
+ Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 218:0-229:1
Visibility: public -/
def transition.provider_stage (stage : transition.Stage) : Result Bool := do
match stage with
- | transition.Stage.Attempting => ok false
+ | transition.Stage.Reserved => ok false
+ | transition.Stage.EntryGranted => ok false
| transition.Stage.NotEntered => ok false
| transition.Stage.Unknown => ok false
| transition.Stage.ResponseRecorded => ok false
| transition.Stage.Observed => ok false
| transition.Stage.ObservedByProvider => ok true
+ | transition.Stage.Acknowledged => ok false
/-- [auths_gateway_kernel::transition::has_reading]:
- Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 203:0-208:1
+ Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 244:0-249:1
Visibility: public -/
def transition.has_reading (reading : transition.Reading) : Result Bool := do
match reading with
@@ -2460,217 +2381,401 @@ def transition.has_reading (reading : transition.Reading) : Result Bool := do
| transition.Reading.Mismatch => ok true
| transition.Reading.EchoMismatch => ok true
-/-- [auths_gateway_kernel::transition::consistent]:
- Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 217:0-228:1
+/-- [auths_gateway_kernel::transition::requires_entry]:
+ Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 336:0-346:1
Visibility: public -/
-def transition.consistent (view : transition.AttemptView) : Result Bool := do
- let s := alloc.vec.Vec.deref view.response
- let b ← transition.absent s
- let b1 ← transition.refusal_stage view.stage
- if b1 = view.refusal
- then
- let b2 ← transition.requires_response view.stage
- if b2
+def transition.requires_entry (stage : transition.Stage) : Result Bool := do
+ match stage with
+ | transition.Stage.Reserved => ok false
+ | transition.Stage.EntryGranted => ok true
+ | transition.Stage.NotEntered => ok false
+ | transition.Stage.Unknown => ok true
+ | transition.Stage.ResponseRecorded => ok true
+ | transition.Stage.Observed => ok true
+ | transition.Stage.ObservedByProvider => ok true
+ | transition.Stage.Acknowledged => ok true
+
+/-- [auths_gateway_kernel::transition::acknowledgement_stage]:
+ Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 278:0-280:1
+ Visibility: public -/
+def transition.acknowledgement_stage
+ (stage : transition.Stage) : Result Bool := do
+ match stage with
+ | transition.Stage.Reserved => ok false
+ | transition.Stage.EntryGranted => ok false
+ | transition.Stage.NotEntered => ok false
+ | transition.Stage.Unknown => ok false
+ | transition.Stage.ResponseRecorded => ok false
+ | transition.Stage.Observed => ok false
+ | transition.Stage.ObservedByProvider => ok false
+ | transition.Stage.Acknowledged => ok true
+
+/-- [auths_gateway_kernel::transition::protocol_consistent]:
+ Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 285:0-321:1
+ Visibility: public -/
+def transition.protocol_consistent
+ (view : transition.AttemptView) : Result Bool := do
+ match view.protocol with
+ | transition.Protocol.Mutation =>
+ let b ← transition.acknowledgement_stage view.stage
+ if b
+ then ok false
+ else let s := alloc.vec.Vec.deref view.acknowledgement
+ transition.absent s
+ | transition.Protocol.Acquisition =>
+ let b ←
+ match view.stage with
+ | transition.Stage.Reserved => ok true
+ | transition.Stage.EntryGranted => ok true
+ | transition.Stage.NotEntered => ok true
+ | transition.Stage.Unknown => ok true
+ | transition.Stage.ResponseRecorded => ok true
+ | transition.Stage.Observed => ok false
+ | transition.Stage.ObservedByProvider => ok false
+ | transition.Stage.Acknowledged => ok false
+ if b
then
- if ¬ b
+ let s := alloc.vec.Vec.deref view.acknowledgement
+ let b1 ← transition.absent s
+ if b1
then
- let b3 ← transition.permits_response view.stage
- if b3
+ let s1 := alloc.vec.Vec.deref view.pre_entry
+ let b2 ← transition.absent s1
+ if b2
then
- let s1 := alloc.vec.Vec.deref view.locator
- let b4 ← transition.absent s1
- if b4
+ let s2 := alloc.vec.Vec.deref view.locator
+ let b3 ← transition.absent s2
+ if b3
then
- let b5 ← transition.observed_stage view.stage
- let b6 ← transition.has_reading view.reading
- if b5 = b6
- then
- if b6
- then
- if view.observation
- then
- let b7 ← transition.provider_stage view.stage
- if b7 = view.evidence
- then
- if view.evidence
- then
- let b8 ← transition.has_link view.link
- if b8
- then ok true
- else ok false
- else ok true
- else ok false
- else ok false
+ if view.observation
+ then ok false
+ else
+ if view.evidence
+ then ok false
else
- let b7 ← transition.provider_stage view.stage
- if b7 = view.evidence
- then
- if view.evidence
- then
- let b8 ← transition.has_link view.link
- if b8
- then ok true
- else ok false
- else ok true
- else ok false
- else ok false
- else
- let b5 ← transition.observed_stage view.stage
- let b6 ← transition.has_reading view.reading
- if b5 = b6
- then
- if b6
- then
- if view.observation
+ let b4 ←
+ match view.reading with
+ | transition.Reading.None => ok true
+ | transition.Reading.Match => ok false
+ | transition.Reading.Mismatch => ok false
+ | transition.Reading.EchoMismatch => ok false
+ if b4
then
- let b7 ← transition.provider_stage view.stage
- if b7 = view.evidence
- then
- if view.evidence
- then
- let b8 ← transition.has_link view.link
- if b8
- then ok true
- else ok false
- else ok true
- else ok false
+ match view.link with
+ | transition.Link.None => ok true
+ | transition.Link.Verified => ok false
+ | transition.Link.AfterResponse => ok false
else ok false
+ else ok false
+ else ok false
+ else ok false
+ else ok false
+ | transition.Protocol.Disclosure =>
+ let b ←
+ match view.stage with
+ | transition.Stage.Reserved => ok true
+ | transition.Stage.EntryGranted => ok true
+ | transition.Stage.NotEntered => ok true
+ | transition.Stage.Unknown => ok true
+ | transition.Stage.ResponseRecorded => ok false
+ | transition.Stage.Observed => ok false
+ | transition.Stage.ObservedByProvider => ok false
+ | transition.Stage.Acknowledged => ok true
+ if b
+ then
+ let s := alloc.vec.Vec.deref view.response
+ let b1 ← transition.absent s
+ if b1
+ then
+ let s1 := alloc.vec.Vec.deref view.pre_entry
+ let b2 ← transition.absent s1
+ if b2
+ then
+ let s2 := alloc.vec.Vec.deref view.locator
+ let b3 ← transition.absent s2
+ if b3
+ then
+ if view.observation
+ then ok false
+ else
+ if view.evidence
+ then ok false
else
- let b7 ← transition.provider_stage view.stage
- if b7 = view.evidence
+ let b4 ←
+ match view.reading with
+ | transition.Reading.None => ok true
+ | transition.Reading.Match => ok false
+ | transition.Reading.Mismatch => ok false
+ | transition.Reading.EchoMismatch => ok false
+ if b4
then
- if view.evidence
- then
- let b8 ← transition.has_link view.link
- if b8
- then ok true
- else ok false
- else ok true
+ match view.link with
+ | transition.Link.None => ok true
+ | transition.Link.Verified => ok false
+ | transition.Link.AfterResponse => ok false
else ok false
- else ok false
+ else ok false
else ok false
else ok false
- else
- if ¬ b
+ else ok false
+
+/-- [auths_gateway_kernel::transition::consistent]:
+ Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 258:0-274:1
+ Visibility: public -/
+def transition.consistent (view : transition.AttemptView) : Result Bool := do
+ let s := alloc.vec.Vec.deref view.response
+ let b ← transition.absent s
+ let b1 ← transition.protocol_consistent view
+ if b1
+ then
+ if view.revision > 0#u64
+ then
+ if view.revision <= 9007199254740991#u64
then
- let b3 ← transition.permits_response view.stage
- if b3
+ let b2 ← transition.requires_entry view.stage
+ let s1 := alloc.vec.Vec.deref view.entry
+ let b3 ← transition.absent s1
+ if b2 != b3
then
- let s1 := alloc.vec.Vec.deref view.locator
- let b4 ← transition.absent s1
- if b4
+ let b4 ← transition.acknowledgement_stage view.stage
+ let s2 := alloc.vec.Vec.deref view.acknowledgement
+ let b5 ← transition.absent s2
+ if b4 != b5
then
- let b5 ← transition.observed_stage view.stage
- let b6 ← transition.has_reading view.reading
- if b5 = b6
+ let b6 ← transition.refusal_stage view.stage
+ if b6 = view.refusal
then
- if b6
+ let b7 ← transition.requires_response view.stage
+ if b7
then
- if view.observation
+ if ¬ b
then
- let b7 ← transition.provider_stage view.stage
- if b7 = view.evidence
+ let b8 ← transition.permits_response view.stage
+ if b8
then
- if view.evidence
+ let s3 := alloc.vec.Vec.deref view.locator
+ let b9 ← transition.absent s3
+ if b9
then
- let b8 ← transition.has_link view.link
- if b8
- then ok true
+ let b10 ← transition.observed_stage view.stage
+ let b11 ← transition.has_reading view.reading
+ if b10 = b11
+ then
+ if b11
+ then
+ if view.observation
+ then
+ let b12 ← transition.provider_stage view.stage
+ if b12 = view.evidence
+ then
+ if view.evidence
+ then
+ let b13 ← transition.has_link view.link
+ if b13
+ then ok true
+ else ok false
+ else ok true
+ else ok false
+ else ok false
+ else
+ let b12 ← transition.provider_stage view.stage
+ if b12 = view.evidence
+ then
+ if view.evidence
+ then
+ let b13 ← transition.has_link view.link
+ if b13
+ then ok true
+ else ok false
+ else ok true
+ else ok false
+ else ok false
+ else
+ let b10 ← transition.observed_stage view.stage
+ let b11 ← transition.has_reading view.reading
+ if b10 = b11
+ then
+ if b11
+ then
+ if view.observation
+ then
+ let b12 ← transition.provider_stage view.stage
+ if b12 = view.evidence
+ then
+ if view.evidence
+ then
+ let b13 ← transition.has_link view.link
+ if b13
+ then ok true
+ else ok false
+ else ok true
+ else ok false
+ else ok false
+ else
+ let b12 ← transition.provider_stage view.stage
+ if b12 = view.evidence
+ then
+ if view.evidence
+ then
+ let b13 ← transition.has_link view.link
+ if b13
+ then ok true
+ else ok false
+ else ok true
+ else ok false
else ok false
- else ok true
else ok false
else ok false
else
- let b7 ← transition.provider_stage view.stage
- if b7 = view.evidence
+ if ¬ b
then
- if view.evidence
+ let b8 ← transition.permits_response view.stage
+ if b8
then
- let b8 ← transition.has_link view.link
- if b8
- then ok true
- else ok false
- else ok true
- else ok false
- else ok false
- else
- let b5 ← transition.observed_stage view.stage
- let b6 ← transition.has_reading view.reading
- if b5 = b6
- then
- if b6
- then
- if view.observation
- then
- let b7 ← transition.provider_stage view.stage
- if b7 = view.evidence
- then
- if view.evidence
+ let s3 := alloc.vec.Vec.deref view.locator
+ let b9 ← transition.absent s3
+ if b9
then
- let b8 ← transition.has_link view.link
- if b8
- then ok true
+ let b10 ← transition.observed_stage view.stage
+ let b11 ← transition.has_reading view.reading
+ if b10 = b11
+ then
+ if b11
+ then
+ if view.observation
+ then
+ let b12 ← transition.provider_stage view.stage
+ if b12 = view.evidence
+ then
+ if view.evidence
+ then
+ let b13 ← transition.has_link view.link
+ if b13
+ then ok true
+ else ok false
+ else ok true
+ else ok false
+ else ok false
+ else
+ let b12 ← transition.provider_stage view.stage
+ if b12 = view.evidence
+ then
+ if view.evidence
+ then
+ let b13 ← transition.has_link view.link
+ if b13
+ then ok true
+ else ok false
+ else ok true
+ else ok false
+ else ok false
+ else
+ let b10 ← transition.observed_stage view.stage
+ let b11 ← transition.has_reading view.reading
+ if b10 = b11
+ then
+ if b11
+ then
+ if view.observation
+ then
+ let b12 ← transition.provider_stage view.stage
+ if b12 = view.evidence
+ then
+ if view.evidence
+ then
+ let b13 ← transition.has_link view.link
+ if b13
+ then ok true
+ else ok false
+ else ok true
+ else ok false
+ else ok false
+ else
+ let b12 ← transition.provider_stage view.stage
+ if b12 = view.evidence
+ then
+ if view.evidence
+ then
+ let b13 ← transition.has_link view.link
+ if b13
+ then ok true
+ else ok false
+ else ok true
+ else ok false
else ok false
- else ok true
else ok false
- else ok false
- else
- let b7 ← transition.provider_stage view.stage
- if b7 = view.evidence
- then
- if view.evidence
- then
- let b8 ← transition.has_link view.link
- if b8
- then ok true
- else ok false
- else ok true
- else ok false
- else ok false
- else ok false
- else
- let s1 := alloc.vec.Vec.deref view.locator
- let b3 ← transition.absent s1
- if b3
- then
- let b4 ← transition.observed_stage view.stage
- let b5 ← transition.has_reading view.reading
- if b4 = b5
- then
- if b5
- then
- if view.observation
- then
- let b6 ← transition.provider_stage view.stage
- if b6 = view.evidence
- then
- if view.evidence
+ else
+ let s3 := alloc.vec.Vec.deref view.locator
+ let b8 ← transition.absent s3
+ if b8
then
- let b7 ← transition.has_link view.link
- if b7
- then transition.resolves_unknown view.link
+ let b9 ← transition.observed_stage view.stage
+ let b10 ← transition.has_reading view.reading
+ if b9 = b10
+ then
+ if b10
+ then
+ if view.observation
+ then
+ let b11 ← transition.provider_stage view.stage
+ if b11 = view.evidence
+ then
+ if view.evidence
+ then
+ let b12 ← transition.has_link view.link
+ if b12
+ then transition.resolves_unknown view.link
+ else ok false
+ else ok true
+ else ok false
+ else ok false
+ else
+ let b11 ← transition.provider_stage view.stage
+ if b11 = view.evidence
+ then
+ if view.evidence
+ then
+ let b12 ← transition.has_link view.link
+ if b12
+ then transition.resolves_unknown view.link
+ else ok false
+ else ok true
+ else ok false
else ok false
- else ok true
- else ok false
- else ok false
- else
- let b6 ← transition.provider_stage view.stage
- if b6 = view.evidence
- then
- if view.evidence
- then
- let b7 ← transition.has_link view.link
- if b7
- then transition.resolves_unknown view.link
else ok false
- else ok true
- else ok false
+ else ok false
else ok false
else ok false
+ else ok false
+ else ok false
else ok false
+/-- [auths_gateway_kernel::transition::same_protocol]:
+ Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 325:0-332:1
+ Visibility: public -/
+def transition.same_protocol
+ (left : transition.Protocol) (right : transition.Protocol) :
+ Result Bool
+ := do
+ match left with
+ | transition.Protocol.Mutation =>
+ match right with
+ | transition.Protocol.Mutation => ok true
+ | transition.Protocol.Acquisition => ok false
+ | transition.Protocol.Disclosure => ok false
+ | transition.Protocol.Acquisition =>
+ match right with
+ | transition.Protocol.Mutation => ok false
+ | transition.Protocol.Acquisition => ok true
+ | transition.Protocol.Disclosure => ok false
+ | transition.Protocol.Disclosure =>
+ match right with
+ | transition.Protocol.Mutation => ok false
+ | transition.Protocol.Acquisition => ok false
+ | transition.Protocol.Disclosure => ok true
+
/-- [auths_gateway_kernel::transition::stage_transition_allowed]:
- Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 235:0-260:1
+ Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 352:0-373:1
Visibility: public -/
def transition.stage_transition_allowed
(from1 : transition.Stage) (to1 : transition.Stage) (link : transition.Link)
@@ -2678,106 +2783,199 @@ def transition.stage_transition_allowed
Result Bool
:= do
match from1 with
- | transition.Stage.Attempting =>
+ | transition.Stage.Reserved =>
match to1 with
- | transition.Stage.Attempting => ok true
+ | transition.Stage.Reserved => ok true
+ | transition.Stage.EntryGranted => ok true
| transition.Stage.NotEntered => ok true
+ | transition.Stage.Unknown => ok false
+ | transition.Stage.ResponseRecorded => ok false
+ | transition.Stage.Observed => ok false
+ | transition.Stage.ObservedByProvider => ok false
+ | transition.Stage.Acknowledged => ok false
+ | transition.Stage.EntryGranted =>
+ match to1 with
+ | transition.Stage.Reserved => ok false
+ | transition.Stage.EntryGranted => ok false
+ | transition.Stage.NotEntered => ok false
| transition.Stage.Unknown => ok true
| transition.Stage.ResponseRecorded => ok true
| transition.Stage.Observed => ok false
| transition.Stage.ObservedByProvider => transition.resolves_unknown link
+ | transition.Stage.Acknowledged => ok true
| transition.Stage.NotEntered => ok false
| transition.Stage.Unknown =>
match to1 with
- | transition.Stage.Attempting => ok false
+ | transition.Stage.Reserved => ok false
+ | transition.Stage.EntryGranted => ok false
| transition.Stage.NotEntered => ok false
| transition.Stage.Unknown => ok false
- | transition.Stage.ResponseRecorded => ok false
+ | transition.Stage.ResponseRecorded => ok true
| transition.Stage.Observed => ok false
| transition.Stage.ObservedByProvider => transition.resolves_unknown link
+ | transition.Stage.Acknowledged => ok true
| transition.Stage.ResponseRecorded =>
match to1 with
- | transition.Stage.Attempting => ok false
+ | transition.Stage.Reserved => ok false
+ | transition.Stage.EntryGranted => ok false
| transition.Stage.NotEntered => ok false
| transition.Stage.Unknown => ok false
| transition.Stage.ResponseRecorded => ok false
| transition.Stage.Observed => ok true
| transition.Stage.ObservedByProvider => ok true
+ | transition.Stage.Acknowledged => ok false
| transition.Stage.Observed => ok false
| transition.Stage.ObservedByProvider => ok false
+ | transition.Stage.Acknowledged => ok false
/-- [auths_gateway_kernel::transition::adds_pre_entry]:
- Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 265:0-280:1
+ Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 377:0-379:1
Visibility: public -/
def transition.adds_pre_entry
(from1 : transition.Stage) (to1 : transition.Stage) : Result Bool := do
- match from1 with
- | transition.Stage.Attempting =>
+ let b ←
+ match from1 with
+ | transition.Stage.Reserved => ok true
+ | transition.Stage.EntryGranted => ok false
+ | transition.Stage.NotEntered => ok false
+ | transition.Stage.Unknown => ok false
+ | transition.Stage.ResponseRecorded => ok false
+ | transition.Stage.Observed => ok false
+ | transition.Stage.ObservedByProvider => ok false
+ | transition.Stage.Acknowledged => ok false
+ if b
+ then
match to1 with
- | transition.Stage.Attempting => ok true
+ | transition.Stage.Reserved => ok true
+ | transition.Stage.EntryGranted => ok false
| transition.Stage.NotEntered => ok true
| transition.Stage.Unknown => ok false
| transition.Stage.ResponseRecorded => ok false
| transition.Stage.Observed => ok false
| transition.Stage.ObservedByProvider => ok false
- | transition.Stage.NotEntered => ok false
- | transition.Stage.Unknown => ok false
- | transition.Stage.ResponseRecorded => ok false
- | transition.Stage.Observed => ok false
- | transition.Stage.ObservedByProvider => ok false
+ | transition.Stage.Acknowledged => ok false
+ else ok false
/-- [auths_gateway_kernel::transition::adds_response]:
- Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 284:0-300:1
+ Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 383:0-385:1
Visibility: public -/
def transition.adds_response
(from1 : transition.Stage) (to1 : transition.Stage) : Result Bool := do
- match from1 with
- | transition.Stage.Attempting =>
+ let b ←
+ match from1 with
+ | transition.Stage.Reserved => ok false
+ | transition.Stage.EntryGranted => ok true
+ | transition.Stage.NotEntered => ok false
+ | transition.Stage.Unknown => ok true
+ | transition.Stage.ResponseRecorded => ok false
+ | transition.Stage.Observed => ok false
+ | transition.Stage.ObservedByProvider => ok false
+ | transition.Stage.Acknowledged => ok false
+ if b
+ then
match to1 with
- | transition.Stage.Attempting => ok false
+ | transition.Stage.Reserved => ok false
+ | transition.Stage.EntryGranted => ok false
| transition.Stage.NotEntered => ok false
| transition.Stage.Unknown => ok false
| transition.Stage.ResponseRecorded => ok true
| transition.Stage.Observed => ok false
| transition.Stage.ObservedByProvider => ok false
- | transition.Stage.NotEntered => ok false
- | transition.Stage.Unknown => ok false
- | transition.Stage.ResponseRecorded => ok false
- | transition.Stage.Observed => ok false
- | transition.Stage.ObservedByProvider => ok false
+ | transition.Stage.Acknowledged => ok false
+ else ok false
/-- [auths_gateway_kernel::transition::checkpoint_valid]:
- Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 305:0-321:1
+ Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 389:0-395:1
Visibility: public -/
def transition.checkpoint_valid
(old : transition.AttemptView) (new : transition.AttemptView) :
Result Bool
:= do
- match old.stage with
- | transition.Stage.Attempting =>
- match new.stage with
- | transition.Stage.Attempting =>
+ let b ←
+ match old.stage with
+ | transition.Stage.Reserved => ok true
+ | transition.Stage.EntryGranted => ok false
+ | transition.Stage.NotEntered => ok false
+ | transition.Stage.Unknown => ok false
+ | transition.Stage.ResponseRecorded => ok false
+ | transition.Stage.Observed => ok false
+ | transition.Stage.ObservedByProvider => ok false
+ | transition.Stage.Acknowledged => ok false
+ if b
+ then
+ let b1 ←
+ match new.stage with
+ | transition.Stage.Reserved => ok true
+ | transition.Stage.EntryGranted => ok false
+ | transition.Stage.NotEntered => ok false
+ | transition.Stage.Unknown => ok false
+ | transition.Stage.ResponseRecorded => ok false
+ | transition.Stage.Observed => ok false
+ | transition.Stage.ObservedByProvider => ok false
+ | transition.Stage.Acknowledged => ok false
+ if b1
+ then
let s := alloc.vec.Vec.deref old.pre_entry
- let b ← transition.absent s
- if b
+ let b2 ← transition.absent s
+ if b2
then
let s1 := alloc.vec.Vec.deref new.pre_entry
- let b1 ← transition.absent s1
- ok (¬ b1)
+ let b3 ← transition.absent s1
+ ok (¬ b3)
else ok false
- | transition.Stage.NotEntered => ok true
- | transition.Stage.Unknown => ok true
- | transition.Stage.ResponseRecorded => ok true
- | transition.Stage.Observed => ok true
- | transition.Stage.ObservedByProvider => ok true
- | transition.Stage.NotEntered => ok true
- | transition.Stage.Unknown => ok true
- | transition.Stage.ResponseRecorded => ok true
- | transition.Stage.Observed => ok true
- | transition.Stage.ObservedByProvider => ok true
+ else ok true
+ else ok true
+
+/-- [auths_gateway_kernel::transition::entry_kept]:
+ Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 399:0-405:1
+ Visibility: public -/
+def transition.entry_kept
+ (old : transition.AttemptView) (new : transition.AttemptView) :
+ Result Bool
+ := do
+ let b ←
+ match old.stage with
+ | transition.Stage.Reserved => ok true
+ | transition.Stage.EntryGranted => ok false
+ | transition.Stage.NotEntered => ok false
+ | transition.Stage.Unknown => ok false
+ | transition.Stage.ResponseRecorded => ok false
+ | transition.Stage.Observed => ok false
+ | transition.Stage.ObservedByProvider => ok false
+ | transition.Stage.Acknowledged => ok false
+ if b
+ then
+ let b1 ←
+ match new.stage with
+ | transition.Stage.Reserved => ok false
+ | transition.Stage.EntryGranted => ok true
+ | transition.Stage.NotEntered => ok false
+ | transition.Stage.Unknown => ok false
+ | transition.Stage.ResponseRecorded => ok false
+ | transition.Stage.Observed => ok false
+ | transition.Stage.ObservedByProvider => ok false
+ | transition.Stage.Acknowledged => ok false
+ if b1
+ then
+ let s := alloc.vec.Vec.deref old.entry
+ let b2 ← transition.absent s
+ if b2
+ then
+ let s1 := alloc.vec.Vec.deref new.entry
+ let b3 ← transition.absent s1
+ ok (¬ b3)
+ else ok false
+ else
+ let s := alloc.vec.Vec.deref old.entry
+ let s1 := alloc.vec.Vec.deref new.entry
+ transition.same_bytes s s1
+ else
+ let s := alloc.vec.Vec.deref old.entry
+ let s1 := alloc.vec.Vec.deref new.entry
+ transition.same_bytes s s1
/-- [auths_gateway_kernel::transition::pre_entry_kept]:
- Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 326:0-332:1
+ Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 410:0-416:1
Visibility: public -/
def transition.pre_entry_kept
(old : transition.AttemptView) (new : transition.AttemptView) :
@@ -2798,7 +2996,7 @@ def transition.pre_entry_kept
transition.same_bytes s1 s2
/-- [auths_gateway_kernel::transition::response_kept]:
- Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 337:0-340:1
+ Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 421:0-424:1
Visibility: public -/
def transition.response_kept
(old : transition.AttemptView) (new : transition.AttemptView) :
@@ -2818,8 +3016,47 @@ def transition.response_kept
transition.same_bytes s2 s3
else ok false
+/-- [auths_gateway_kernel::transition::acknowledgement_kept]:
+ Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 429:0-436:1
+ Visibility: public -/
+def transition.acknowledgement_kept
+ (old : transition.AttemptView) (new : transition.AttemptView) :
+ Result Bool
+ := do
+ let b ←
+ match old.stage with
+ | transition.Stage.Reserved => ok false
+ | transition.Stage.EntryGranted => ok true
+ | transition.Stage.NotEntered => ok false
+ | transition.Stage.Unknown => ok true
+ | transition.Stage.ResponseRecorded => ok false
+ | transition.Stage.Observed => ok false
+ | transition.Stage.ObservedByProvider => ok false
+ | transition.Stage.Acknowledged => ok false
+ if b
+ then
+ let b1 ← transition.acknowledgement_stage new.stage
+ if b1
+ then
+ let s := alloc.vec.Vec.deref old.acknowledgement
+ let b2 ← transition.absent s
+ if b2
+ then
+ let s1 := alloc.vec.Vec.deref new.acknowledgement
+ let b3 ← transition.absent s1
+ ok (¬ b3)
+ else ok false
+ else
+ let s := alloc.vec.Vec.deref old.acknowledgement
+ let s1 := alloc.vec.Vec.deref new.acknowledgement
+ transition.same_bytes s s1
+ else
+ let s := alloc.vec.Vec.deref old.acknowledgement
+ let s1 := alloc.vec.Vec.deref new.acknowledgement
+ transition.same_bytes s s1
+
/-- [auths_gateway_kernel::transition::link_equal]:
- Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 350:0-365:1
+ Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 459:0-474:1
Visibility: public -/
def transition.link_equal
(left : transition.Link) (right : transition.Link) : Result Bool := do
@@ -2841,7 +3078,7 @@ def transition.link_equal
| transition.Link.AfterResponse => ok true
/-- [auths_gateway_kernel::transition::plan_kept]:
- Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 344:0-346:1
+ Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 453:0-455:1
Visibility: public -/
def transition.plan_kept
(old : transition.AttemptView) (new : transition.AttemptView) :
@@ -2852,33 +3089,57 @@ def transition.plan_kept
else ok false
/-- [auths_gateway_kernel::transition::valid_transition]:
- Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 373:0-381:1
+ Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 482:0-496:1
Visibility: public -/
def transition.valid_transition
(old : transition.AttemptView) (new : transition.AttemptView) :
Result Bool
:= do
- let s := alloc.vec.Vec.deref old.fixed
- let s1 := alloc.vec.Vec.deref new.fixed
- let b ← transition.same_bytes s s1
+ let b ← transition.same_protocol old.protocol new.protocol
if b
then
- let b1 ← transition.plan_kept old new
- if b1
+ if old.revision < 9007199254740991#u64
then
- let b2 ←
- transition.stage_transition_allowed old.stage new.stage old.link
- if b2
+ let i ← old.revision + 1#u64
+ if new.revision = i
then
- let b3 ← transition.checkpoint_valid old new
- if b3
+ let b1 ← transition.consistent old
+ if b1
then
- let b4 ← transition.pre_entry_kept old new
- if b4
+ let b2 ← transition.entry_kept old new
+ if b2
then
- let b5 ← transition.response_kept old new
- if b5
- then transition.consistent new
+ let s := alloc.vec.Vec.deref old.fixed
+ let s1 := alloc.vec.Vec.deref new.fixed
+ let b3 ← transition.same_bytes s s1
+ if b3
+ then
+ let b4 ← transition.plan_kept old new
+ if b4
+ then
+ let b5 ←
+ transition.stage_transition_allowed old.stage new.stage
+ old.link
+ if b5
+ then
+ let b6 ← transition.checkpoint_valid old new
+ if b6
+ then
+ let b7 ← transition.pre_entry_kept old new
+ if b7
+ then
+ let b8 ← transition.response_kept old new
+ if b8
+ then
+ let b9 ← transition.acknowledgement_kept old new
+ if b9
+ then transition.consistent new
+ else ok false
+ else ok false
+ else ok false
+ else ok false
+ else ok false
+ else ok false
else ok false
else ok false
else ok false
diff --git a/formal/qualification/aeneas/generated/gateway/Types.lean b/formal/qualification/aeneas/generated/gateway/Types.lean
index 2782c58c8..38767b75d 100644
--- a/formal/qualification/aeneas/generated/gateway/Types.lean
+++ b/formal/qualification/aeneas/generated/gateway/Types.lean
@@ -147,17 +147,8 @@ inductive construct.ConstructError where
| HeaderValue : construct.ConstructError
| BodySize : construct.ConstructError
-/-- [auths_gateway_kernel::order::Mode]
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 34:0-41:1
- Visibility: public -/
-@[discriminant isize]
-inductive order.Mode where
-| Entry : order.Mode
-| ReadBack : order.Mode
-| Reobserve : order.Mode
-
/-- [auths_gateway_kernel::order::SubmitPlan]
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 50:0-63:1
+ Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 39:0-52:1
Visibility: public -/
structure order.SubmitPlan where
account_scope : Bool
@@ -168,7 +159,7 @@ structure order.SubmitPlan where
basis_points : Std.U16
/-- [auths_gateway_kernel::order::Phase]
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 67:0-116:1
+ Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 56:0-103:1
Visibility: public -/
@[discriminant isize]
inductive order.Phase where
@@ -179,26 +170,25 @@ inductive order.Phase where
| Scope : order.Phase
| Prepare : order.Phase
| Claim : order.Phase
-| Resume : order.Phase
-| Reload : order.Mode → order.Phase
-| Lease : order.Mode → order.Phase
-| Prefix : order.Mode → order.Phase
-| Account : order.Mode → order.Phase
-| Denied : order.Mode → Std.U8 → order.Phase
+| Reload : order.Phase
+| Lease : order.Phase
+| Prefix : order.Phase
+| Account : order.Phase
+| Denied : Std.U8 → order.Phase
| PreEntry : order.Phase
| Ceiling : order.Phase
| Checkpoint : order.Phase
| EntryReload : order.Phase
| Deadline : order.Phase
+| Grant : order.Phase
| Send : order.Phase
| RecordResponse : order.Phase
| RecordUnknown : order.Phase
| RecordNotEntered : order.Phase
-| ReadBack : order.Mode → order.Phase
| Done : order.Phase
/-- [auths_gateway_kernel::order::SubmitState]
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 120:0-128:1
+ Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 107:0-115:1
Visibility: public -/
structure order.SubmitState where
plan : order.SubmitPlan
@@ -206,7 +196,7 @@ structure order.SubmitState where
argument : Std.U64
/-- [auths_gateway_kernel::order::Verification]
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 132:0-141:1
+ Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 119:0-128:1
Visibility: public -/
@[discriminant isize]
inductive order.Verification where
@@ -214,17 +204,17 @@ inductive order.Verification where
| Authorized : Std.U64 → order.Verification
/-- [auths_gateway_kernel::order::ClaimResult]
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 145:0-154:1
+ Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 132:0-141:1
Visibility: public -/
@[discriminant isize]
inductive order.ClaimResult where
-| Inserted : order.ClaimResult
+| Reserved : order.ClaimResult
| Refused : order.ClaimResult
| Replay : order.ClaimResult
| Unavailable : order.ClaimResult
/-- [auths_gateway_kernel::order::AccountResult]
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 158:0-165:1
+ Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 145:0-152:1
Visibility: public -/
@[discriminant isize]
inductive order.AccountResult where
@@ -233,7 +223,7 @@ inductive order.AccountResult where
| Unavailable : order.AccountResult
/-- [auths_gateway_kernel::order::DeniedResult]
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 169:0-176:1
+ Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 156:0-163:1
Visibility: public -/
@[discriminant isize]
inductive order.DeniedResult where
@@ -242,7 +232,7 @@ inductive order.DeniedResult where
| Unavailable : order.DeniedResult
/-- [auths_gateway_kernel::order::PreEntryResult]
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 180:0-187:1
+ Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 167:0-174:1
Visibility: public -/
@[discriminant isize]
inductive order.PreEntryResult where
@@ -251,7 +241,7 @@ inductive order.PreEntryResult where
| Unavailable : order.PreEntryResult
/-- [auths_gateway_kernel::order::CeilingRead]
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 191:0-201:1
+ Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 178:0-188:1
Visibility: public -/
@[discriminant isize]
inductive order.CeilingRead where
@@ -259,24 +249,23 @@ inductive order.CeilingRead where
| Read : Std.U64 → Bool → order.CeilingRead
/-- [auths_gateway_kernel::order::WriteResult]
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 205:0-212:1
+ Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 192:0-197:1
Visibility: public -/
@[discriminant isize]
inductive order.WriteResult where
-| NotEntered : order.WriteResult
| Unknown : order.WriteResult
| Response : order.WriteResult
/-- [auths_gateway_kernel::order::ResponseRecord]
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 216:0-224:1
+ Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 201:0-206:1
Visibility: public -/
@[discriminant isize]
inductive order.ResponseRecord where
| Failed : order.ResponseRecord
-| Recorded : Bool → order.ResponseRecord
+| Recorded : order.ResponseRecord
/-- [auths_gateway_kernel::order::SubmitEvent]
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 228:0-267:1
+ Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 210:0-249:1
Visibility: public -/
@[discriminant isize]
inductive order.SubmitEvent where
@@ -287,7 +276,6 @@ inductive order.SubmitEvent where
| Scope : Bool → order.SubmitEvent
| Preparation : Bool → order.SubmitEvent
| Claim : order.ClaimResult → order.SubmitEvent
-| Resume : Bool → order.SubmitEvent
| Reload : Bool → order.SubmitEvent
| Lease : Bool → order.SubmitEvent
| Prefix : Bool → order.SubmitEvent
@@ -297,11 +285,12 @@ inductive order.SubmitEvent where
| Ceiling : order.CeilingRead → order.SubmitEvent
| Recorded : Bool → order.SubmitEvent
| Deadline : Bool → order.SubmitEvent
+| Granted : Bool → order.SubmitEvent
| Write : order.WriteResult → order.SubmitEvent
| Response : order.ResponseRecord → order.SubmitEvent
/-- [auths_gateway_kernel::order::Refusal]
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 271:0-300:1
+ Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 253:0-282:1
Visibility: public -/
@[discriminant isize]
inductive order.Refusal where
@@ -321,7 +310,7 @@ inductive order.Refusal where
| TransportNotEntered : order.Refusal
/-- [auths_gateway_kernel::order::Stop]
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 304:0-323:1
+ Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 286:0-305:1
Visibility: public -/
@[discriminant isize]
inductive order.Stop where
@@ -331,12 +320,12 @@ inductive order.Stop where
| NotEntered : order.Stop
| Unknown : order.Stop
| Response : order.Stop
-| Observed : order.Stop
-| ReplayRefused : order.Stop
+| Stored : order.Stop
+| GrantUnavailable : order.Stop
| Halted : order.Stop
/-- [auths_gateway_kernel::order::SubmitAction]
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 327:0-374:1
+ Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 309:0-354:1
Visibility: public -/
@[discriminant isize]
inductive order.SubmitAction where
@@ -346,26 +335,25 @@ inductive order.SubmitAction where
| BindScope : order.SubmitAction
| Prepare : order.SubmitAction
| Claim : order.SubmitAction
-| Resume : order.SubmitAction
-| Reload : order.Mode → order.SubmitAction
-| Lease : order.Mode → order.SubmitAction
-| CheckPrefix : order.Mode → order.SubmitAction
-| ReadAccount : order.Mode → order.SubmitAction
-| DeniedRead : order.Mode → Std.U8 → order.SubmitAction
+| Reload : order.SubmitAction
+| Lease : order.SubmitAction
+| CheckPrefix : order.SubmitAction
+| ReadAccount : order.SubmitAction
+| DeniedRead : Std.U8 → order.SubmitAction
| PreEntryRead : order.SubmitAction
| CeilingRead : order.SubmitAction
| RecordCheckpoint : order.SubmitAction
| ReloadBeforeEntry : order.SubmitAction
| CheckDeadline : order.SubmitAction
+| GrantEntry : order.SubmitAction
| Send : order.SubmitAction
| RecordResponse : order.SubmitAction
| RecordUnknown : order.SubmitAction
| RecordNotEntered : order.Refusal → order.SubmitAction
-| ReadBack : order.Mode → order.SubmitAction
| Stop : order.Stop → order.SubmitAction
/-- [auths_gateway_kernel::order::SubmitDecision]
- Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 378:0-383:1
+ Source: 'product/runtime/auths-gateway-kernel/src/order.rs', lines 358:0-363:1
Visibility: public -/
structure order.SubmitDecision where
state : order.SubmitState
@@ -435,20 +423,31 @@ structure recovery.RecoveryCapability where
pre_entry_reread : Bool
write_is_conditional : Bool
+/-- [auths_gateway_kernel::transition::Protocol]
+ Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 19:0-26:1
+ Visibility: public -/
+@[discriminant isize]
+inductive transition.Protocol where
+| Mutation : transition.Protocol
+| Acquisition : transition.Protocol
+| Disclosure : transition.Protocol
+
/-- [auths_gateway_kernel::transition::Stage]
- Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 19:0-32:1
+ Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 30:0-47:1
Visibility: public -/
@[discriminant isize]
inductive transition.Stage where
-| Attempting : transition.Stage
+| Reserved : transition.Stage
+| EntryGranted : transition.Stage
| NotEntered : transition.Stage
| Unknown : transition.Stage
| ResponseRecorded : transition.Stage
| Observed : transition.Stage
| ObservedByProvider : transition.Stage
+| Acknowledged : transition.Stage
/-- [auths_gateway_kernel::transition::Reading]
- Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 36:0-45:1
+ Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 51:0-60:1
Visibility: public -/
@[discriminant isize]
inductive transition.Reading where
@@ -458,7 +457,7 @@ inductive transition.Reading where
| EchoMismatch : transition.Reading
/-- [auths_gateway_kernel::transition::Link]
- Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 49:0-56:1
+ Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 64:0-71:1
Visibility: public -/
@[discriminant isize]
inductive transition.Link where
@@ -467,14 +466,18 @@ inductive transition.Link where
| AfterResponse : transition.Link
/-- [auths_gateway_kernel::transition::AttemptView]
- Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 60:0-81:1
+ Source: 'product/runtime/auths-gateway-kernel/src/transition.rs', lines 75:0-105:1
Visibility: public -/
structure transition.AttemptView where
+ protocol : transition.Protocol
+ revision : Std.U64
fixed : alloc.vec.Vec Std.U8
+ entry : alloc.vec.Vec Std.U8
stage : transition.Stage
refusal : Bool
pre_entry : alloc.vec.Vec Std.U8
response : alloc.vec.Vec Std.U8
+ acknowledgement : alloc.vec.Vec Std.U8
locator : alloc.vec.Vec Std.U8
reading : transition.Reading
evidence : Bool
diff --git a/formal/qualification/aeneas/generated/gateway/translation.json b/formal/qualification/aeneas/generated/gateway/translation.json
index fff14d276..2a69a139d 100644
--- a/formal/qualification/aeneas/generated/gateway/translation.json
+++ b/formal/qualification/aeneas/generated/gateway/translation.json
@@ -4,7 +4,7 @@
"crate": "auths_gateway_kernel",
"functions": [
{
- "def_id": 37,
+ "def_id": 40,
"lean_name": "auths_gateway_kernel.construct.append_bytes_loop.body",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::append_bytes",
@@ -29,7 +29,7 @@
"parent_lean_name": "auths_gateway_kernel.construct.append_bytes"
},
{
- "def_id": 37,
+ "def_id": 40,
"lean_name": "auths_gateway_kernel.construct.append_bytes_loop",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::append_bytes",
@@ -54,7 +54,7 @@
"parent_lean_name": "auths_gateway_kernel.construct.append_bytes"
},
{
- "def_id": 37,
+ "def_id": 40,
"lean_name": "auths_gateway_kernel.construct.append_bytes",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::append_bytes",
@@ -88,7 +88,7 @@
"reducible": false
},
{
- "def_id": 113,
+ "def_id": 116,
"lean_name": "auths_gateway_kernel.construct.is_alphanumeric",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::is_alphanumeric",
@@ -105,7 +105,7 @@
"reducible": false
},
{
- "def_id": 109,
+ "def_id": 112,
"lean_name": "auths_gateway_kernel.construct.is_path_literal",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::is_path_literal",
@@ -122,7 +122,7 @@
"reducible": false
},
{
- "def_id": 111,
+ "def_id": 114,
"lean_name": "auths_gateway_kernel.construct.is_form_literal",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::is_form_literal",
@@ -139,7 +139,7 @@
"reducible": false
},
{
- "def_id": 114,
+ "def_id": 117,
"lean_name": "auths_gateway_kernel.construct.hex_upper",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::hex_upper",
@@ -156,7 +156,7 @@
"reducible": false
},
{
- "def_id": 115,
+ "def_id": 118,
"lean_name": "auths_gateway_kernel.construct.hex_lower",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::hex_lower",
@@ -173,7 +173,7 @@
"reducible": false
},
{
- "def_id": 110,
+ "def_id": 113,
"lean_name": "auths_gateway_kernel.construct.append_percent",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::append_percent",
@@ -190,7 +190,7 @@
"reducible": false
},
{
- "def_id": 103,
+ "def_id": 107,
"lean_name": "auths_gateway_kernel.construct.append_path_byte",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::append_path_byte",
@@ -207,7 +207,7 @@
"reducible": false
},
{
- "def_id": 96,
+ "def_id": 100,
"lean_name": "auths_gateway_kernel.construct.append_path_encoded_loop.body",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::append_path_encoded",
@@ -232,7 +232,7 @@
"parent_lean_name": "auths_gateway_kernel.construct.append_path_encoded"
},
{
- "def_id": 96,
+ "def_id": 100,
"lean_name": "auths_gateway_kernel.construct.append_path_encoded_loop",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::append_path_encoded",
@@ -257,7 +257,7 @@
"parent_lean_name": "auths_gateway_kernel.construct.append_path_encoded"
},
{
- "def_id": 96,
+ "def_id": 100,
"lean_name": "auths_gateway_kernel.construct.append_path_encoded",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::append_path_encoded",
@@ -274,7 +274,7 @@
"reducible": true
},
{
- "def_id": 95,
+ "def_id": 99,
"lean_name": "auths_gateway_kernel.construct.segment_value_valid",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::segment_value_valid",
@@ -291,7 +291,7 @@
"reducible": false
},
{
- "def_id": 82,
+ "def_id": 86,
"lean_name": "auths_gateway_kernel.construct.segment_error",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::segment_error",
@@ -308,7 +308,7 @@
"reducible": false
},
{
- "def_id": 34,
+ "def_id": 37,
"lean_name": "auths_gateway_kernel.construct.path_error_loop.body",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::path_error",
@@ -333,7 +333,7 @@
"parent_lean_name": "auths_gateway_kernel.construct.path_error"
},
{
- "def_id": 34,
+ "def_id": 37,
"lean_name": "auths_gateway_kernel.construct.path_error_loop",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::path_error",
@@ -358,7 +358,7 @@
"parent_lean_name": "auths_gateway_kernel.construct.path_error"
},
{
- "def_id": 34,
+ "def_id": 37,
"lean_name": "auths_gateway_kernel.construct.path_error",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::path_error",
@@ -375,7 +375,7 @@
"reducible": true
},
{
- "def_id": 84,
+ "def_id": 88,
"lean_name": "auths_gateway_kernel.construct.append_segment",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::append_segment",
@@ -392,7 +392,7 @@
"reducible": false
},
{
- "def_id": 35,
+ "def_id": 38,
"lean_name": "auths_gateway_kernel.construct.append_segments_loop.body",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::append_segments",
@@ -417,7 +417,7 @@
"parent_lean_name": "auths_gateway_kernel.construct.append_segments"
},
{
- "def_id": 35,
+ "def_id": 38,
"lean_name": "auths_gateway_kernel.construct.append_segments_loop",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::append_segments",
@@ -442,7 +442,7 @@
"parent_lean_name": "auths_gateway_kernel.construct.append_segments"
},
{
- "def_id": 35,
+ "def_id": 38,
"lean_name": "auths_gateway_kernel.construct.append_segments",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::append_segments",
@@ -543,7 +543,7 @@
"reducible": false
},
{
- "def_id": 52,
+ "def_id": 55,
"lean_name": "auths_gateway_kernel.construct.copy_header",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::copy_header",
@@ -627,7 +627,7 @@
"reducible": true
},
{
- "def_id": 106,
+ "def_id": 110,
"lean_name": "auths_gateway_kernel.construct.is_account_byte",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::is_account_byte",
@@ -644,7 +644,7 @@
"reducible": false
},
{
- "def_id": 100,
+ "def_id": 104,
"lean_name": "auths_gateway_kernel.construct.account_bytes_valid_loop.body",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::account_bytes_valid",
@@ -669,7 +669,7 @@
"parent_lean_name": "auths_gateway_kernel.construct.account_bytes_valid"
},
{
- "def_id": 100,
+ "def_id": 104,
"lean_name": "auths_gateway_kernel.construct.account_bytes_valid_loop",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::account_bytes_valid",
@@ -694,7 +694,7 @@
"parent_lean_name": "auths_gateway_kernel.construct.account_bytes_valid"
},
{
- "def_id": 100,
+ "def_id": 104,
"lean_name": "auths_gateway_kernel.construct.account_bytes_valid",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::account_bytes_valid",
@@ -711,7 +711,7 @@
"reducible": true
},
{
- "def_id": 91,
+ "def_id": 95,
"lean_name": "auths_gateway_kernel.construct.scope_value_valid",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::scope_value_valid",
@@ -728,7 +728,7 @@
"reducible": false
},
{
- "def_id": 47,
+ "def_id": 50,
"lean_name": "auths_gateway_kernel.construct.account_scope_header",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::account_scope_header",
@@ -762,7 +762,7 @@
"reducible": false
},
{
- "def_id": 112,
+ "def_id": 115,
"lean_name": "auths_gateway_kernel.construct.append_json_byte",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::append_json_byte",
@@ -779,7 +779,7 @@
"reducible": false
},
{
- "def_id": 105,
+ "def_id": 109,
"lean_name": "auths_gateway_kernel.construct.append_json_escaped_loop.body",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::append_json_escaped",
@@ -804,7 +804,7 @@
"parent_lean_name": "auths_gateway_kernel.construct.append_json_escaped"
},
{
- "def_id": 105,
+ "def_id": 109,
"lean_name": "auths_gateway_kernel.construct.append_json_escaped_loop",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::append_json_escaped",
@@ -829,7 +829,7 @@
"parent_lean_name": "auths_gateway_kernel.construct.append_json_escaped"
},
{
- "def_id": 105,
+ "def_id": 109,
"lean_name": "auths_gateway_kernel.construct.append_json_escaped",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::append_json_escaped",
@@ -846,7 +846,7 @@
"reducible": true
},
{
- "def_id": 98,
+ "def_id": 102,
"lean_name": "auths_gateway_kernel.construct.append_json_string",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::append_json_string",
@@ -863,7 +863,7 @@
"reducible": false
},
{
- "def_id": 99,
+ "def_id": 103,
"lean_name": "auths_gateway_kernel.construct.append_json_argument",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::append_json_argument",
@@ -880,7 +880,7 @@
"reducible": false
},
{
- "def_id": 87,
+ "def_id": 91,
"lean_name": "auths_gateway_kernel.construct.json_piece_valid",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::json_piece_valid",
@@ -897,7 +897,7 @@
"reducible": false
},
{
- "def_id": 44,
+ "def_id": 47,
"lean_name": "auths_gateway_kernel.construct.json_pieces_valid_loop.body",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::json_pieces_valid",
@@ -922,7 +922,7 @@
"parent_lean_name": "auths_gateway_kernel.construct.json_pieces_valid"
},
{
- "def_id": 44,
+ "def_id": 47,
"lean_name": "auths_gateway_kernel.construct.json_pieces_valid_loop",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::json_pieces_valid",
@@ -947,7 +947,7 @@
"parent_lean_name": "auths_gateway_kernel.construct.json_pieces_valid"
},
{
- "def_id": 44,
+ "def_id": 47,
"lean_name": "auths_gateway_kernel.construct.json_pieces_valid",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::json_pieces_valid",
@@ -964,7 +964,7 @@
"reducible": true
},
{
- "def_id": 89,
+ "def_id": 93,
"lean_name": "auths_gateway_kernel.construct.append_json_piece",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::append_json_piece",
@@ -981,7 +981,7 @@
"reducible": false
},
{
- "def_id": 46,
+ "def_id": 49,
"lean_name": "auths_gateway_kernel.construct.append_json_pieces_loop.body",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::append_json_pieces",
@@ -1006,7 +1006,7 @@
"parent_lean_name": "auths_gateway_kernel.construct.append_json_pieces"
},
{
- "def_id": 46,
+ "def_id": 49,
"lean_name": "auths_gateway_kernel.construct.append_json_pieces_loop",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::append_json_pieces",
@@ -1031,7 +1031,7 @@
"parent_lean_name": "auths_gateway_kernel.construct.append_json_pieces"
},
{
- "def_id": 46,
+ "def_id": 49,
"lean_name": "auths_gateway_kernel.construct.append_json_pieces",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::append_json_pieces",
@@ -1048,7 +1048,7 @@
"reducible": true
},
{
- "def_id": 104,
+ "def_id": 108,
"lean_name": "auths_gateway_kernel.construct.append_form_byte",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::append_form_byte",
@@ -1065,7 +1065,7 @@
"reducible": false
},
{
- "def_id": 97,
+ "def_id": 101,
"lean_name": "auths_gateway_kernel.construct.append_form_encoded_loop.body",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::append_form_encoded",
@@ -1090,7 +1090,7 @@
"parent_lean_name": "auths_gateway_kernel.construct.append_form_encoded"
},
{
- "def_id": 97,
+ "def_id": 101,
"lean_name": "auths_gateway_kernel.construct.append_form_encoded_loop",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::append_form_encoded",
@@ -1115,7 +1115,7 @@
"parent_lean_name": "auths_gateway_kernel.construct.append_form_encoded"
},
{
- "def_id": 97,
+ "def_id": 101,
"lean_name": "auths_gateway_kernel.construct.append_form_encoded",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::append_form_encoded",
@@ -1132,7 +1132,7 @@
"reducible": true
},
{
- "def_id": 86,
+ "def_id": 90,
"lean_name": "auths_gateway_kernel.construct.form_piece_valid",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::form_piece_valid",
@@ -1149,7 +1149,7 @@
"reducible": false
},
{
- "def_id": 43,
+ "def_id": 46,
"lean_name": "auths_gateway_kernel.construct.form_pieces_valid_loop.body",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::form_pieces_valid",
@@ -1174,7 +1174,7 @@
"parent_lean_name": "auths_gateway_kernel.construct.form_pieces_valid"
},
{
- "def_id": 43,
+ "def_id": 46,
"lean_name": "auths_gateway_kernel.construct.form_pieces_valid_loop",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::form_pieces_valid",
@@ -1199,7 +1199,7 @@
"parent_lean_name": "auths_gateway_kernel.construct.form_pieces_valid"
},
{
- "def_id": 43,
+ "def_id": 46,
"lean_name": "auths_gateway_kernel.construct.form_pieces_valid",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::form_pieces_valid",
@@ -1216,7 +1216,7 @@
"reducible": true
},
{
- "def_id": 88,
+ "def_id": 92,
"lean_name": "auths_gateway_kernel.construct.append_form_piece",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::append_form_piece",
@@ -1233,7 +1233,7 @@
"reducible": false
},
{
- "def_id": 45,
+ "def_id": 48,
"lean_name": "auths_gateway_kernel.construct.append_form_pieces_loop.body",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::append_form_pieces",
@@ -1258,7 +1258,7 @@
"parent_lean_name": "auths_gateway_kernel.construct.append_form_pieces"
},
{
- "def_id": 45,
+ "def_id": 48,
"lean_name": "auths_gateway_kernel.construct.append_form_pieces_loop",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::append_form_pieces",
@@ -1283,7 +1283,7 @@
"parent_lean_name": "auths_gateway_kernel.construct.append_form_pieces"
},
{
- "def_id": 45,
+ "def_id": 48,
"lean_name": "auths_gateway_kernel.construct.append_form_pieces",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::construct::append_form_pieces",
@@ -1375,8 +1375,8 @@
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 387,
- "end_line": 393
+ "begin_line": 367,
+ "end_line": 373
},
"is_opaque": false,
"can_fail": true,
@@ -1385,15 +1385,15 @@
"reducible": false
},
{
- "def_id": 54,
+ "def_id": 57,
"lean_name": "auths_gateway_kernel.order.go",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::order::go",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 397,
- "end_line": 406
+ "begin_line": 377,
+ "end_line": 386
},
"is_opaque": false,
"can_fail": true,
@@ -1402,15 +1402,15 @@
"reducible": false
},
{
- "def_id": 53,
+ "def_id": 56,
"lean_name": "auths_gateway_kernel.order.stop",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::order::stop",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 410,
- "end_line": 412
+ "begin_line": 390,
+ "end_line": 392
},
"is_opaque": false,
"can_fail": true,
@@ -1426,8 +1426,8 @@
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 416,
- "end_line": 418
+ "begin_line": 396,
+ "end_line": 398
},
"is_opaque": false,
"can_fail": true,
@@ -1436,32 +1436,15 @@
"reducible": false
},
{
- "def_id": 67,
+ "def_id": 63,
"lean_name": "auths_gateway_kernel.order.not_entered",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::order::not_entered",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 422,
- "end_line": 428
- },
- "is_opaque": false,
- "can_fail": true,
- "can_diverge": false,
- "is_rec": false,
- "reducible": false
- },
- {
- "def_id": 60,
- "lean_name": "auths_gateway_kernel.order.lease_failed",
- "lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
- "rust_name": "auths_gateway_kernel::order::lease_failed",
- "is_local": true,
- "source": {
- "file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 433,
- "end_line": 439
+ "begin_line": 402,
+ "end_line": 408
},
"is_opaque": false,
"can_fail": true,
@@ -1470,15 +1453,15 @@
"reducible": false
},
{
- "def_id": 94,
+ "def_id": 98,
"lean_name": "auths_gateway_kernel.order.after_reads",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::order::after_reads",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 445,
- "end_line": 451
+ "begin_line": 414,
+ "end_line": 420
},
"is_opaque": false,
"can_fail": true,
@@ -1487,15 +1470,15 @@
"reducible": false
},
{
- "def_id": 93,
+ "def_id": 97,
"lean_name": "auths_gateway_kernel.order.ceiling_from",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::order::ceiling_from",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 455,
- "end_line": 461
+ "begin_line": 424,
+ "end_line": 430
},
"is_opaque": false,
"can_fail": true,
@@ -1504,32 +1487,15 @@
"reducible": false
},
{
- "def_id": 107,
+ "def_id": 105,
"lean_name": "auths_gateway_kernel.order.pre_entry_from",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::order::pre_entry_from",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 465,
- "end_line": 471
- },
- "is_opaque": false,
- "can_fail": true,
- "can_diverge": false,
- "is_rec": false,
- "reducible": false
- },
- {
- "def_id": 101,
- "lean_name": "auths_gateway_kernel.order.after_credentials",
- "lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
- "rust_name": "auths_gateway_kernel::order::after_credentials",
- "is_local": true,
- "source": {
- "file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 476,
- "end_line": 483
+ "begin_line": 434,
+ "end_line": 440
},
"is_opaque": false,
"can_fail": true,
@@ -1538,15 +1504,15 @@
"reducible": false
},
{
- "def_id": 92,
+ "def_id": 96,
"lean_name": "auths_gateway_kernel.order.denied_from",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::order::denied_from",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 487,
- "end_line": 497
+ "begin_line": 444,
+ "end_line": 450
},
"is_opaque": false,
"can_fail": true,
@@ -1555,15 +1521,15 @@
"reducible": false
},
{
- "def_id": 61,
+ "def_id": 64,
"lean_name": "auths_gateway_kernel.order.after_prefix",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::order::after_prefix",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 501,
- "end_line": 507
+ "begin_line": 454,
+ "end_line": 460
},
"is_opaque": false,
"can_fail": true,
@@ -1572,15 +1538,15 @@
"reducible": false
},
{
- "def_id": 55,
+ "def_id": 58,
"lean_name": "auths_gateway_kernel.order.after_admission",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::order::after_admission",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 511,
- "end_line": 517
+ "begin_line": 464,
+ "end_line": 470
},
"is_opaque": false,
"can_fail": true,
@@ -1589,15 +1555,15 @@
"reducible": false
},
{
- "def_id": 57,
+ "def_id": 60,
"lean_name": "auths_gateway_kernel.order.pre_claim",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::order::pre_claim",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 521,
- "end_line": 527
+ "begin_line": 474,
+ "end_line": 480
},
"is_opaque": false,
"can_fail": true,
@@ -1606,15 +1572,15 @@
"reducible": false
},
{
- "def_id": 56,
+ "def_id": 59,
"lean_name": "auths_gateway_kernel.order.after_claim",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::order::after_claim",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 531,
- "end_line": 542
+ "begin_line": 484,
+ "end_line": 491
},
"is_opaque": false,
"can_fail": true,
@@ -1623,15 +1589,15 @@
"reducible": false
},
{
- "def_id": 58,
+ "def_id": 61,
"lean_name": "auths_gateway_kernel.order.after_account",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::order::after_account",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 546,
- "end_line": 556
+ "begin_line": 495,
+ "end_line": 501
},
"is_opaque": false,
"can_fail": true,
@@ -1640,15 +1606,15 @@
"reducible": false
},
{
- "def_id": 59,
+ "def_id": 62,
"lean_name": "auths_gateway_kernel.order.after_denied",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::order::after_denied",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 560,
- "end_line": 577
+ "begin_line": 505,
+ "end_line": 517
},
"is_opaque": false,
"can_fail": true,
@@ -1657,15 +1623,15 @@
"reducible": false
},
{
- "def_id": 62,
+ "def_id": 65,
"lean_name": "auths_gateway_kernel.order.after_pre_entry",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::order::after_pre_entry",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 581,
- "end_line": 587
+ "begin_line": 521,
+ "end_line": 527
},
"is_opaque": false,
"can_fail": true,
@@ -1691,15 +1657,15 @@
"reducible": false
},
{
- "def_id": 63,
+ "def_id": 66,
"lean_name": "auths_gateway_kernel.order.after_ceiling",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::order::after_ceiling",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 592,
- "end_line": 605
+ "begin_line": 532,
+ "end_line": 545
},
"is_opaque": false,
"can_fail": true,
@@ -1708,15 +1674,15 @@
"reducible": false
},
{
- "def_id": 64,
+ "def_id": 67,
"lean_name": "auths_gateway_kernel.order.after_write",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::order::after_write",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 609,
- "end_line": 615
+ "begin_line": 549,
+ "end_line": 554
},
"is_opaque": false,
"can_fail": true,
@@ -1725,32 +1691,15 @@
"reducible": false
},
{
- "def_id": 65,
+ "def_id": 68,
"lean_name": "auths_gateway_kernel.order.after_response",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::order::after_response",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 619,
- "end_line": 634
- },
- "is_opaque": false,
- "can_fail": true,
- "can_diverge": false,
- "is_rec": false,
- "reducible": false
- },
- {
- "def_id": 66,
- "lean_name": "auths_gateway_kernel.order.after_read_back",
- "lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
- "rust_name": "auths_gateway_kernel::order::after_read_back",
- "is_local": true,
- "source": {
- "file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 638,
- "end_line": 648
+ "begin_line": 558,
+ "end_line": 563
},
"is_opaque": false,
"can_fail": true,
@@ -1766,8 +1715,8 @@
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 653,
- "end_line": 705
+ "begin_line": 568,
+ "end_line": 620
},
"is_opaque": false,
"can_fail": true,
@@ -1783,8 +1732,8 @@
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 710,
- "end_line": 766
+ "begin_line": 625,
+ "end_line": 667
},
"is_opaque": false,
"can_fail": true,
@@ -1800,8 +1749,8 @@
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 772,
- "end_line": 840
+ "begin_line": 673,
+ "end_line": 742
},
"is_opaque": false,
"can_fail": true,
@@ -1817,8 +1766,8 @@
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 846,
- "end_line": 873
+ "begin_line": 748,
+ "end_line": 772
},
"is_opaque": false,
"can_fail": true,
@@ -1912,15 +1861,15 @@
"reducible": false
},
{
- "def_id": 71,
+ "def_id": 69,
"lean_name": "auths_gateway_kernel.transition.absent",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::transition::absent",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/transition.rs",
- "begin_line": 89,
- "end_line": 91
+ "begin_line": 113,
+ "end_line": 115
},
"is_opaque": false,
"can_fail": true,
@@ -1929,15 +1878,15 @@
"reducible": false
},
{
- "def_id": 68,
+ "def_id": 81,
"lean_name": "auths_gateway_kernel.transition.same_bytes_from_loop.body",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::transition::same_bytes_from",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/transition.rs",
- "begin_line": 97,
- "end_line": 104
+ "begin_line": 121,
+ "end_line": 128
},
"is_opaque": false,
"can_fail": true,
@@ -1954,15 +1903,15 @@
"parent_lean_name": "auths_gateway_kernel.transition.same_bytes_from"
},
{
- "def_id": 68,
+ "def_id": 81,
"lean_name": "auths_gateway_kernel.transition.same_bytes_from_loop",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::transition::same_bytes_from",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/transition.rs",
- "begin_line": 97,
- "end_line": 104
+ "begin_line": 121,
+ "end_line": 128
},
"is_opaque": false,
"can_fail": true,
@@ -1979,15 +1928,15 @@
"parent_lean_name": "auths_gateway_kernel.transition.same_bytes_from"
},
{
- "def_id": 68,
+ "def_id": 81,
"lean_name": "auths_gateway_kernel.transition.same_bytes_from",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::transition::same_bytes_from",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/transition.rs",
- "begin_line": 95,
- "end_line": 104
+ "begin_line": 119,
+ "end_line": 128
},
"is_opaque": false,
"can_fail": true,
@@ -1996,15 +1945,15 @@
"reducible": true
},
{
- "def_id": 27,
+ "def_id": 30,
"lean_name": "auths_gateway_kernel.transition.same_bytes",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::transition::same_bytes",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/transition.rs",
- "begin_line": 108,
- "end_line": 114
+ "begin_line": 132,
+ "end_line": 138
},
"is_opaque": false,
"can_fail": true,
@@ -2013,15 +1962,15 @@
"reducible": false
},
{
- "def_id": 70,
+ "def_id": 80,
"lean_name": "auths_gateway_kernel.transition.resolves_unknown",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::transition::resolves_unknown",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/transition.rs",
- "begin_line": 119,
- "end_line": 124
+ "begin_line": 143,
+ "end_line": 148
},
"is_opaque": false,
"can_fail": true,
@@ -2030,15 +1979,15 @@
"reducible": false
},
{
- "def_id": 80,
+ "def_id": 79,
"lean_name": "auths_gateway_kernel.transition.has_link",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::transition::has_link",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/transition.rs",
- "begin_line": 128,
- "end_line": 133
+ "begin_line": 152,
+ "end_line": 157
},
"is_opaque": false,
"can_fail": true,
@@ -2047,15 +1996,15 @@
"reducible": false
},
{
- "def_id": 74,
+ "def_id": 73,
"lean_name": "auths_gateway_kernel.transition.refusal_stage",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::transition::refusal_stage",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/transition.rs",
- "begin_line": 137,
- "end_line": 146
+ "begin_line": 161,
+ "end_line": 172
},
"is_opaque": false,
"can_fail": true,
@@ -2064,15 +2013,15 @@
"reducible": false
},
{
- "def_id": 75,
+ "def_id": 74,
"lean_name": "auths_gateway_kernel.transition.requires_response",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::transition::requires_response",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/transition.rs",
- "begin_line": 150,
- "end_line": 155
+ "begin_line": 176,
+ "end_line": 186
},
"is_opaque": false,
"can_fail": true,
@@ -2081,15 +2030,15 @@
"reducible": false
},
{
- "def_id": 76,
+ "def_id": 75,
"lean_name": "auths_gateway_kernel.transition.permits_response",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::transition::permits_response",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/transition.rs",
- "begin_line": 159,
- "end_line": 164
+ "begin_line": 190,
+ "end_line": 199
},
"is_opaque": false,
"can_fail": true,
@@ -2098,15 +2047,15 @@
"reducible": false
},
{
- "def_id": 77,
+ "def_id": 76,
"lean_name": "auths_gateway_kernel.transition.observed_stage",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::transition::observed_stage",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/transition.rs",
- "begin_line": 168,
- "end_line": 177
+ "begin_line": 203,
+ "end_line": 214
},
"is_opaque": false,
"can_fail": true,
@@ -2115,15 +2064,15 @@
"reducible": false
},
{
- "def_id": 79,
+ "def_id": 78,
"lean_name": "auths_gateway_kernel.transition.provider_stage",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::transition::provider_stage",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/transition.rs",
- "begin_line": 181,
- "end_line": 190
+ "begin_line": 218,
+ "end_line": 229
},
"is_opaque": false,
"can_fail": true,
@@ -2132,15 +2081,15 @@
"reducible": false
},
{
- "def_id": 78,
+ "def_id": 77,
"lean_name": "auths_gateway_kernel.transition.has_reading",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::transition::has_reading",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/transition.rs",
- "begin_line": 203,
- "end_line": 208
+ "begin_line": 244,
+ "end_line": 249
},
"is_opaque": false,
"can_fail": true,
@@ -2149,15 +2098,66 @@
"reducible": false
},
{
- "def_id": 33,
+ "def_id": 71,
+ "lean_name": "auths_gateway_kernel.transition.requires_entry",
+ "lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
+ "rust_name": "auths_gateway_kernel::transition::requires_entry",
+ "is_local": true,
+ "source": {
+ "file": "product/runtime/auths-gateway-kernel/src/transition.rs",
+ "begin_line": 336,
+ "end_line": 346
+ },
+ "is_opaque": false,
+ "can_fail": true,
+ "can_diverge": false,
+ "is_rec": false,
+ "reducible": false
+ },
+ {
+ "def_id": 72,
+ "lean_name": "auths_gateway_kernel.transition.acknowledgement_stage",
+ "lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
+ "rust_name": "auths_gateway_kernel::transition::acknowledgement_stage",
+ "is_local": true,
+ "source": {
+ "file": "product/runtime/auths-gateway-kernel/src/transition.rs",
+ "begin_line": 278,
+ "end_line": 280
+ },
+ "is_opaque": false,
+ "can_fail": true,
+ "can_diverge": false,
+ "is_rec": false,
+ "reducible": false
+ },
+ {
+ "def_id": 70,
+ "lean_name": "auths_gateway_kernel.transition.protocol_consistent",
+ "lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
+ "rust_name": "auths_gateway_kernel::transition::protocol_consistent",
+ "is_local": true,
+ "source": {
+ "file": "product/runtime/auths-gateway-kernel/src/transition.rs",
+ "begin_line": 285,
+ "end_line": 321
+ },
+ "is_opaque": false,
+ "can_fail": true,
+ "can_diverge": false,
+ "is_rec": false,
+ "reducible": false
+ },
+ {
+ "def_id": 28,
"lean_name": "auths_gateway_kernel.transition.consistent",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::transition::consistent",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/transition.rs",
- "begin_line": 217,
- "end_line": 228
+ "begin_line": 258,
+ "end_line": 274
},
"is_opaque": false,
"can_fail": true,
@@ -2166,15 +2166,32 @@
"reducible": false
},
{
- "def_id": 29,
+ "def_id": 27,
+ "lean_name": "auths_gateway_kernel.transition.same_protocol",
+ "lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
+ "rust_name": "auths_gateway_kernel::transition::same_protocol",
+ "is_local": true,
+ "source": {
+ "file": "product/runtime/auths-gateway-kernel/src/transition.rs",
+ "begin_line": 325,
+ "end_line": 332
+ },
+ "is_opaque": false,
+ "can_fail": true,
+ "can_diverge": false,
+ "is_rec": false,
+ "reducible": false
+ },
+ {
+ "def_id": 32,
"lean_name": "auths_gateway_kernel.transition.stage_transition_allowed",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::transition::stage_transition_allowed",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/transition.rs",
- "begin_line": 235,
- "end_line": 260
+ "begin_line": 352,
+ "end_line": 373
},
"is_opaque": false,
"can_fail": true,
@@ -2183,15 +2200,15 @@
"reducible": false
},
{
- "def_id": 72,
+ "def_id": 83,
"lean_name": "auths_gateway_kernel.transition.adds_pre_entry",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::transition::adds_pre_entry",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/transition.rs",
- "begin_line": 265,
- "end_line": 280
+ "begin_line": 377,
+ "end_line": 379
},
"is_opaque": false,
"can_fail": true,
@@ -2200,15 +2217,15 @@
"reducible": false
},
{
- "def_id": 73,
+ "def_id": 84,
"lean_name": "auths_gateway_kernel.transition.adds_response",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::transition::adds_response",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/transition.rs",
- "begin_line": 284,
- "end_line": 300
+ "begin_line": 383,
+ "end_line": 385
},
"is_opaque": false,
"can_fail": true,
@@ -2217,15 +2234,15 @@
"reducible": false
},
{
- "def_id": 30,
+ "def_id": 33,
"lean_name": "auths_gateway_kernel.transition.checkpoint_valid",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::transition::checkpoint_valid",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/transition.rs",
- "begin_line": 305,
- "end_line": 321
+ "begin_line": 389,
+ "end_line": 395
},
"is_opaque": false,
"can_fail": true,
@@ -2234,15 +2251,32 @@
"reducible": false
},
{
- "def_id": 31,
+ "def_id": 29,
+ "lean_name": "auths_gateway_kernel.transition.entry_kept",
+ "lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
+ "rust_name": "auths_gateway_kernel::transition::entry_kept",
+ "is_local": true,
+ "source": {
+ "file": "product/runtime/auths-gateway-kernel/src/transition.rs",
+ "begin_line": 399,
+ "end_line": 405
+ },
+ "is_opaque": false,
+ "can_fail": true,
+ "can_diverge": true,
+ "is_rec": false,
+ "reducible": false
+ },
+ {
+ "def_id": 34,
"lean_name": "auths_gateway_kernel.transition.pre_entry_kept",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::transition::pre_entry_kept",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/transition.rs",
- "begin_line": 326,
- "end_line": 332
+ "begin_line": 410,
+ "end_line": 416
},
"is_opaque": false,
"can_fail": true,
@@ -2251,15 +2285,15 @@
"reducible": false
},
{
- "def_id": 32,
+ "def_id": 35,
"lean_name": "auths_gateway_kernel.transition.response_kept",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::transition::response_kept",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/transition.rs",
- "begin_line": 337,
- "end_line": 340
+ "begin_line": 421,
+ "end_line": 424
},
"is_opaque": false,
"can_fail": true,
@@ -2268,15 +2302,32 @@
"reducible": false
},
{
- "def_id": 69,
+ "def_id": 36,
+ "lean_name": "auths_gateway_kernel.transition.acknowledgement_kept",
+ "lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
+ "rust_name": "auths_gateway_kernel::transition::acknowledgement_kept",
+ "is_local": true,
+ "source": {
+ "file": "product/runtime/auths-gateway-kernel/src/transition.rs",
+ "begin_line": 429,
+ "end_line": 436
+ },
+ "is_opaque": false,
+ "can_fail": true,
+ "can_diverge": true,
+ "is_rec": false,
+ "reducible": false
+ },
+ {
+ "def_id": 82,
"lean_name": "auths_gateway_kernel.transition.link_equal",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::transition::link_equal",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/transition.rs",
- "begin_line": 350,
- "end_line": 365
+ "begin_line": 459,
+ "end_line": 474
},
"is_opaque": false,
"can_fail": true,
@@ -2285,15 +2336,15 @@
"reducible": false
},
{
- "def_id": 28,
+ "def_id": 31,
"lean_name": "auths_gateway_kernel.transition.plan_kept",
"lean_file": "qualification/aeneas/generated/gateway/Funs.lean",
"rust_name": "auths_gateway_kernel::transition::plan_kept",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/transition.rs",
- "begin_line": 344,
- "end_line": 346
+ "begin_line": 453,
+ "end_line": 455
},
"is_opaque": false,
"can_fail": true,
@@ -2309,8 +2360,8 @@
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/transition.rs",
- "begin_line": 373,
- "end_line": 381
+ "begin_line": 482,
+ "end_line": 496
},
"is_opaque": false,
"can_fail": true,
@@ -2381,7 +2432,7 @@
}
},
{
- "def_id": 34,
+ "def_id": 35,
"lean_name": "auths_gateway_kernel.construct.AccountScopePlan",
"lean_file": "qualification/aeneas/generated/gateway/Types.lean",
"rust_name": "auths_gateway_kernel::construct::AccountScopePlan",
@@ -2405,7 +2456,7 @@
}
},
{
- "def_id": 32,
+ "def_id": 33,
"lean_name": "auths_gateway_kernel.construct.JsonPiece",
"lean_file": "qualification/aeneas/generated/gateway/Types.lean",
"rust_name": "auths_gateway_kernel::construct::JsonPiece",
@@ -2417,7 +2468,7 @@
}
},
{
- "def_id": 33,
+ "def_id": 34,
"lean_name": "auths_gateway_kernel.construct.FormPiece",
"lean_file": "qualification/aeneas/generated/gateway/Types.lean",
"rust_name": "auths_gateway_kernel::construct::FormPiece",
@@ -2500,18 +2551,6 @@
"end_line": 207
}
},
- {
- "def_id": 35,
- "lean_name": "auths_gateway_kernel.order.Mode",
- "lean_file": "qualification/aeneas/generated/gateway/Types.lean",
- "rust_name": "auths_gateway_kernel::order::Mode",
- "is_local": true,
- "source": {
- "file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 34,
- "end_line": 41
- }
- },
{
"def_id": 23,
"lean_name": "auths_gateway_kernel.order.SubmitPlan",
@@ -2520,8 +2559,8 @@
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 50,
- "end_line": 63
+ "begin_line": 39,
+ "end_line": 52
}
},
{
@@ -2532,8 +2571,8 @@
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 67,
- "end_line": 116
+ "begin_line": 56,
+ "end_line": 103
}
},
{
@@ -2544,8 +2583,8 @@
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 120,
- "end_line": 128
+ "begin_line": 107,
+ "end_line": 115
}
},
{
@@ -2556,8 +2595,8 @@
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 132,
- "end_line": 141
+ "begin_line": 119,
+ "end_line": 128
}
},
{
@@ -2568,8 +2607,8 @@
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 145,
- "end_line": 154
+ "begin_line": 132,
+ "end_line": 141
}
},
{
@@ -2580,8 +2619,8 @@
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 158,
- "end_line": 165
+ "begin_line": 145,
+ "end_line": 152
}
},
{
@@ -2592,8 +2631,8 @@
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 169,
- "end_line": 176
+ "begin_line": 156,
+ "end_line": 163
}
},
{
@@ -2604,8 +2643,8 @@
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 180,
- "end_line": 187
+ "begin_line": 167,
+ "end_line": 174
}
},
{
@@ -2616,8 +2655,8 @@
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 191,
- "end_line": 201
+ "begin_line": 178,
+ "end_line": 188
}
},
{
@@ -2628,8 +2667,8 @@
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 205,
- "end_line": 212
+ "begin_line": 192,
+ "end_line": 197
}
},
{
@@ -2640,8 +2679,8 @@
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 216,
- "end_line": 224
+ "begin_line": 201,
+ "end_line": 206
}
},
{
@@ -2652,8 +2691,8 @@
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 228,
- "end_line": 267
+ "begin_line": 210,
+ "end_line": 249
}
},
{
@@ -2664,8 +2703,8 @@
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 271,
- "end_line": 300
+ "begin_line": 253,
+ "end_line": 282
}
},
{
@@ -2676,8 +2715,8 @@
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 304,
- "end_line": 323
+ "begin_line": 286,
+ "end_line": 305
}
},
{
@@ -2688,8 +2727,8 @@
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 327,
- "end_line": 374
+ "begin_line": 309,
+ "end_line": 354
}
},
{
@@ -2700,8 +2739,8 @@
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/order.rs",
- "begin_line": 378,
- "end_line": 383
+ "begin_line": 358,
+ "end_line": 363
}
},
{
@@ -2717,7 +2756,7 @@
}
},
{
- "def_id": 31,
+ "def_id": 32,
"lean_name": "auths_gateway_kernel.recovery.IdempotencyKind",
"lean_file": "qualification/aeneas/generated/gateway/Types.lean",
"rust_name": "auths_gateway_kernel::recovery::IdempotencyKind",
@@ -2790,14 +2829,26 @@
},
{
"def_id": 29,
+ "lean_name": "auths_gateway_kernel.transition.Protocol",
+ "lean_file": "qualification/aeneas/generated/gateway/Types.lean",
+ "rust_name": "auths_gateway_kernel::transition::Protocol",
+ "is_local": true,
+ "source": {
+ "file": "product/runtime/auths-gateway-kernel/src/transition.rs",
+ "begin_line": 19,
+ "end_line": 26
+ }
+ },
+ {
+ "def_id": 30,
"lean_name": "auths_gateway_kernel.transition.Stage",
"lean_file": "qualification/aeneas/generated/gateway/Types.lean",
"rust_name": "auths_gateway_kernel::transition::Stage",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/transition.rs",
- "begin_line": 19,
- "end_line": 32
+ "begin_line": 30,
+ "end_line": 47
}
},
{
@@ -2808,20 +2859,20 @@
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/transition.rs",
- "begin_line": 36,
- "end_line": 45
+ "begin_line": 51,
+ "end_line": 60
}
},
{
- "def_id": 30,
+ "def_id": 31,
"lean_name": "auths_gateway_kernel.transition.Link",
"lean_file": "qualification/aeneas/generated/gateway/Types.lean",
"rust_name": "auths_gateway_kernel::transition::Link",
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/transition.rs",
- "begin_line": 49,
- "end_line": 56
+ "begin_line": 64,
+ "end_line": 71
}
},
{
@@ -2832,8 +2883,8 @@
"is_local": true,
"source": {
"file": "product/runtime/auths-gateway-kernel/src/transition.rs",
- "begin_line": 60,
- "end_line": 81
+ "begin_line": 75,
+ "end_line": 105
}
}
],
diff --git a/formal/qualification/aeneas/qualification.toml b/formal/qualification/aeneas/qualification.toml
index 62cc62aad..e1ca72405 100644
--- a/formal/qualification/aeneas/qualification.toml
+++ b/formal/qualification/aeneas/qualification.toml
@@ -397,7 +397,7 @@ allowed_external_symbols = []
[[translations]]
crate_name = "auths_gateway_kernel"
translation_json = "formal/qualification/aeneas/generated/gateway/translation.json"
-local_functions = 123
+local_functions = 126
external_functions = 0
opaque_local_functions = 0
required_symbols = [
@@ -414,6 +414,12 @@ required_symbols = [
"auths_gateway_kernel::order::after_ceiling",
"auths_gateway_kernel::transition::valid_transition",
"auths_gateway_kernel::transition::consistent",
+ "auths_gateway_kernel::transition::protocol_consistent",
+ "auths_gateway_kernel::transition::entry_kept",
+ "auths_gateway_kernel::transition::acknowledgement_kept",
+ "auths_gateway_kernel::transition::acknowledgement_stage",
+ "auths_gateway_kernel::transition::requires_entry",
+ "auths_gateway_kernel::transition::same_protocol",
"auths_gateway_kernel::transition::stage_transition_allowed",
"auths_gateway_kernel::ratio::relative_basis",
"auths_gateway_kernel::ratio::relative_ceiling_admits",
diff --git a/formal/qualification/aeneas/source-closure.json b/formal/qualification/aeneas/source-closure.json
index f0600223b..dcd2dd29a 100644
--- a/formal/qualification/aeneas/source-closure.json
+++ b/formal/qualification/aeneas/source-closure.json
@@ -1,6 +1,6 @@
{
"schema": "auths-proof-translation-source-closure/v2",
- "digest": "5cf292b7c37fbc6183c496fd348e885b87204d515353fb5088f2d181b69b7720",
+ "digest": "09fb135d91c47ff1b2a873b583564db3467e76973c9763090bac2affd73b6837",
"files": [
{
"path": ".cargo/config.toml",
@@ -8,12 +8,12 @@
},
{
"path": "Cargo.lock",
- "sha256": "f84825e2ac0d4af57a406ab4b49ae5ce4e9360caf0a23e092906257743509b04",
+ "sha256": "ab4ed8625de4336a74db8583cc3fb11b285f1e66b9d0aee349f9fb30a813536c",
"normalization": "translated-cargo-closure-v1"
},
{
"path": "Cargo.toml",
- "sha256": "6e0240aa432bcdb5b4f855618994582d2e428afe5ee4e4f6a3f8d21a518f8d24"
+ "sha256": "ccc17deac72da945940a9ab4be2ce27628c056b977d19429bc1679b19fae30f3"
},
{
"path": "core/crates/auths-algebra-kernel/Cargo.toml",
@@ -81,7 +81,7 @@
},
{
"path": "formal/qualification/aeneas/cases/Gateway.lean",
- "sha256": "a6d3148ab9ad313a0d00b92313d6fb2bc70a131759e5ccffb9d6f5e028738264"
+ "sha256": "094648470bdda3587c346a4c7ae1a29c115be613d589bd7a69dc2141d34303bc"
},
{
"path": "formal/qualification/aeneas/cases/Lifecycle.lean",
@@ -93,7 +93,7 @@
},
{
"path": "formal/qualification/aeneas/qualification.toml",
- "sha256": "38aa0babb5be3ac958c7e91f0640ad0ae3b6e4f265257642976dfe8bd3330850"
+ "sha256": "1da6c96a8d64598ff8a6999d6834e4f78fc0dee69d8c56ab3f69c37ecf436d58"
},
{
"path": "formal/translation-toolchain.lock",
@@ -157,7 +157,7 @@
},
{
"path": "product/runtime/auths-connections/src/credential.rs",
- "sha256": "82921413bb053f0f6c7931fbd4e6a86c8cc1d8c6b51debd9c2cef33a8d5b1c51"
+ "sha256": "7e6fbca8911d5c0c1baddf35572236d419566dd6e3fcbbdfda77c8dd114dbcad"
},
{
"path": "product/runtime/auths-connections/src/kernel.rs",
@@ -189,11 +189,11 @@
},
{
"path": "product/runtime/auths-gateway-kernel/src/order.rs",
- "sha256": "e0045f877c4b7fe3860c291039853885064ddcdebba7c9b345ab65e81e646dda"
+ "sha256": "efc2f2eef68232fe3082e1607e6f1e87d4a19cbb2f492bdc6e71916a2461b6a8"
},
{
"path": "product/runtime/auths-gateway-kernel/src/outcome.rs",
- "sha256": "e9785b1afe70788c89b51b18655bcd5acff7b0c8e3d12fb79c6f05a202e67e6b"
+ "sha256": "ac0245699457c802a70cdd6f52a01dcb4777c487b58ed7dc9cfc99345465ccaa"
},
{
"path": "product/runtime/auths-gateway-kernel/src/ratio.rs",
@@ -205,7 +205,7 @@
},
{
"path": "product/runtime/auths-gateway-kernel/src/transition.rs",
- "sha256": "cefb8f01d915880f6921f2501ff881ad94b9de61072c34d9cd19220992f97374"
+ "sha256": "9c9c292397aa640c2f002ffb5befe110f86d0aa246f3ffe99440718e5174af3c"
},
{
"path": "product/runtime/auths-lifecycle/Cargo.toml",
@@ -273,7 +273,7 @@
},
{
"path": "xtask/ci-plan/src/lib.rs",
- "sha256": "aeae292998e347454a2396933d4e4968460c70c165f88b7a6586985582a0463c"
+ "sha256": "27ed7fb84e45c5ef7a81edfc350f3299f410522abb10364c0374aba8751c9732"
},
{
"path": "xtask/ci-plan/src/main.rs",
@@ -325,7 +325,7 @@
},
{
"path": "xtask/src/formal.rs",
- "sha256": "99d0378a44c59d6ce227033a5aa8af193d6b55fe76a519ba06e28d0f2ae2094d"
+ "sha256": "0663dbad2e21a0b9a47d5787c24d193a7b85348f656228ba8d489a9d6028e985"
},
{
"path": "xtask/src/formal_coverage.rs",
@@ -337,11 +337,11 @@
},
{
"path": "xtask/src/formal_qualification.rs",
- "sha256": "1826078f21dbf946801413cb6cf1f66dfd7495c9de3853455fbf2eb151f899e6"
+ "sha256": "5565f25fb5e321d32baf8bd9ac028348aa47fe16864824973fceca7c69ed45cd"
},
{
"path": "xtask/src/fuzz.rs",
- "sha256": "d7f5286aa0eccca894b2b8ea4b2d10b8e5261a6c7010ef696403a7550f5225bf"
+ "sha256": "25f87f093f4efb89d7818d8dea9d29bb79c55b7dcb80156db5b8b700863e62f8"
},
{
"path": "xtask/src/kani_harness.rs",
@@ -353,7 +353,7 @@
},
{
"path": "xtask/src/main.rs",
- "sha256": "e897303e906ad3eab43badd52a60142ee1350a6f53e8e7ad8b2e505ac88240fe"
+ "sha256": "6b3bfb08ee10977f16741369b2c321b4e16d9f0c6e091bfcdca08471430eb459"
},
{
"path": "xtask/src/mcp_session_contract.rs",
@@ -401,7 +401,7 @@
},
{
"path": "xtask/src/semantic_freeze.rs",
- "sha256": "9e36c78ae7ed8ce11bbc545ef7db8c1741c826038d01a7fb66d6e7fa5c91b18b"
+ "sha256": "f98ace0550ead760eeb850fdb2a120ce0ea953ab2b6112da3a5f80175b222c9a"
},
{
"path": "xtask/src/stripe.rs",
diff --git a/product/integrations/auths-custody-aws-kms/src/lib.rs b/product/integrations/auths-custody-aws-kms/src/lib.rs
index 389779425..1585f47b1 100644
--- a/product/integrations/auths-custody-aws-kms/src/lib.rs
+++ b/product/integrations/auths-custody-aws-kms/src/lib.rs
@@ -7,7 +7,7 @@ use auths_custody::{
CustodyProviderError, ExternalSigner, KeyLifecycleState, KeyVersionId, P256SignatureVerifier,
RawSigningResponse, SigningIntent, UntrustedSigningResponse,
};
-use p256::{EncodedPoint, pkcs8::DecodePublicKey as _};
+use p256::{Sec1Point, pkcs8::DecodePublicKey as _};
use sha2::{Digest as _, Sha256};
const ADAPTER_ID: &str = "aws-kms-p256-v1";
@@ -231,7 +231,7 @@ impl AwsKmsP256Adapter {
.map_err(AwsKmsConfigurationError::Provider)?;
let key = p256::PublicKey::from_public_key_der(&public_key_der)
.map_err(|_| AwsKmsConfigurationError::InvalidPublicKey)?;
- let sec1 = EncodedPoint::from(key).compress();
+ let sec1 = Sec1Point::from(key).compress();
let identity = CustodyIdentity::p256(form, sec1.as_bytes())
.map_err(|_| AwsKmsConfigurationError::InvalidPublicKey)?;
let key_version = key_version(key_arn, &public_key_der)?;
@@ -407,7 +407,7 @@ mod tests {
}
fn get_public_key(&self, _key_arn: &str) -> Result, AwsKmsFailure> {
- let encoded = self.key.verifying_key().to_encoded_point(false);
+ let encoded = self.key.verifying_key().to_sec1_point(false);
Ok(p256::PublicKey::from_sec1_bytes(encoded.as_bytes())
.unwrap()
.to_public_key_der()
@@ -424,7 +424,7 @@ mod tests {
_message_type: AwsKmsMessageType,
) -> Result {
let signature: Signature = self.key.sign(message);
- let signature = signature.normalize_s().unwrap_or(signature);
+ let signature = signature.normalize_s();
Ok(AwsKmsSignOutput {
key_arn: self.arn.clone(),
algorithm,
diff --git a/product/integrations/auths-custody-pkcs11/src/lib.rs b/product/integrations/auths-custody-pkcs11/src/lib.rs
index a6afb2bfe..46e5a9a09 100644
--- a/product/integrations/auths-custody-pkcs11/src/lib.rs
+++ b/product/integrations/auths-custody-pkcs11/src/lib.rs
@@ -411,7 +411,7 @@ mod tests {
public_key_sec1: self
.key
.verifying_key()
- .to_encoded_point(true)
+ .to_sec1_point(true)
.as_bytes()
.to_vec(),
p256: true,
@@ -427,7 +427,7 @@ mod tests {
message: &[u8],
) -> Result {
let signature: Signature = self.key.sign(message);
- let signature = signature.normalize_s().unwrap_or(signature);
+ let signature = signature.normalize_s();
Ok(Pkcs11SignOutput {
signature: signature.to_bytes().to_vec(),
})
diff --git a/product/integrations/auths-custody/src/conformance.rs b/product/integrations/auths-custody/src/conformance.rs
index 2532b6242..f62007c3d 100644
--- a/product/integrations/auths-custody/src/conformance.rs
+++ b/product/integrations/auths-custody/src/conformance.rs
@@ -143,7 +143,7 @@ impl ConformanceSigner {
) -> Self {
let key = SigningKey::from_slice(&KEY).expect("fixed P-256 key");
let identity =
- CustodyIdentity::p256(form, key.verifying_key().to_encoded_point(true).as_bytes())
+ CustodyIdentity::p256(form, key.verifying_key().to_sec1_point(true).as_bytes())
.expect("fixed identity");
let descriptor = CustodyDescriptor::new(
kind,
@@ -165,7 +165,7 @@ impl ConformanceSigner {
fn signature(key: &SigningKey, message: &[u8]) -> Signature {
let signature: Signature = key.sign(message);
- signature.normalize_s().unwrap_or(signature)
+ signature.normalize_s()
}
}
diff --git a/product/integrations/auths-custody/src/key.rs b/product/integrations/auths-custody/src/key.rs
index 119888a0a..682692b26 100644
--- a/product/integrations/auths-custody/src/key.rs
+++ b/product/integrations/auths-custody/src/key.rs
@@ -20,7 +20,7 @@ use auths_model::{
};
use auths_multikey::{Multikey, MultikeyType};
use auths_raw_key::{RAW_KEY_MEDIA_TYPE, RAW_KEY_V1, RawKeyDescriptor, RawKeyType};
-use p256::{EncodedPoint, PublicKey};
+use p256::{PublicKey, Sec1Point};
const P256_SUITE: &str = "p256-sha256-v1";
@@ -53,7 +53,7 @@ impl CustodyIdentity {
/// valid P-256 point or cannot be represented under `form`.
pub fn p256(form: CustodyPrincipalForm, sec1: &[u8]) -> Result {
let key = PublicKey::from_sec1_bytes(sec1).map_err(|_| CustodyError::EvidenceMismatch)?;
- let public_key = EncodedPoint::from(key).compress().as_bytes().to_vec();
+ let public_key = Sec1Point::from(key).compress().as_bytes().to_vec();
let invalid = |_| CustodyError::EvidenceMismatch;
let suite = SignatureSuiteId::parse(P256_SUITE).map_err(invalid)?;
let (principal, method, verification, evidence_type, media_type, bytes) = match form {
diff --git a/product/integrations/auths-custody/src/lib.rs b/product/integrations/auths-custody/src/lib.rs
index 60af2209a..337fb7447 100644
--- a/product/integrations/auths-custody/src/lib.rs
+++ b/product/integrations/auths-custody/src/lib.rs
@@ -604,7 +604,7 @@ fn canonical_signature(
P256Signature::from_der(&bytes)
}
.map_err(|_| CustodyError::MalformedSignature)?;
- if signature.normalize_s().is_some() {
+ if signature.normalize_s() != signature {
return Err(CustodyError::NonCanonicalSignature);
}
SignatureBytes::new(signature.to_bytes().to_vec()).map_err(|_| CustodyError::MalformedSignature)
@@ -748,7 +748,7 @@ mod tests {
request: &SigningIntent<'_>,
) -> Result {
let signature: P256Signature = self.key.sign(request.signing_preimage());
- let signature = signature.normalize_s().unwrap_or(signature);
+ let signature = signature.normalize_s();
let mut transaction_digest = *request.transaction_digest();
if self.transaction_mismatch {
transaction_digest[0] ^= 1;
@@ -798,7 +798,7 @@ mod tests {
P256SignatureVerifier,
) {
let key = SigningKey::from_slice(&[7; 32]).unwrap();
- let verification = key.verifying_key().to_encoded_point(true);
+ let verification = key.verifying_key().to_sec1_point(true);
let principal = PrincipalId::parse("raw:p256-test").unwrap();
let signature = SignatureDescriptor::new(
PrincipalMethodId::parse("raw-key-v1").unwrap(),
@@ -858,7 +858,7 @@ mod tests {
fn observation(principal: &PrincipalId) -> auths_model::ObservationStatement {
auths_model::ObservationStatement::new(
principal.clone(),
- auths_model::ObservationSchemaId::parse("auths.gateway-outcome/2").unwrap(),
+ auths_model::ObservationSchemaId::parse("auths.gateway-outcome/3").unwrap(),
ResourceId::parse("auths-gateway://ns/operations/op-1").unwrap(),
auths_model::Timestamp::new(1),
auths_model::ObservationFacts::new(vec![auths_model::ObservationFact::new(
@@ -955,7 +955,7 @@ mod tests {
let other = SigningKey::from_slice(&[5; 32]).unwrap();
let identity = CustodyIdentity::p256(
CustodyPrincipalForm::RawKeyV1,
- other.verifying_key().to_encoded_point(false).as_bytes(),
+ other.verifying_key().to_sec1_point(false).as_bytes(),
)
.unwrap();
assert!(matches!(
diff --git a/product/integrations/auths-git-signing/src/custody_tests.rs b/product/integrations/auths-git-signing/src/custody_tests.rs
index 5fc8a0f09..c8b344b75 100644
--- a/product/integrations/auths-git-signing/src/custody_tests.rs
+++ b/product/integrations/auths-git-signing/src/custody_tests.rs
@@ -139,7 +139,7 @@ impl Pkcs11Api for Token {
public_key_sec1: self
.0
.verifying_key()
- .to_encoded_point(true)
+ .to_sec1_point(true)
.as_bytes()
.to_vec(),
p256: true,
@@ -156,7 +156,7 @@ impl Pkcs11Api for Token {
) -> Result {
let signature: Signature = self.0.sign(message);
Ok(Pkcs11SignOutput {
- signature: signature.normalize_s().unwrap_or(signature).to_vec(),
+ signature: signature.normalize_s().to_vec(),
})
}
}
diff --git a/product/integrations/auths-git-signing/src/workload.rs b/product/integrations/auths-git-signing/src/workload.rs
index 77c240dcb..6653f4cbd 100644
--- a/product/integrations/auths-git-signing/src/workload.rs
+++ b/product/integrations/auths-git-signing/src/workload.rs
@@ -263,7 +263,7 @@ pub const WORKLOAD_CUSTODY: &str = "workload";
/// encoding Fulcio and Rekor take.
fn sign_p256(key: &P256SigningKey, message: &[u8]) -> P256Signature {
let signature: P256Signature = p256::ecdsa::signature::Signer::sign(key, message);
- signature.normalize_s().unwrap_or(signature)
+ signature.normalize_s()
}
fn sign_ed25519(key: &SigningKey, preimage: &[u8]) -> Result {
@@ -443,7 +443,7 @@ impl<'c> SigstoreKeylessSigner<'c> {
) -> Result {
let key = ephemeral_p256_key()?;
let mut spki = P256_SPKI_PREFIX.to_vec();
- spki.extend_from_slice(key.verifying_key().to_encoded_point(false).as_bytes());
+ spki.extend_from_slice(key.verifying_key().to_sec1_point(false).as_bytes());
let token = tokens.token(SIGSTORE_AUDIENCE)?;
let identity = TokenIdentity::read(token.as_bytes(), SIGSTORE_AUDIENCE)?;
let proof = sign_p256(&key, identity.subject.as_bytes()).to_der();
diff --git a/product/integrations/auths-git-signing/src/workload_fakes.rs b/product/integrations/auths-git-signing/src/workload_fakes.rs
index 5f528706d..5bf893f0f 100644
--- a/product/integrations/auths-git-signing/src/workload_fakes.rs
+++ b/product/integrations/auths-git-signing/src/workload_fakes.rs
@@ -10,7 +10,7 @@ use crate::workload::{
use auths_oidc_workload::jws::CompactJws;
use auths_sigstore_keyless::merkle::leaf_hash;
use base64ct::{Base64, Base64UrlUnpadded, Encoding as _};
-use ed25519_dalek::SigningKey;
+use ed25519_dalek::{Signer as _, SigningKey};
use p256::ecdsa::signature::{Signer as _, Verifier as _};
use p256::ecdsa::{Signature as P256Signature, SigningKey as P256SigningKey, VerifyingKey};
use rcgen::{
@@ -213,7 +213,7 @@ impl FakeLogKey {
}
Self::P256(key) => {
let mut spki = P256_SPKI_PREFIX.to_vec();
- spki.extend_from_slice(key.verifying_key().to_encoded_point(false).as_bytes());
+ spki.extend_from_slice(key.verifying_key().to_sec1_point(false).as_bytes());
spki
}
}
@@ -226,12 +226,12 @@ impl FakeLogKey {
.to_vec(),
Self::P256(key) => {
let signature: P256Signature = key.sign(message);
- let low = signature.normalize_s().unwrap_or(signature);
+ let low = signature.normalize_s();
let (r, s) = low.split_scalars();
let high = P256Signature::from_scalars(r.to_bytes(), (-*s).to_bytes())
.expect("high-S pair");
assert!(
- high.normalize_s().is_some(),
+ high.normalize_s() != high,
"the emitted signature is high-S"
);
high.to_der().as_bytes().to_vec()
@@ -404,7 +404,7 @@ impl SigstoreClient for FakeSigstore {
),
];
set_validity(&mut params, self.leaf_window.0, self.leaf_window.1);
- let point = key.to_encoded_point(false).as_bytes().to_vec();
+ let point = key.to_sec1_point(false).as_bytes().to_vec();
let leaf = params
.signed_by(&RawP256(point), &self.ca, &self.ca_key)
.map_err(|_| refuse("leaf"))?;
diff --git a/product/qualification/auths-recipe-qualification-issuance/src/execution.rs b/product/qualification/auths-recipe-qualification-issuance/src/execution.rs
index 1dc21ff9b..01ca64cfc 100644
--- a/product/qualification/auths-recipe-qualification-issuance/src/execution.rs
+++ b/product/qualification/auths-recipe-qualification-issuance/src/execution.rs
@@ -467,30 +467,15 @@ impl RunCase {
{
return Err(IssuanceError::CaseFailed);
}
- let mut unresolved = false;
for (step, actual) in self.steps.iter().zip(observations) {
- if step.operation == Operation::Replay {
- let read_only_completion = unresolved
- && actual.credential_leases <= 1
- && matches!(
- actual.verdict.outcome,
- RunOutcome::Unknown | RunOutcome::ResponseRecorded | RunOutcome::Observed
- );
- if actual.provider_entries != 0
- || (actual.credential_leases != 0 && !read_only_completion)
- {
- return Err(IssuanceError::CaseFailed);
- }
- }
- if step.operation != Operation::Oracle {
- if actual.provider_entries > 0 {
- unresolved = matches!(
- actual.verdict.outcome,
- RunOutcome::Unknown | RunOutcome::ResponseRecorded
- );
- } else if actual.verdict.outcome == RunOutcome::Observed {
- unresolved = false;
- }
+ // Replay may return historical evidence, but cannot count it as a
+ // new observation or authorize another credential lease or entry.
+ if step.operation == Operation::Replay
+ && (actual.provider_entries != 0
+ || actual.credential_leases != 0
+ || actual.confirmed_by_read_back != 0)
+ {
+ return Err(IssuanceError::CaseFailed);
}
}
if matches!(
diff --git a/product/qualification/auths-recipe-qualification-issuance/tests/execution.rs b/product/qualification/auths-recipe-qualification-issuance/tests/execution.rs
index 19c134b45..d48c135ae 100644
--- a/product/qualification/auths-recipe-qualification-issuance/tests/execution.rs
+++ b/product/qualification/auths-recipe-qualification-issuance/tests/execution.rs
@@ -136,7 +136,7 @@ fn commissioning_client_refusal_cannot_replace_an_ordinary_installed_effect() {
}
#[test]
-fn replay_can_complete_an_unresolved_read_back_but_never_write_or_reacquire_after_observation() {
+fn replay_returns_history_without_a_lease_entry_or_new_confirmation() {
use auths_recipe_qualification::BoundedText;
let mut case = executable_corpus()
.cases
@@ -144,26 +144,21 @@ fn replay_can_complete_an_unresolved_read_back_but_never_write_or_reacquire_afte
.find(|case| case.scenario == Scenario::AmbiguousResponse)
.expect("ambiguous response");
let replay = &mut case.steps[1].expected;
- replay.credential_leases = 1;
+ replay.credential_leases = 0;
replay.verdict.outcome = RunOutcome::Unknown;
replay.verdict.code = BoundedText::parse("unknown").expect("code");
assert!(run(&case, |_, _| {}).is_ok());
let replay = &mut case.steps[1].expected;
replay.verdict.outcome = RunOutcome::Observed;
replay.verdict.evidence_sha256 = Some(tuple().profile_lock_sha256);
- replay.confirmed_by_read_back = 1;
+ replay.confirmed_by_read_back = 0;
assert!(run(&case, |_, _| {}).is_ok());
for mutation in 0..3 {
let mut changed = case.clone();
match mutation {
0 => changed.steps[1].expected.provider_entries = 1,
- 1 => changed.steps[1].expected.credential_leases = 2,
- _ => {
- changed.steps[0].expected.verdict.outcome = RunOutcome::Observed;
- changed.steps[0].expected.verdict.evidence_sha256 =
- Some(tuple().profile_lock_sha256);
- changed.steps[0].expected.confirmed_by_read_back = 1;
- }
+ 1 => changed.steps[1].expected.credential_leases = 1,
+ _ => changed.steps[1].expected.confirmed_by_read_back = 1,
}
assert_eq!(run(&changed, |_, _| {}), Err(IssuanceError::CaseFailed));
}
diff --git a/product/qualification/auths-recipe-qualification-issuance/tests/script_pipeline.rs b/product/qualification/auths-recipe-qualification-issuance/tests/script_pipeline.rs
index 79a77aea8..f1e6b9028 100644
--- a/product/qualification/auths-recipe-qualification-issuance/tests/script_pipeline.rs
+++ b/product/qualification/auths-recipe-qualification-issuance/tests/script_pipeline.rs
@@ -70,7 +70,11 @@ fn stage_reports_flow_through_redaction_and_assembly_and_missing_execution_refus
}
succeed(Command::new("git").args(["init", "--quiet"]).arg(&root));
succeed(Command::new("git").arg("-C").arg(&root).args(["add", "."]));
+ // This disposable corpus needs a source revision, not the developer's
+ // signing identity. Qualification signing is exercised separately.
succeed(Command::new("git").arg("-C").arg(&root).args([
+ "-c",
+ "commit.gpgsign=false",
"-c",
"user.name=Qualification test",
"-c",
diff --git a/product/qualification/auths-recipe-qualification/src/vectors/mod.rs b/product/qualification/auths-recipe-qualification/src/vectors/mod.rs
index 47d3314c2..c2d1e5a55 100644
--- a/product/qualification/auths-recipe-qualification/src/vectors/mod.rs
+++ b/product/qualification/auths-recipe-qualification/src/vectors/mod.rs
@@ -280,7 +280,7 @@ pub(super) fn target() -> QualificationTarget {
gateway_version: bounded("1.0.0-rc.1"),
gateway_build_sha256: digest_of("gateway build"),
store_kind: LifecycleStoreKind::PostgresqlV1,
- store_schema: bounded("auths.lifecycle.postgresql/6"),
+ store_schema: bounded("auths.lifecycle.postgresql/8"),
credential_store_kind: CredentialStoreKind::AwsSecretsManagerV1,
}
}
diff --git a/product/runtime/auths-connections/src/credential.rs b/product/runtime/auths-connections/src/credential.rs
index 744c4bc84..e593575b8 100644
--- a/product/runtime/auths-connections/src/credential.rs
+++ b/product/runtime/auths-connections/src/credential.rs
@@ -6,11 +6,11 @@ use std::{
collections::BTreeMap,
convert::Infallible,
fmt,
- fs::{self, File},
- io::Write as _,
+ fs::{self, File, OpenOptions},
+ io::{Read as _, Write as _},
num::NonZeroU64,
path::{Path, PathBuf},
- sync::{Mutex, RwLock},
+ sync::{Mutex, MutexGuard, RwLock},
time::Instant,
};
use subtle::ConstantTimeEq as _;
@@ -304,6 +304,8 @@ pub struct InMemoryCredentialStore {
maximum_bytes: usize,
}
+type CredentialEntries = BTreeMap<(String, u64), StoredSecret>;
+
/// One retained generation. Every clone, including the copy-on-write map in
/// [`PersistentCredentialStore`], zeroizes its bytes when dropped.
#[derive(Clone)]
@@ -541,7 +543,10 @@ impl ConnectionCredentialStore for InMemoryCredentialStore {
/// only internal connection IDs, generations, commitment-bound secret bytes,
/// and no provider/account/profile metadata. Deployments that require an HSM,
/// OS keychain, or external secret manager can implement the same mechanism
-/// contract without changing callers.
+/// contract without changing callers. Every operation reloads bounded state
+/// under a stable cross-process file lock. File-lock contention fails immediately with
+/// `Unavailable`; stale handles cannot overwrite a successor or lease a
+/// generation already removed by a completed custody mutation.
pub struct PersistentCredentialStore {
path: PathBuf,
entries: Mutex>,
@@ -587,10 +592,7 @@ impl PersistentCredentialStore {
let parent = path.parent().ok_or(CredentialStoreError::InvalidCapacity)?;
validate_parent(parent)?;
let entries = if path.exists() {
- validate_secret_file(&path, maximum_bytes)?;
- let bytes =
- Zeroizing::new(fs::read(&path).map_err(|_| CredentialStoreError::Unavailable)?);
- decode_persistent_entries(&bytes, maximum_entries, maximum_bytes)?
+ read_persistent_entries(&path, maximum_entries, maximum_bytes)?
} else {
BTreeMap::new()
};
@@ -612,6 +614,8 @@ impl PersistentCredentialStore {
.entries
.lock()
.map_err(|_| CredentialStoreError::Unavailable)?;
+ let _writer = self.writer_lock()?;
+ self.refresh(&mut entries)?;
let mut next = entries.clone();
let result = mutation(&mut next)?;
persist_entries(&self.path, &next, self.maximum_bytes)?;
@@ -619,6 +623,80 @@ impl PersistentCredentialStore {
Ok(result)
}
+ /// Cross-process mutations use one stable sidecar, not the database inode
+ /// that atomic rename replaces. Contention fails immediately and cannot
+ /// acknowledge custody. A reader's snapshot is ordered with these writers.
+ fn writer_lock(&self) -> Result {
+ let parent = self
+ .path
+ .parent()
+ .ok_or(CredentialStoreError::UnsafeStorage)?;
+ validate_parent(parent)?;
+ let mut name = self.path.as_os_str().to_os_string();
+ name.push(".lock");
+ let path = PathBuf::from(name);
+ let mut options = OpenOptions::new();
+ options.read(true).write(true).create_new(true);
+ #[cfg(unix)]
+ {
+ use std::os::unix::fs::OpenOptionsExt as _;
+ options.mode(0o600);
+ }
+ let lock = match options.open(&path) {
+ Ok(lock) => lock,
+ Err(error) if error.kind() == std::io::ErrorKind::AlreadyExists => {
+ validate_secret_file(&path, 0)?;
+ OpenOptions::new()
+ .read(true)
+ .write(true)
+ .open(&path)
+ .map_err(|_| CredentialStoreError::Unavailable)?
+ }
+ Err(_) => return Err(CredentialStoreError::Unavailable),
+ };
+ lock.try_lock()
+ .map_err(|_| CredentialStoreError::Unavailable)?;
+ validate_secret_file(&path, 0)?;
+ #[cfg(unix)]
+ {
+ use std::os::unix::fs::MetadataExt as _;
+ let opened = lock
+ .metadata()
+ .map_err(|_| CredentialStoreError::Unavailable)?;
+ let named =
+ fs::symlink_metadata(&path).map_err(|_| CredentialStoreError::Unavailable)?;
+ if opened.dev() != named.dev() || opened.ino() != named.ino() || opened.nlink() != 1 {
+ return Err(CredentialStoreError::UnsafeStorage);
+ }
+ }
+ Ok(lock)
+ }
+
+ fn refresh(
+ &self,
+ entries: &mut BTreeMap<(String, u64), StoredSecret>,
+ ) -> Result<(), CredentialStoreError> {
+ match fs::symlink_metadata(&self.path) {
+ Err(error) if error.kind() == std::io::ErrorKind::NotFound && entries.is_empty() => {
+ return Ok(());
+ }
+ Err(_) => return Err(CredentialStoreError::Unavailable),
+ Ok(_) => {}
+ }
+ *entries = read_persistent_entries(&self.path, self.maximum_entries, self.maximum_bytes)?;
+ Ok(())
+ }
+
+ fn current_entries(&self) -> Result, CredentialStoreError> {
+ let mut entries = self
+ .entries
+ .lock()
+ .map_err(|_| CredentialStoreError::Unavailable)?;
+ let _writer = self.writer_lock()?;
+ self.refresh(&mut entries)?;
+ Ok(entries)
+ }
+
fn total_bytes(entries: &BTreeMap<(String, u64), StoredSecret>) -> usize {
entries.values().map(|entry| entry.bytes.len()).sum()
}
@@ -641,10 +719,7 @@ impl PersistentCredentialStore {
connection_id: &ConnectionId,
generation: NonZeroU64,
) -> Result {
- let entries = self
- .entries
- .lock()
- .map_err(|_| CredentialStoreError::Unavailable)?;
+ let entries = self.current_entries()?;
retained_entry(&entries, connection_id, generation)
.map(|(_, entry)| entry.commitment)
.ok_or(CredentialStoreError::Unavailable)
@@ -661,10 +736,7 @@ impl PersistentCredentialStore {
&self,
connection_id: &ConnectionId,
) -> Result, CredentialStoreError> {
- let entries = self
- .entries
- .lock()
- .map_err(|_| CredentialStoreError::Unavailable)?;
+ let entries = self.current_entries()?;
Ok(connection_generations(&entries, connection_id)
.filter_map(NonZeroU64::new)
.collect())
@@ -689,10 +761,7 @@ impl PersistentCredentialStore {
record: &ConnectionRecord,
deadline: Instant,
) -> Result {
- let entries = self
- .entries
- .lock()
- .map_err(|_| CredentialStoreError::Unavailable)?;
+ let entries = self.current_entries()?;
let stored = record_entry(&entries, record)?;
Ok(StoredSecretLease {
bytes: stored.bytes.clone(),
@@ -710,10 +779,7 @@ impl PersistentCredentialStore {
&self,
record: &ConnectionRecord,
) -> Result<(), CredentialStoreError> {
- let entries = self
- .entries
- .lock()
- .map_err(|_| CredentialStoreError::Unavailable)?;
+ let entries = self.current_entries()?;
record_entry(&entries, record).map(|_| ())
}
@@ -976,10 +1042,7 @@ impl ConnectionCredentialStore for PersistentCredentialStore {
binding: &CredentialBinding,
deadline: Instant,
) -> Result {
- let entries = self
- .entries
- .lock()
- .map_err(|_| CredentialStoreError::Unavailable)?;
+ let entries = self.current_entries()?;
let stored = binding_entry(&entries, binding)?;
Ok(StoredSecretLease {
bytes: stored.bytes.clone(),
@@ -1047,10 +1110,7 @@ impl ConnectionCredentialStore for PersistentCredentialStore {
}
async fn holds(&self, binding: &CredentialBinding) -> Result<(), CredentialStoreError> {
- let entries = self
- .entries
- .lock()
- .map_err(|_| CredentialStoreError::Unavailable)?;
+ let entries = self.current_entries()?;
binding_entry(&entries, binding).map(|_| ())
}
@@ -1124,6 +1184,37 @@ fn validate_secret_file(path: &Path, maximum_bytes: usize) -> Result<(), Credent
Ok(())
}
+fn read_persistent_entries(
+ path: &Path,
+ maximum_entries: usize,
+ maximum_bytes: usize,
+) -> Result, CredentialStoreError> {
+ validate_secret_file(path, maximum_bytes)?;
+ let mut file = File::open(path).map_err(|_| CredentialStoreError::Unavailable)?;
+ let length = usize::try_from(
+ file.metadata()
+ .map_err(|_| CredentialStoreError::Unavailable)?
+ .len(),
+ )
+ .map_err(|_| CredentialStoreError::Capacity)?;
+ if length > maximum_bytes {
+ return Err(CredentialStoreError::Capacity);
+ }
+ // Allocate once: growing a secret-bearing Vec can strand an unwiped copy.
+ let mut bytes = Zeroizing::new(vec![0; length]);
+ file.read_exact(&mut bytes)
+ .map_err(|_| CredentialStoreError::Unavailable)?;
+ let mut extra = Zeroizing::new([0_u8; 1]);
+ if file
+ .read(&mut *extra)
+ .map_err(|_| CredentialStoreError::Unavailable)?
+ != 0
+ {
+ return Err(CredentialStoreError::UnsafeStorage);
+ }
+ decode_persistent_entries(&bytes, maximum_entries, maximum_bytes)
+}
+
fn persist_entries(
path: &Path,
entries: &BTreeMap<(String, u64), StoredSecret>,
diff --git a/product/runtime/auths-gateway-kernel/src/order.rs b/product/runtime/auths-gateway-kernel/src/order.rs
index fdd14a801..747058ff0 100644
--- a/product/runtime/auths-gateway-kernel/src/order.rs
+++ b/product/runtime/auths-gateway-kernel/src/order.rs
@@ -7,18 +7,18 @@
//! read, a store write, and the provider write each happen only when the
//! machine names them, and the machine names them only in this order:
//!
-//! 1. read the clock, verify natively, admit (with the account-scope binding
-//! when declared), and prepare the transport, storing nothing;
-//! 2. claim the logical operation; a replay goes to one read-only
-//! re-observation and never to a write;
+//! 1. read the clock, verify native authority and exact effect approval, admit
+//! (with the account-scope binding when declared), and prepare the transport,
+//! revalidating the approved effect against that installation, storing nothing;
+//! 2. claim the logical operation; a replay returns retained state without I/O;
//! 3. reload the connection record, lease the credential, check the prefix,
//! the account read, and each denied read;
//! 4. run the pre-entry re-read and the relative-ceiling read, when declared,
//! and record what they read in one checkpoint;
-//! 5. reload the connection record again, refuse entry after the deadline,
-//! and send the one write;
-//! 6. record the response or `unknown`, then at most one read-back under a
-//! fresh lease that passes the same credential checks.
+//! 5. reload the connection record again, refuse entry after the deadline or
+//! failed fresh authority/approval validation, and send the one write;
+//! 6. record the response or `unknown`, then stop. Provider observation requires
+//! a separate approved acquisition and cannot be entered by this machine.
//!
//! Every refusal before the claim stores nothing; every refusal after it and
//! before transport entry records `not-entered` with its code. An event the
@@ -29,17 +29,6 @@
use crate::ratio::relative_ceiling_admits;
-/// Why a lease is taken.
-#[derive(Clone, Copy, Debug, Eq, PartialEq)]
-pub enum Mode {
- /// The lease for the one write.
- Entry,
- /// The lease for the read-back after a recorded response.
- ReadBack,
- /// The lease for a replay's read-only re-observation.
- Reobserve,
-}
-
/// The declarations and verified values the order depends on, fixed before
/// the machine starts.
#[derive(Clone, Copy, Debug, Eq, PartialEq)]
@@ -79,18 +68,16 @@ pub enum Phase {
Prepare,
/// Awaiting the claim.
Claim,
- /// Awaiting whether a replayed attempt can be re-observed.
- Resume,
/// Awaiting the connection reload before a lease.
- Reload(Mode),
+ Reload,
/// Awaiting the credential lease.
- Lease(Mode),
+ Lease,
/// Awaiting the prefix guard.
- Prefix(Mode),
+ Prefix,
/// Awaiting the account read.
- Account(Mode),
+ Account,
/// Awaiting the denied read at this index.
- Denied(Mode, u8),
+ Denied(u8),
/// Awaiting the pre-entry re-read.
PreEntry,
/// Awaiting the relative-ceiling read.
@@ -101,6 +88,8 @@ pub enum Phase {
EntryReload,
/// Awaiting the entry-deadline check.
Deadline,
+ /// Awaiting a newly acknowledged durable entry grant.
+ Grant,
/// Awaiting the provider write.
Send,
/// Awaiting the store write of the response.
@@ -109,8 +98,6 @@ pub enum Phase {
RecordUnknown,
/// Awaiting the store write of `not-entered`.
RecordNotEntered,
- /// Awaiting the read-back.
- ReadBack(Mode),
/// Stopped.
Done,
}
@@ -143,8 +130,8 @@ pub enum Verification {
/// The result of the claim.
#[derive(Clone, Copy, Debug, Eq, PartialEq)]
pub enum ClaimResult {
- /// This submission stored the claim with every slot.
- Inserted,
+ /// This submission holds a new or reopened reservation with every slot.
+ Reserved,
/// The claim step stored the claim as `not-entered`.
Refused,
/// The logical operation was already claimed.
@@ -203,8 +190,6 @@ pub enum CeilingRead {
/// The result of the provider write.
#[derive(Clone, Copy, Debug, Eq, PartialEq)]
pub enum WriteResult {
- /// Refused before network entry.
- NotEntered,
/// Transport may have been entered; no complete response.
Unknown,
/// A complete bounded response was received.
@@ -217,10 +202,7 @@ pub enum ResponseRecord {
/// The store write failed.
Failed,
/// The response was recorded.
- Recorded {
- /// Whether a read-back request exists for it.
- observable: bool,
- },
+ Recorded,
}
/// The result of the action the previous decision named.
@@ -240,8 +222,6 @@ pub enum SubmitEvent {
Preparation(bool),
/// The claim result.
Claim(ClaimResult),
- /// The replayed attempt can be re-observed.
- Resume(bool),
/// The connection record is unchanged.
Reload(bool),
/// The credential was leased.
@@ -260,6 +240,8 @@ pub enum SubmitEvent {
Recorded(bool),
/// Entry is within the deadline.
Deadline(bool),
+ /// A new durable grant was acknowledged and yielded its one-use permit.
+ Granted(bool),
/// The provider write result.
Write(WriteResult),
/// The response record result.
@@ -304,20 +286,20 @@ pub enum Refusal {
pub enum Stop {
/// Refused before the claim; nothing was stored.
Refused,
- /// The claim could not be stored; nothing was stored.
+ /// The claim was not acknowledged; durable state may be unavailable.
StoreUnavailable,
/// The claim step recorded `not-entered`.
ClaimRefused,
/// `not-entered` was recorded after the claim.
NotEntered,
- /// The stored record is `attempting` or `unknown`.
+ /// The stored record is `entry-granted` or `unknown`.
Unknown,
- /// The response is recorded and no read-back transition was.
+ /// The response is recorded; observation requires a separate acquisition.
Response,
- /// A read-back recorded its transition.
- Observed,
- /// A replay recorded nothing.
- ReplayRefused,
+ /// An identical replay returned only the matching journal state.
+ Stored,
+ /// A grant was not acknowledged; only a fresh journal read can report state.
+ GrantUnavailable,
/// An event the phase does not expect.
Halted,
}
@@ -337,28 +319,28 @@ pub enum SubmitAction {
Prepare,
/// Claim the logical operation.
Claim,
- /// Load the replayed attempt and decide whether it can be re-observed.
- Resume,
/// Reload the connection record before a lease.
- Reload(Mode),
+ Reload,
/// Lease the credential.
- Lease(Mode),
+ Lease,
/// Check the leased secret's prefix.
- CheckPrefix(Mode),
+ CheckPrefix,
/// Run the account read.
- ReadAccount(Mode),
+ ReadAccount,
/// Run the denied read at this index.
- DeniedRead(Mode, u8),
+ DeniedRead(u8),
/// Run the pre-entry re-read.
PreEntryRead,
/// Run the relative-ceiling read.
CeilingRead,
- /// Record what was read in one `attempting → attempting`.
+ /// Record what was read in one `reserved → reserved`.
RecordCheckpoint,
/// Reload the connection record before entry.
ReloadBeforeEntry,
/// Check the entry deadline.
CheckDeadline,
+ /// Validate and durably grant entry under the common writer fence.
+ GrantEntry,
/// Send the one write.
Send,
/// Record the response.
@@ -367,8 +349,6 @@ pub enum SubmitAction {
RecordUnknown,
/// Record `not-entered` with this refusal.
RecordNotEntered(Refusal),
- /// Perform one read-only observation and record what it shows.
- ReadBack(Mode),
/// Stop.
Stop(Stop),
}
@@ -427,17 +407,6 @@ pub const fn not_entered(state: SubmitState, refusal: Refusal) -> SubmitDecision
)
}
-/// A failed lease-time step: an entry records `not-entered`; a read-back or
-/// re-observation records nothing.
-#[must_use]
-pub const fn lease_failed(state: SubmitState, mode: Mode, refusal: Refusal) -> SubmitDecision {
- match mode {
- Mode::Entry => not_entered(state, refusal),
- Mode::ReadBack => stop(state, Stop::Response),
- Mode::Reobserve => stop(state, Stop::ReplayRefused),
- }
-}
-
/// After the relative ceiling, or after the pre-entry re-read when no
/// ceiling is declared: the checkpoint when anything was read, else the
/// reload before entry.
@@ -470,39 +439,23 @@ pub const fn pre_entry_from(state: SubmitState) -> SubmitDecision {
}
}
-/// After every credential check passed: the entry's reads, or the one
-/// read-back of a read-back or re-observation lease.
-#[must_use]
-pub const fn after_credentials(state: SubmitState, mode: Mode) -> SubmitDecision {
- match mode {
- Mode::Entry => pre_entry_from(state),
- Mode::ReadBack | Mode::Reobserve => {
- go(state, Phase::ReadBack(mode), SubmitAction::ReadBack(mode))
- }
- }
-}
-
/// The denied read at `index` when one remains.
#[must_use]
-pub const fn denied_from(state: SubmitState, mode: Mode, index: u8) -> SubmitDecision {
+pub const fn denied_from(state: SubmitState, index: u8) -> SubmitDecision {
if index < state.plan.denied_reads {
- go(
- state,
- Phase::Denied(mode, index),
- SubmitAction::DeniedRead(mode, index),
- )
+ go(state, Phase::Denied(index), SubmitAction::DeniedRead(index))
} else {
- after_credentials(state, mode)
+ pre_entry_from(state)
}
}
/// After the prefix guard: the account read when declared.
#[must_use]
-pub const fn after_prefix(state: SubmitState, mode: Mode) -> SubmitDecision {
+pub const fn after_prefix(state: SubmitState) -> SubmitDecision {
if state.plan.account_read {
- go(state, Phase::Account(mode), SubmitAction::ReadAccount(mode))
+ go(state, Phase::Account, SubmitAction::ReadAccount)
} else {
- denied_from(state, mode, 0)
+ denied_from(state, 0)
}
}
@@ -530,49 +483,36 @@ pub const fn pre_claim(state: SubmitState, passed: bool, next: SubmitDecision) -
#[must_use]
pub const fn after_claim(state: SubmitState, result: ClaimResult) -> SubmitDecision {
match result {
- ClaimResult::Inserted => go(
- state,
- Phase::Reload(Mode::Entry),
- SubmitAction::Reload(Mode::Entry),
- ),
+ ClaimResult::Reserved => go(state, Phase::Reload, SubmitAction::Reload),
ClaimResult::Refused => stop(state, Stop::ClaimRefused),
- ClaimResult::Replay => go(state, Phase::Resume, SubmitAction::Resume),
+ ClaimResult::Replay => stop(state, Stop::Stored),
ClaimResult::Unavailable => stop(state, Stop::StoreUnavailable),
}
}
/// The account read result.
#[must_use]
-pub const fn after_account(
- state: SubmitState,
- mode: Mode,
- result: AccountResult,
-) -> SubmitDecision {
+pub const fn after_account(state: SubmitState, result: AccountResult) -> SubmitDecision {
match result {
- AccountResult::Equal => denied_from(state, mode, 0),
- AccountResult::Mismatch => lease_failed(state, mode, Refusal::AccountMismatch),
- AccountResult::Unavailable => lease_failed(state, mode, Refusal::AccountUnavailable),
+ AccountResult::Equal => denied_from(state, 0),
+ AccountResult::Mismatch => not_entered(state, Refusal::AccountMismatch),
+ AccountResult::Unavailable => not_entered(state, Refusal::AccountUnavailable),
}
}
/// The result of the denied read at `index`.
#[must_use]
-pub const fn after_denied(
- state: SubmitState,
- mode: Mode,
- index: u8,
- result: DeniedResult,
-) -> SubmitDecision {
+pub const fn after_denied(state: SubmitState, index: u8, result: DeniedResult) -> SubmitDecision {
match result {
DeniedResult::Refused => {
if index < state.plan.denied_reads {
- denied_from(state, mode, index + 1)
+ denied_from(state, index + 1)
} else {
halt(state)
}
}
- DeniedResult::Answered => lease_failed(state, mode, Refusal::CapabilityExcess),
- DeniedResult::Unavailable => lease_failed(state, mode, Refusal::CapabilityUnavailable),
+ DeniedResult::Answered => not_entered(state, Refusal::CapabilityExcess),
+ DeniedResult::Unavailable => not_entered(state, Refusal::CapabilityUnavailable),
}
}
@@ -608,42 +548,17 @@ pub const fn after_ceiling(state: SubmitState, read: CeilingRead) -> SubmitDecis
#[must_use]
pub const fn after_write(state: SubmitState, result: WriteResult) -> SubmitDecision {
match result {
- WriteResult::NotEntered => not_entered(state, Refusal::TransportNotEntered),
WriteResult::Unknown => go(state, Phase::RecordUnknown, SubmitAction::RecordUnknown),
WriteResult::Response => go(state, Phase::RecordResponse, SubmitAction::RecordResponse),
}
}
-/// The response record result: at most one read-back, under a fresh lease.
+/// Recording a response stops the submission without granting observation I/O.
#[must_use]
pub const fn after_response(state: SubmitState, record: ResponseRecord) -> SubmitDecision {
match record {
ResponseRecord::Failed => stop(state, Stop::Unknown),
- ResponseRecord::Recorded { observable } => {
- if observable {
- go(
- state,
- Phase::Reload(Mode::ReadBack),
- SubmitAction::Reload(Mode::ReadBack),
- )
- } else {
- stop(state, Stop::Response)
- }
- }
- }
-}
-
-/// The read-back result.
-#[must_use]
-pub const fn after_read_back(state: SubmitState, mode: Mode, recorded: bool) -> SubmitDecision {
- if recorded {
- stop(state, Stop::Observed)
- } else {
- match mode {
- Mode::Entry => halt(state),
- Mode::ReadBack => stop(state, Stop::Response),
- Mode::Reobserve => stop(state, Stop::ReplayRefused),
- }
+ ResponseRecord::Recorded => stop(state, Stop::Response),
}
}
@@ -709,56 +624,42 @@ pub const fn pre_claim_step(state: SubmitState, event: SubmitEvent) -> SubmitDec
#[must_use]
pub const fn lease_step(state: SubmitState, event: SubmitEvent) -> SubmitDecision {
match state.phase {
- Phase::Resume => match event {
- SubmitEvent::Resume(resumable) => {
- if resumable {
- go(
- state,
- Phase::Reload(Mode::Reobserve),
- SubmitAction::Reload(Mode::Reobserve),
- )
- } else {
- stop(state, Stop::ReplayRefused)
- }
- }
- _ => halt(state),
- },
- Phase::Reload(mode) => match event {
+ Phase::Reload => match event {
SubmitEvent::Reload(unchanged) => {
if unchanged {
- go(state, Phase::Lease(mode), SubmitAction::Lease(mode))
+ go(state, Phase::Lease, SubmitAction::Lease)
} else {
- lease_failed(state, mode, Refusal::ConnectionChanged)
+ not_entered(state, Refusal::ConnectionChanged)
}
}
_ => halt(state),
},
- Phase::Lease(mode) => match event {
+ Phase::Lease => match event {
SubmitEvent::Lease(leased) => {
if leased {
- go(state, Phase::Prefix(mode), SubmitAction::CheckPrefix(mode))
+ go(state, Phase::Prefix, SubmitAction::CheckPrefix)
} else {
- lease_failed(state, mode, Refusal::CredentialUnavailable)
+ not_entered(state, Refusal::CredentialUnavailable)
}
}
_ => halt(state),
},
- Phase::Prefix(mode) => match event {
+ Phase::Prefix => match event {
SubmitEvent::Prefix(allowed) => {
if allowed {
- after_prefix(state, mode)
+ after_prefix(state)
} else {
- lease_failed(state, mode, Refusal::ModeGuard)
+ not_entered(state, Refusal::ModeGuard)
}
}
_ => halt(state),
},
- Phase::Account(mode) => match event {
- SubmitEvent::Account(result) => after_account(state, mode, result),
+ Phase::Account => match event {
+ SubmitEvent::Account(result) => after_account(state, result),
_ => halt(state),
},
- Phase::Denied(mode, index) => match event {
- SubmitEvent::Denied(result) => after_denied(state, mode, index, result),
+ Phase::Denied(index) => match event {
+ SubmitEvent::Denied(result) => after_denied(state, index, result),
_ => halt(state),
},
_ => halt(state),
@@ -767,7 +668,7 @@ pub const fn lease_step(state: SubmitState, event: SubmitEvent) -> SubmitDecisio
/// A step between the credential checks and the stop: the entry's reads,
/// the checkpoint, the reload and deadline before entry, the write, and the
-/// store writes and read-back after it.
+/// store writes after it.
#[must_use]
pub const fn entry_step(state: SubmitState, event: SubmitEvent) -> SubmitDecision {
match state.phase {
@@ -802,13 +703,18 @@ pub const fn entry_step(state: SubmitState, event: SubmitEvent) -> SubmitDecisio
Phase::Deadline => match event {
SubmitEvent::Deadline(within) => {
if within {
- go(state, Phase::Send, SubmitAction::Send)
+ go(state, Phase::Grant, SubmitAction::GrantEntry)
} else {
not_entered(state, Refusal::EntryDeadline)
}
}
_ => halt(state),
},
+ Phase::Grant => match event {
+ SubmitEvent::Granted(true) => go(state, Phase::Send, SubmitAction::Send),
+ SubmitEvent::Granted(false) => stop(state, Stop::GrantUnavailable),
+ _ => halt(state),
+ },
Phase::Send => match event {
SubmitEvent::Write(result) => after_write(state, result),
_ => halt(state),
@@ -831,10 +737,6 @@ pub const fn entry_step(state: SubmitState, event: SubmitEvent) -> SubmitDecisio
}
_ => halt(state),
},
- Phase::ReadBack(mode) => match event {
- SubmitEvent::Recorded(recorded) => after_read_back(state, mode, recorded),
- _ => halt(state),
- },
_ => halt(state),
}
}
@@ -852,22 +754,19 @@ pub const fn next_step(state: SubmitState, event: SubmitEvent) -> SubmitDecision
| Phase::Scope
| Phase::Prepare
| Phase::Claim => pre_claim_step(state, event),
- Phase::Resume
- | Phase::Reload(_)
- | Phase::Lease(_)
- | Phase::Prefix(_)
- | Phase::Account(_)
- | Phase::Denied(_, _) => lease_step(state, event),
+ Phase::Reload | Phase::Lease | Phase::Prefix | Phase::Account | Phase::Denied(_) => {
+ lease_step(state, event)
+ }
Phase::PreEntry
| Phase::Ceiling
| Phase::Checkpoint
| Phase::EntryReload
| Phase::Deadline
+ | Phase::Grant
| Phase::Send
| Phase::RecordResponse
| Phase::RecordUnknown
- | Phase::RecordNotEntered
- | Phase::ReadBack(_) => entry_step(state, event),
+ | Phase::RecordNotEntered => entry_step(state, event),
Phase::Done => halt(state),
}
}
@@ -905,7 +804,7 @@ mod tests {
SubmitEvent::Verification(Verification::Authorized { argument: 500 }),
SubmitEvent::Admission(true),
SubmitEvent::Preparation(true),
- SubmitEvent::Claim(ClaimResult::Inserted),
+ SubmitEvent::Claim(ClaimResult::Reserved),
SubmitEvent::Reload(true),
SubmitEvent::Lease(true),
SubmitEvent::Prefix(true),
@@ -926,18 +825,10 @@ mod tests {
SubmitEvent::Recorded(true),
SubmitEvent::Reload(true),
SubmitEvent::Deadline(true),
+ SubmitEvent::Granted(true),
SubmitEvent::Write(WriteResult::Response),
- SubmitEvent::Response(ResponseRecord::Recorded { observable: true }),
- SubmitEvent::Reload(true),
- SubmitEvent::Lease(true),
- SubmitEvent::Prefix(true),
- SubmitEvent::Account(AccountResult::Equal),
- SubmitEvent::Denied(DeniedResult::Refused),
- SubmitEvent::Denied(DeniedResult::Refused),
- SubmitEvent::Recorded(true),
+ SubmitEvent::Response(ResponseRecord::Recorded),
]);
- let entry = Mode::Entry;
- let back = Mode::ReadBack;
assert_eq!(
run(STRIPE, &events),
[
@@ -946,26 +837,20 @@ mod tests {
SubmitAction::Admit,
SubmitAction::Prepare,
SubmitAction::Claim,
- SubmitAction::Reload(entry),
- SubmitAction::Lease(entry),
- SubmitAction::CheckPrefix(entry),
- SubmitAction::ReadAccount(entry),
- SubmitAction::DeniedRead(entry, 0),
- SubmitAction::DeniedRead(entry, 1),
+ SubmitAction::Reload,
+ SubmitAction::Lease,
+ SubmitAction::CheckPrefix,
+ SubmitAction::ReadAccount,
+ SubmitAction::DeniedRead(0),
+ SubmitAction::DeniedRead(1),
SubmitAction::CeilingRead,
SubmitAction::RecordCheckpoint,
SubmitAction::ReloadBeforeEntry,
SubmitAction::CheckDeadline,
+ SubmitAction::GrantEntry,
SubmitAction::Send,
SubmitAction::RecordResponse,
- SubmitAction::Reload(back),
- SubmitAction::Lease(back),
- SubmitAction::CheckPrefix(back),
- SubmitAction::ReadAccount(back),
- SubmitAction::DeniedRead(back, 0),
- SubmitAction::DeniedRead(back, 1),
- SubmitAction::ReadBack(back),
- SubmitAction::Stop(Stop::Observed),
+ SubmitAction::Stop(Stop::Response),
]
);
}
@@ -1020,18 +905,15 @@ mod tests {
SubmitEvent::Admission(true),
SubmitEvent::Preparation(true),
SubmitEvent::Claim(ClaimResult::Replay),
- SubmitEvent::Resume(true),
- SubmitEvent::Reload(true),
- SubmitEvent::Lease(true),
- SubmitEvent::Prefix(true),
- SubmitEvent::Account(AccountResult::Mismatch),
];
let actions = run(STRIPE, &replay);
assert!(!actions.contains(&SubmitAction::Send));
- assert_eq!(
- actions.last(),
- Some(&SubmitAction::Stop(Stop::ReplayRefused))
+ assert!(
+ !actions
+ .iter()
+ .any(|action| matches!(action, SubmitAction::Lease))
);
+ assert_eq!(actions.last(), Some(&SubmitAction::Stop(Stop::Stored)));
let refused = run(
STRIPE,
&[
@@ -1045,6 +927,36 @@ mod tests {
assert!(!refused.contains(&SubmitAction::Claim));
}
+ #[test]
+ fn sending_requires_a_new_acknowledged_grant() {
+ let state = SubmitState {
+ plan: STRIPE,
+ phase: Phase::Grant,
+ argument: 500,
+ };
+ assert_eq!(
+ next_step(state, SubmitEvent::Granted(true)).action,
+ SubmitAction::Send
+ );
+ assert_eq!(
+ next_step(state, SubmitEvent::Granted(false)).action,
+ SubmitAction::Stop(Stop::GrantUnavailable)
+ );
+ assert_eq!(
+ next_step(state, SubmitEvent::Write(WriteResult::Response)).action,
+ SubmitAction::Stop(Stop::Halted)
+ );
+ let after = SubmitState {
+ plan: STRIPE,
+ phase: Phase::Send,
+ argument: 500,
+ };
+ assert_eq!(
+ next_step(after, SubmitEvent::Granted(true)).action,
+ SubmitAction::Stop(Stop::Halted)
+ );
+ }
+
#[test]
fn an_unexpected_event_halts_without_io() {
let decision = next_step(start(STRIPE), SubmitEvent::Write(WriteResult::Response));
diff --git a/product/runtime/auths-gateway-kernel/src/outcome.rs b/product/runtime/auths-gateway-kernel/src/outcome.rs
index 9a323fb68..2b459f801 100644
--- a/product/runtime/auths-gateway-kernel/src/outcome.rs
+++ b/product/runtime/auths-gateway-kernel/src/outcome.rs
@@ -42,13 +42,14 @@ pub struct OutcomeFacts {
pub relative_basis_digest: bool,
}
-/// Whether an outcome may name `stage`: every stage but `attempting`, which
-/// an outcome reports as `unknown`.
+/// Whether a provider outcome may name `stage`. Entry grants project unknown;
+/// recipient acknowledgement belongs to the distinct disclosure projection.
#[must_use]
pub const fn outcome_stage(stage: Stage) -> bool {
match stage {
- Stage::Attempting => false,
- Stage::NotEntered
+ Stage::EntryGranted | Stage::Acknowledged => false,
+ Stage::Reserved
+ | Stage::NotEntered
| Stage::Unknown
| Stage::ResponseRecorded
| Stage::Observed
@@ -122,7 +123,8 @@ mod tests {
..NONE
}
));
- assert!(!outcome_facts_present(Stage::Attempting, NONE));
+ assert!(!outcome_facts_present(Stage::EntryGranted, NONE));
+ assert!(!outcome_facts_present(Stage::Acknowledged, NONE));
}
#[test]
@@ -166,13 +168,15 @@ mod proofs {
use super::*;
fn any_stage() -> Stage {
- match kani::any::() % 6 {
- 0 => Stage::Attempting,
+ match kani::any::() % 8 {
+ 0 => Stage::EntryGranted,
1 => Stage::NotEntered,
2 => Stage::Unknown,
3 => Stage::ResponseRecorded,
4 => Stage::Observed,
- _ => Stage::ObservedByProvider,
+ 5 => Stage::Reserved,
+ 6 => Stage::ObservedByProvider,
+ _ => Stage::Acknowledged,
}
}
@@ -196,11 +200,11 @@ mod proofs {
let pairs = facts.http_status == facts.response_digest
&& facts.relative_basis == facts.relative_basis_digest;
let table = match stage {
- Stage::Attempting => false,
+ Stage::EntryGranted | Stage::Acknowledged => false,
Stage::NotEntered => {
facts.refusal && !facts.http_status && !facts.observation && !facts.evidence_digest
}
- Stage::Unknown => {
+ Stage::Reserved | Stage::Unknown => {
!facts.refusal && !facts.http_status && !facts.observation && !facts.evidence_digest
}
Stage::ResponseRecorded => {
diff --git a/product/runtime/auths-gateway-kernel/src/transition.rs b/product/runtime/auths-gateway-kernel/src/transition.rs
index d5064a032..a3a6dace5 100644
--- a/product/runtime/auths-gateway-kernel/src/transition.rs
+++ b/product/runtime/auths-gateway-kernel/src/transition.rs
@@ -14,11 +14,24 @@
use alloc::vec::Vec;
+/// Closed external boundary whose facts a journal record may carry.
+#[derive(Clone, Copy, Debug, Eq, PartialEq)]
+pub enum Protocol {
+ /// A provider mutation may be refined by separate observation evidence.
+ Mutation,
+ /// A bounded acquisition records a response, never mutation success.
+ Acquisition,
+ /// Recipient delivery records authenticated acknowledgement, never a provider response.
+ Disclosure,
+}
+
/// The stage of an attempt record.
#[derive(Clone, Copy, Debug, Eq, PartialEq)]
pub enum Stage {
- /// Claimed; transport may have been entered after the last checkpoint.
- Attempting,
+ /// Capacity is reserved, but external entry is forbidden.
+ Reserved,
+ /// A unique durable grant exists; external entry may have occurred.
+ EntryGranted,
/// Transport entry was excluded; terminal.
NotEntered,
/// Transport may have been entered and no complete response exists.
@@ -29,6 +42,8 @@ pub enum Stage {
Observed,
/// A read-back found this action's token and value; terminal.
ObservedByProvider,
+ /// The exact recipient acknowledged this delivery; terminal.
+ Acknowledged,
}
/// The read-back comparison recorded with an `observed` stage.
@@ -58,8 +73,14 @@ pub enum Link {
/// The fields of one stored attempt record the transition rule inspects.
#[derive(Clone, Debug, Eq, PartialEq)]
pub struct AttemptView {
+ /// Typed protocol; transitions cannot change it.
+ pub protocol: Protocol,
+ /// Monotonic journal revision, bounded to exact JSON integers.
+ pub revision: u64,
/// Canonical bytes of the fields no transition may change.
pub fixed: Vec,
+ /// Immutable canonical grant/audit bytes; empty before entry.
+ pub entry: Vec,
/// The stage.
pub stage: Stage,
/// Whether a refusal code is recorded.
@@ -68,6 +89,9 @@ pub struct AttemptView {
pub pre_entry: Vec,
/// Canonical bytes of the response status and digest; empty when absent.
pub response: Vec,
+ /// Canonical authenticated recipient evidence; empty until acknowledged.
+ /// The runtime verifies its binding and signature before projecting it here.
+ pub acknowledgement: Vec,
/// Canonical bytes of the response locator; empty when absent.
pub locator: Vec,
/// The recorded comparison.
@@ -113,7 +137,7 @@ pub fn same_bytes(left: &[u8], right: &[u8]) -> bool {
}
}
-/// Whether a link lets an `attempting` or `unknown` record reach
+/// Whether a link lets an `entry-granted` or `unknown` record reach
/// `observed-by-provider`: only a verified-locator link.
#[must_use]
pub const fn resolves_unknown(link: Link) -> bool {
@@ -137,11 +161,13 @@ pub const fn has_link(link: Link) -> bool {
pub const fn refusal_stage(stage: Stage) -> bool {
match stage {
Stage::NotEntered => true,
- Stage::Attempting
+ Stage::Reserved
+ | Stage::EntryGranted
| Stage::Unknown
| Stage::ResponseRecorded
| Stage::Observed
- | Stage::ObservedByProvider => false,
+ | Stage::ObservedByProvider
+ | Stage::Acknowledged => false,
}
}
@@ -150,7 +176,12 @@ pub const fn refusal_stage(stage: Stage) -> bool {
pub const fn requires_response(stage: Stage) -> bool {
match stage {
Stage::ResponseRecorded | Stage::Observed => true,
- Stage::Attempting | Stage::NotEntered | Stage::Unknown | Stage::ObservedByProvider => false,
+ Stage::Reserved
+ | Stage::EntryGranted
+ | Stage::NotEntered
+ | Stage::Unknown
+ | Stage::ObservedByProvider
+ | Stage::Acknowledged => false,
}
}
@@ -159,7 +190,11 @@ pub const fn requires_response(stage: Stage) -> bool {
pub const fn permits_response(stage: Stage) -> bool {
match stage {
Stage::ResponseRecorded | Stage::Observed | Stage::ObservedByProvider => true,
- Stage::Attempting | Stage::NotEntered | Stage::Unknown => false,
+ Stage::Reserved
+ | Stage::EntryGranted
+ | Stage::NotEntered
+ | Stage::Unknown
+ | Stage::Acknowledged => false,
}
}
@@ -168,11 +203,13 @@ pub const fn permits_response(stage: Stage) -> bool {
pub const fn observed_stage(stage: Stage) -> bool {
match stage {
Stage::Observed => true,
- Stage::Attempting
+ Stage::Reserved
+ | Stage::EntryGranted
| Stage::NotEntered
| Stage::Unknown
| Stage::ResponseRecorded
- | Stage::ObservedByProvider => false,
+ | Stage::ObservedByProvider
+ | Stage::Acknowledged => false,
}
}
@@ -181,11 +218,13 @@ pub const fn observed_stage(stage: Stage) -> bool {
pub const fn provider_stage(stage: Stage) -> bool {
match stage {
Stage::ObservedByProvider => true,
- Stage::Attempting
+ Stage::Reserved
+ | Stage::EntryGranted
| Stage::NotEntered
| Stage::Unknown
| Stage::ResponseRecorded
- | Stage::Observed => false,
+ | Stage::Observed
+ | Stage::Acknowledged => false,
}
}
@@ -193,8 +232,10 @@ pub const fn provider_stage(stage: Stage) -> bool {
#[must_use]
pub const fn terminal_stage(stage: Stage) -> bool {
match stage {
- Stage::NotEntered | Stage::Observed | Stage::ObservedByProvider => true,
- Stage::Attempting | Stage::Unknown | Stage::ResponseRecorded => false,
+ Stage::NotEntered | Stage::Observed | Stage::ObservedByProvider | Stage::Acknowledged => {
+ true
+ }
+ Stage::Reserved | Stage::EntryGranted | Stage::Unknown | Stage::ResponseRecorded => false,
}
}
@@ -216,7 +257,12 @@ pub const fn has_reading(reading: Reading) -> bool {
#[must_use]
pub fn consistent(view: &AttemptView) -> bool {
let has_response = !absent(&view.response);
- refusal_stage(view.stage) == view.refusal
+ protocol_consistent(view)
+ && view.revision > 0
+ && view.revision <= 9_007_199_254_740_991
+ && requires_entry(view.stage) != absent(&view.entry)
+ && acknowledgement_stage(view.stage) != absent(&view.acknowledgement)
+ && refusal_stage(view.stage) == view.refusal
&& (!requires_response(view.stage) || has_response)
&& (!has_response || permits_response(view.stage))
&& (absent(&view.locator) || has_response)
@@ -227,96 +273,134 @@ pub fn consistent(view: &AttemptView) -> bool {
&& (!view.evidence || has_response || resolves_unknown(view.link))
}
-/// The closed stage table: which stage may follow which, given the stored
-/// plan's link. `attempting → attempting` is the single checkpoint of
-/// pre-entry evidence; an `attempting` or `unknown` record reaches
-/// `observed-by-provider` only for a verified-locator link.
+/// Whether a stage requires authenticated recipient evidence.
+#[must_use]
+pub const fn acknowledgement_stage(stage: Stage) -> bool {
+ matches!(stage, Stage::Acknowledged)
+}
+
+/// Acquisition and disclosure cannot acquire each other's evidence or mutation
+/// observations. A common stage does not make their external boundaries equal.
+#[must_use]
+pub fn protocol_consistent(view: &AttemptView) -> bool {
+ match view.protocol {
+ Protocol::Mutation => !acknowledgement_stage(view.stage) && absent(&view.acknowledgement),
+ Protocol::Acquisition => {
+ matches!(
+ view.stage,
+ Stage::Reserved
+ | Stage::EntryGranted
+ | Stage::NotEntered
+ | Stage::Unknown
+ | Stage::ResponseRecorded
+ ) && absent(&view.acknowledgement)
+ && absent(&view.pre_entry)
+ && absent(&view.locator)
+ && !view.observation
+ && !view.evidence
+ && matches!(view.reading, Reading::None)
+ && matches!(view.link, Link::None)
+ }
+ Protocol::Disclosure => {
+ matches!(
+ view.stage,
+ Stage::Reserved
+ | Stage::EntryGranted
+ | Stage::NotEntered
+ | Stage::Unknown
+ | Stage::Acknowledged
+ ) && absent(&view.response)
+ && absent(&view.pre_entry)
+ && absent(&view.locator)
+ && !view.observation
+ && !view.evidence
+ && matches!(view.reading, Reading::None)
+ && matches!(view.link, Link::None)
+ }
+ }
+}
+
+/// Protocol identity is immutable independently of the encoded fixed fields.
+#[must_use]
+pub const fn same_protocol(left: Protocol, right: Protocol) -> bool {
+ matches!(
+ (left, right),
+ (Protocol::Mutation, Protocol::Mutation)
+ | (Protocol::Acquisition, Protocol::Acquisition)
+ | (Protocol::Disclosure, Protocol::Disclosure)
+ )
+}
+
+/// Whether the state requires an immutable entry grant.
+#[must_use]
+pub const fn requires_entry(stage: Stage) -> bool {
+ match stage {
+ Stage::Reserved | Stage::NotEntered => false,
+ Stage::EntryGranted
+ | Stage::Unknown
+ | Stage::ResponseRecorded
+ | Stage::Observed
+ | Stage::ObservedByProvider
+ | Stage::Acknowledged => true,
+ }
+}
+
+/// The one transition table: a reservation cannot acquire response evidence
+/// or become unknown until it has received a durable entry grant. Protocol
+/// consistency separately confines ACKs to disclosure and responses to providers.
#[must_use]
pub const fn stage_transition_allowed(from: Stage, to: Stage, link: Link) -> bool {
match from {
- Stage::Attempting => match to {
- Stage::Attempting | Stage::NotEntered | Stage::Unknown | Stage::ResponseRecorded => {
- true
- }
+ Stage::Reserved => matches!(
+ to,
+ Stage::Reserved | Stage::NotEntered | Stage::EntryGranted
+ ),
+ Stage::EntryGranted => match to {
+ Stage::Unknown | Stage::ResponseRecorded | Stage::Acknowledged => true,
Stage::ObservedByProvider => resolves_unknown(link),
- Stage::Observed => false,
+ _ => false,
},
Stage::Unknown => match to {
+ Stage::ResponseRecorded | Stage::Acknowledged => true,
Stage::ObservedByProvider => resolves_unknown(link),
- Stage::Attempting
- | Stage::NotEntered
- | Stage::Unknown
- | Stage::ResponseRecorded
- | Stage::Observed => false,
+ _ => false,
},
- Stage::ResponseRecorded => match to {
- Stage::Observed | Stage::ObservedByProvider => true,
- Stage::Attempting | Stage::NotEntered | Stage::Unknown | Stage::ResponseRecorded => {
- false
- }
- },
- Stage::NotEntered | Stage::Observed | Stage::ObservedByProvider => false,
+ Stage::ResponseRecorded => matches!(to, Stage::Observed | Stage::ObservedByProvider),
+ Stage::NotEntered | Stage::Observed | Stage::ObservedByProvider | Stage::Acknowledged => {
+ false
+ }
}
}
-/// Whether a transition from `from` to `to` may add pre-entry evidence:
-/// only the checkpoint and a refusal before transport entry.
+/// Only a reservation may add pre-entry observations.
#[must_use]
pub const fn adds_pre_entry(from: Stage, to: Stage) -> bool {
- match from {
- Stage::Attempting => match to {
- Stage::Attempting | Stage::NotEntered => true,
- Stage::Unknown
- | Stage::ResponseRecorded
- | Stage::Observed
- | Stage::ObservedByProvider => false,
- },
- Stage::NotEntered
- | Stage::Unknown
- | Stage::ResponseRecorded
- | Stage::Observed
- | Stage::ObservedByProvider => false,
- }
+ matches!(from, Stage::Reserved) && matches!(to, Stage::Reserved | Stage::NotEntered)
}
-/// Whether a transition from `from` to `to` records the response.
+/// Only an existing grant may acquire a late or immediate response.
#[must_use]
pub const fn adds_response(from: Stage, to: Stage) -> bool {
- match from {
- Stage::Attempting => match to {
- Stage::ResponseRecorded => true,
- Stage::Attempting
- | Stage::NotEntered
- | Stage::Unknown
- | Stage::Observed
- | Stage::ObservedByProvider => false,
- },
- Stage::NotEntered
- | Stage::Unknown
- | Stage::ResponseRecorded
- | Stage::Observed
- | Stage::ObservedByProvider => false,
- }
+ matches!(from, Stage::EntryGranted | Stage::Unknown) && matches!(to, Stage::ResponseRecorded)
}
-/// Whether the checkpoint condition holds: `attempting → attempting` adds
-/// pre-entry evidence exactly once and changes nothing else.
+/// A reservation checkpoint adds its pre-entry record exactly once.
#[must_use]
pub fn checkpoint_valid(old: &AttemptView, new: &AttemptView) -> bool {
- match old.stage {
- Stage::Attempting => match new.stage {
- Stage::Attempting => absent(&old.pre_entry) && !absent(&new.pre_entry),
- Stage::NotEntered
- | Stage::Unknown
- | Stage::ResponseRecorded
- | Stage::Observed
- | Stage::ObservedByProvider => true,
- },
- Stage::NotEntered
- | Stage::Unknown
- | Stage::ResponseRecorded
- | Stage::Observed
- | Stage::ObservedByProvider => true,
+ if matches!(old.stage, Stage::Reserved) && matches!(new.stage, Stage::Reserved) {
+ absent(&old.pre_entry) && !absent(&new.pre_entry)
+ } else {
+ true
+ }
+}
+
+/// Grant bytes are added only at the single reservation-to-entry transition.
+#[must_use]
+pub fn entry_kept(old: &AttemptView, new: &AttemptView) -> bool {
+ if matches!(old.stage, Stage::Reserved) && matches!(new.stage, Stage::EntryGranted) {
+ absent(&old.entry) && !absent(&new.entry)
+ } else {
+ same_bytes(&old.entry, &new.entry)
}
}
@@ -339,6 +423,31 @@ pub fn response_kept(old: &AttemptView, new: &AttemptView) -> bool {
|| (same_bytes(&old.response, &new.response) && same_bytes(&old.locator, &new.locator))
}
+/// Authenticated ACK bytes may be added only after a grant, including late
+/// evidence for an unknown delivery. They can never be replaced or removed.
+#[must_use]
+pub fn acknowledgement_kept(old: &AttemptView, new: &AttemptView) -> bool {
+ if matches!(old.stage, Stage::EntryGranted | Stage::Unknown) && acknowledgement_stage(new.stage)
+ {
+ absent(&old.acknowledgement) && !absent(&new.acknowledgement)
+ } else {
+ same_bytes(&old.acknowledgement, &new.acknowledgement)
+ }
+}
+
+/// Exact repeated evidence is an observation of existing history, not a state
+/// transition. Callers return the retained revision and must not mint a permit,
+/// charge again, rewrite evidence or consult current execution authority.
+/// Authentication and exact effect/source/recipient/channel/content matching
+/// precede this comparison at the runtime boundary.
+#[must_use]
+pub fn exact_acknowledgement_replay(view: &AttemptView, acknowledgement: &[u8]) -> bool {
+ matches!(view.protocol, Protocol::Disclosure)
+ && acknowledgement_stage(view.stage)
+ && consistent(view)
+ && same_bytes(&view.acknowledgement, acknowledgement)
+}
+
/// Whether the stored plan's summary is unchanged.
#[must_use]
pub const fn plan_kept(old: &AttemptView, new: &AttemptView) -> bool {
@@ -371,12 +480,18 @@ pub const fn link_equal(left: Link, right: Link) -> bool {
/// consistent with its stage.
#[must_use]
pub fn valid_transition(old: &AttemptView, new: &AttemptView) -> bool {
- same_bytes(&old.fixed, &new.fixed)
+ same_protocol(old.protocol, new.protocol)
+ && old.revision < 9_007_199_254_740_991
+ && new.revision == old.revision + 1
+ && consistent(old)
+ && entry_kept(old, new)
+ && same_bytes(&old.fixed, &new.fixed)
&& plan_kept(old, new)
&& stage_transition_allowed(old.stage, new.stage, old.link)
&& checkpoint_valid(old, new)
&& pre_entry_kept(old, new)
&& response_kept(old, new)
+ && acknowledgement_kept(old, new)
&& consistent(new)
}
@@ -385,13 +500,15 @@ mod tests {
use super::*;
use alloc::vec;
- const STAGES: [Stage; 6] = [
- Stage::Attempting,
+ const STAGES: [Stage; 8] = [
+ Stage::Reserved,
+ Stage::EntryGranted,
Stage::NotEntered,
Stage::Unknown,
Stage::ResponseRecorded,
Stage::Observed,
Stage::ObservedByProvider,
+ Stage::Acknowledged,
];
fn view(stage: Stage, link: Link) -> AttemptView {
@@ -401,11 +518,19 @@ mod tests {
Vec::new()
};
AttemptView {
+ protocol: Protocol::Mutation,
+ revision: 1,
+ entry: if requires_entry(stage) {
+ alloc::vec![7]
+ } else {
+ Vec::new()
+ },
fixed: vec![7; 4],
stage,
refusal: refusal_stage(stage),
pre_entry: Vec::new(),
response,
+ acknowledgement: Vec::new(),
locator: Vec::new(),
reading: if observed_stage(stage) {
Reading::Match
@@ -418,6 +543,64 @@ mod tests {
}
}
+ fn transition(old: &AttemptView, new: &AttemptView) -> bool {
+ let mut next = new.clone();
+ next.revision = old.revision + 1;
+ valid_transition(old, &next)
+ }
+
+ #[test]
+ fn acquisition_cannot_gain_mutation_facts_or_change_protocol() {
+ let acquisition = |stage| {
+ let mut value = view(stage, Link::None);
+ value.protocol = Protocol::Acquisition;
+ value.observation = false;
+ value
+ };
+ let reserved = acquisition(Stage::Reserved);
+ let granted = acquisition(Stage::EntryGranted);
+ let response = acquisition(Stage::ResponseRecorded);
+ assert!(transition(&reserved, &granted));
+ assert!(transition(&granted, &response));
+ assert!(!transition(
+ &reserved,
+ &view(Stage::EntryGranted, Link::None)
+ ));
+ assert!(!transition(&view(Stage::Reserved, Link::None), &granted));
+ for stage in [Stage::Observed, Stage::ObservedByProvider] {
+ assert!(!consistent(&acquisition(stage)));
+ assert!(!transition(&response, &acquisition(stage)));
+ }
+ let mut checkpoint = reserved.clone();
+ checkpoint.pre_entry = vec![9];
+ assert!(!consistent(&checkpoint));
+ assert!(!transition(&reserved, &checkpoint));
+ let mut changed = response.clone();
+ changed.locator = vec![1];
+ assert!(!transition(&granted, &changed));
+ changed = response.clone();
+ changed.observation = true;
+ assert!(!consistent(&changed));
+ changed = response.clone();
+ changed.link = Link::Verified;
+ assert!(!consistent(&changed));
+ }
+
+ #[test]
+ fn revisions_advance_once_and_never_wrap() {
+ let old = view(Stage::Reserved, Link::None);
+ let mut new = view(Stage::EntryGranted, Link::None);
+ assert!(!valid_transition(&old, &new));
+ new.revision = 2;
+ assert!(valid_transition(&old, &new));
+ new.revision = 3;
+ assert!(!valid_transition(&old, &new));
+ let mut saturated = old.clone();
+ saturated.revision = 9_007_199_254_740_991;
+ new.revision = 9_007_199_254_740_992;
+ assert!(!valid_transition(&saturated, &new));
+ }
+
#[test]
fn terminal_stages_have_no_outgoing_transition() {
for link in [Link::None, Link::Verified, Link::AfterResponse] {
@@ -425,7 +608,7 @@ mod tests {
for to in STAGES {
if terminal_stage(from) {
assert!(!stage_transition_allowed(from, to, link));
- assert!(!valid_transition(&view(from, link), &view(to, link)));
+ assert!(!transition(&view(from, link), &view(to, link)));
}
}
}
@@ -436,10 +619,11 @@ mod tests {
fn unknown_resolves_only_for_a_verified_link_with_evidence() {
for link in [Link::None, Link::Verified, Link::AfterResponse] {
for to in STAGES {
- let allowed = valid_transition(&view(Stage::Unknown, link), &view(to, link));
+ let allowed = transition(&view(Stage::Unknown, link), &view(to, link));
assert_eq!(
allowed,
- to == Stage::ObservedByProvider && link == Link::Verified,
+ to == Stage::ResponseRecorded
+ || (to == Stage::ObservedByProvider && link == Link::Verified),
"{link:?} {to:?}"
);
}
@@ -448,28 +632,31 @@ mod tests {
#[test]
fn checkpoint_adds_pre_entry_once_and_changes_nothing_else() {
- let old = view(Stage::Attempting, Link::Verified);
+ let old = view(Stage::Reserved, Link::Verified);
let mut checkpoint = old.clone();
checkpoint.pre_entry = vec![9];
- assert!(valid_transition(&old, &checkpoint));
- assert!(!valid_transition(&checkpoint, &checkpoint));
- assert!(
- !valid_transition(&old, &old),
- "a checkpoint must add evidence"
- );
+ assert!(transition(&old, &checkpoint));
+ assert!(!transition(&checkpoint, &checkpoint));
+ assert!(!transition(&old, &old), "a checkpoint must add evidence");
let mut refused = view(Stage::NotEntered, Link::Verified);
refused.pre_entry = vec![9];
- assert!(valid_transition(&old, &refused));
- assert!(valid_transition(&checkpoint, &refused));
+ assert!(transition(&old, &refused));
+ assert!(transition(&checkpoint, &refused));
refused.pre_entry = vec![8];
- assert!(!valid_transition(&checkpoint, &refused), "kept once set");
+ assert!(!transition(&checkpoint, &refused), "kept once set");
let mut responded = view(Stage::ResponseRecorded, Link::Verified);
responded.pre_entry = vec![9];
- assert!(
- !valid_transition(&old, &responded),
- "added only before entry"
- );
- assert!(valid_transition(&checkpoint, &responded));
+ assert!(!transition(&old, &responded), "added only before entry");
+ assert!(!transition(&checkpoint, &responded));
+ let mut grant = view(Stage::EntryGranted, Link::Verified);
+ grant.pre_entry = checkpoint.pre_entry.clone();
+ assert!(transition(&checkpoint, &grant));
+ assert!(transition(&grant, &responded));
+ assert!(!transition(&grant, &refused));
+ assert!(!transition(&grant, &grant), "no second grant");
+ let mut changed = responded.clone();
+ changed.entry = vec![8];
+ assert!(!transition(&grant, &changed), "grant is immutable");
}
#[test]
@@ -477,20 +664,20 @@ mod tests {
let old = view(Stage::ResponseRecorded, Link::AfterResponse);
let mut moved = view(Stage::ObservedByProvider, Link::AfterResponse);
moved.response = vec![1];
- assert!(valid_transition(&old, &moved));
+ assert!(transition(&old, &moved));
moved.fixed = vec![7; 5];
- assert!(!valid_transition(&old, &moved));
+ assert!(!transition(&old, &moved));
let mut changed = view(Stage::Observed, Link::AfterResponse);
changed.response = vec![2];
- assert!(!valid_transition(&old, &changed));
- let attempting = view(Stage::Attempting, Link::AfterResponse);
- assert!(!valid_transition(
+ assert!(!transition(&old, &changed));
+ let attempting = view(Stage::EntryGranted, Link::AfterResponse);
+ assert!(!transition(
&attempting,
&view(Stage::ObservedByProvider, Link::AfterResponse)
));
let mut located = view(Stage::ResponseRecorded, Link::AfterResponse);
located.locator = vec![3];
- assert!(valid_transition(&attempting, &located));
+ assert!(transition(&attempting, &located));
}
}
@@ -500,13 +687,15 @@ mod proofs {
use alloc::vec::Vec;
fn any_stage() -> Stage {
- match kani::any::() % 6 {
- 0 => Stage::Attempting,
+ match kani::any::() % 8 {
+ 0 => Stage::EntryGranted,
1 => Stage::NotEntered,
2 => Stage::Unknown,
3 => Stage::ResponseRecorded,
4 => Stage::Observed,
- _ => Stage::ObservedByProvider,
+ 5 => Stage::Reserved,
+ 6 => Stage::ObservedByProvider,
+ _ => Stage::Acknowledged,
}
}
@@ -539,11 +728,19 @@ mod proofs {
fn any_view() -> AttemptView {
AttemptView {
+ protocol: match kani::any::() % 3 {
+ 0 => Protocol::Mutation,
+ 1 => Protocol::Acquisition,
+ _ => Protocol::Disclosure,
+ },
+ revision: kani::any(),
+ entry: any_bytes(),
fixed: any_bytes(),
stage: any_stage(),
refusal: kani::any(),
pre_entry: any_bytes(),
response: any_bytes(),
+ acknowledgement: any_bytes(),
locator: any_bytes(),
reading: any_reading(),
evidence: kani::any(),
@@ -557,8 +754,14 @@ mod proofs {
matches!(
(from, to),
(
- Stage::Attempting,
- Stage::Attempting | Stage::NotEntered | Stage::Unknown | Stage::ResponseRecorded
+ Stage::EntryGranted,
+ Stage::Unknown | Stage::ResponseRecorded | Stage::Acknowledged
+ ) | (
+ Stage::Reserved,
+ Stage::Reserved | Stage::NotEntered | Stage::EntryGranted
+ ) | (
+ Stage::Unknown,
+ Stage::ResponseRecorded | Stage::Acknowledged
) | (
Stage::ResponseRecorded,
Stage::Observed | Stage::ObservedByProvider
@@ -566,7 +769,7 @@ mod proofs {
) || (matches!(
(from, to),
(
- Stage::Attempting | Stage::Unknown,
+ Stage::EntryGranted | Stage::Unknown,
Stage::ObservedByProvider
)
) && link == Link::Verified)
@@ -578,9 +781,10 @@ mod proofs {
let reading = view.reading != Reading::None;
let shape = match view.stage {
Stage::NotEntered => view.refusal && !response && !reading && !view.evidence,
- Stage::Attempting | Stage::Unknown => {
+ Stage::Reserved | Stage::EntryGranted | Stage::Unknown => {
!view.refusal && !response && !reading && !view.evidence
}
+ Stage::Acknowledged => !view.refusal && !response && !reading && !view.evidence,
Stage::ResponseRecorded => !view.refusal && response && !reading && !view.evidence,
Stage::Observed => {
!view.refusal && response && reading && view.observation && !view.evidence
@@ -593,7 +797,46 @@ mod proofs {
&& (response || view.link == Link::Verified)
}
};
- shape && (view.locator.is_empty() || response)
+ let protocol = match view.protocol {
+ Protocol::Mutation => {
+ view.stage != Stage::Acknowledged && view.acknowledgement.is_empty()
+ }
+ Protocol::Acquisition => {
+ !matches!(
+ view.stage,
+ Stage::Observed | Stage::ObservedByProvider | Stage::Acknowledged
+ ) && view.acknowledgement.is_empty()
+ && view.pre_entry.is_empty()
+ && view.locator.is_empty()
+ && view.reading == Reading::None
+ && !view.evidence
+ && !view.observation
+ && view.link == Link::None
+ }
+ Protocol::Disclosure => {
+ matches!(
+ view.stage,
+ Stage::Reserved
+ | Stage::EntryGranted
+ | Stage::NotEntered
+ | Stage::Unknown
+ | Stage::Acknowledged
+ ) && view.response.is_empty()
+ && view.pre_entry.is_empty()
+ && view.locator.is_empty()
+ && view.reading == Reading::None
+ && !view.evidence
+ && !view.observation
+ && view.link == Link::None
+ }
+ };
+ shape
+ && protocol
+ && ((view.stage == Stage::Acknowledged) == !view.acknowledgement.is_empty())
+ && view.revision > 0
+ && view.revision <= 9_007_199_254_740_991
+ && (view.locator.is_empty() || response)
+ && (matches!(view.stage, Stage::Reserved | Stage::NotEntered) == view.entry.is_empty())
}
/// Exhaustive over every stage, link, and reading pair: the closed stage
@@ -612,6 +855,20 @@ mod proofs {
}
}
+ /// Historical ACK replay is exact and read-only; all providers and all
+ /// pre-ACK stages are refused even if byte fields are otherwise equal.
+ #[kani::proof]
+ #[kani::unwind(3)]
+ fn acknowledgement_replay_requires_exact_terminal_disclosure() {
+ let view = any_view();
+ let evidence = any_bytes();
+ let expected = view.protocol == Protocol::Disclosure
+ && view.stage == Stage::Acknowledged
+ && fields_match_stage(&view)
+ && view.acknowledgement == evidence;
+ assert_eq!(exact_acknowledgement_replay(&view, &evidence), expected);
+ }
+
/// Exhaustive over every pair of bounded records (each byte field absent
/// or one of two values): `valid_transition` holds exactly when the
/// fixed bytes and plan are kept, the stage change is in the table, the
@@ -625,24 +882,45 @@ mod proofs {
let old = any_view();
let new = any_view();
let pre_entry = match (old.stage, new.stage) {
- (Stage::Attempting, Stage::Attempting) => {
+ (Stage::Reserved, Stage::Reserved) => {
old.pre_entry.is_empty() && !new.pre_entry.is_empty()
}
- (Stage::Attempting, Stage::NotEntered) => {
+ (Stage::Reserved, Stage::NotEntered) => {
old.pre_entry.is_empty() || old.pre_entry == new.pre_entry
}
_ => old.pre_entry == new.pre_entry,
};
let response = matches!(
(old.stage, new.stage),
- (Stage::Attempting, Stage::ResponseRecorded)
+ (
+ Stage::EntryGranted | Stage::Unknown,
+ Stage::ResponseRecorded
+ )
) || (old.response == new.response && old.locator == new.locator);
- let expected = old.fixed == new.fixed
+ let entry = if old.stage == Stage::Reserved && new.stage == Stage::EntryGranted {
+ old.entry.is_empty() && !new.entry.is_empty()
+ } else {
+ old.entry == new.entry
+ };
+ let acknowledgement = if matches!(old.stage, Stage::EntryGranted | Stage::Unknown)
+ && new.stage == Stage::Acknowledged
+ {
+ old.acknowledgement.is_empty() && !new.acknowledgement.is_empty()
+ } else {
+ old.acknowledgement == new.acknowledgement
+ };
+ let expected = old.revision < 9_007_199_254_740_991
+ && new.revision == old.revision + 1
+ && old.fixed == new.fixed
+ && old.protocol == new.protocol
+ && entry
+ && fields_match_stage(&old)
&& old.observation == new.observation
&& old.link == new.link
&& table(old.stage, new.stage, old.link)
&& pre_entry
&& response
+ && acknowledgement
&& fields_match_stage(&new);
assert_eq!(valid_transition(&old, &new), expected);
}
diff --git a/product/runtime/auths-gateway/Cargo.toml b/product/runtime/auths-gateway/Cargo.toml
index 682d8611c..23e613565 100644
--- a/product/runtime/auths-gateway/Cargo.toml
+++ b/product/runtime/auths-gateway/Cargo.toml
@@ -8,8 +8,8 @@ description = "Closed developer-authored request recipes for the Auths gateway"
publish = false
[features]
-# Development and CI only: lets `serve --loopback-provider ` send
-# provider requests to a plain-HTTP provider double on 127.0.0.1.
+# Development and CI only: lets serve/candidate-acquire select a plain-HTTP
+# provider double on 127.0.0.1 with --loopback-provider.
loopback-provider = []
testkit-harness = []
# Tests only: lets a production installation keep its provider secret in the
@@ -63,6 +63,7 @@ base64ct.workspace = true
clap.workspace = true
ed25519-dalek.workspace = true
hex.workspace = true
+hmac.workspace = true
minicbor.workspace = true
getrandom.workspace = true
rustix.workspace = true
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_app_frame/content b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_app_frame/content
new file mode 100644
index 000000000..2d5b23d0b
--- /dev/null
+++ b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_app_frame/content
@@ -0,0 +1 @@
+{"schema":"auths.gateway-content/1","submission":{"schema":"auths.gateway-submit/2","proof_b64":"AA","action_b64":"AA","approval_b64":"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"}}
\ No newline at end of file
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_app_frame/content-history b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_app_frame/content-history
new file mode 100644
index 000000000..7a52a4a2b
--- /dev/null
+++ b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_app_frame/content-history
@@ -0,0 +1 @@
+{"schema":"auths.gateway-content-history/1","operation":"fixture-1"}
\ No newline at end of file
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_app_frame/disclose b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_app_frame/disclose
new file mode 100644
index 000000000..dfb556a2e
--- /dev/null
+++ b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_app_frame/disclose
@@ -0,0 +1 @@
+{"schema":"auths.gateway-disclose/1","target":{"source_operation":"fixture-source","recipient_key":[19,19,19,19,19,19,19,19,19,19,19,19,19,19,19,19,19,19,19,19,19,19,19,19,19,19,19,19,19,19,19,19],"channel":{"schema":"auths.gateway-unix-disclosure-channel/1","gateway_uid":100,"recipient_uid":101,"channel_id":[19,19,19,19,19,19,19,19,19,19,19,19,19,19,19,19,19,19,19,19,19,19,19,19,19,19,19,19,19,19,19,19]},"delivery_instance":[20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20]},"submission":{"schema":"auths.gateway-submit/2","proof_b64":"AA","action_b64":"AA","approval_b64":"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"}}
\ No newline at end of file
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_app_frame/disclosure-ack b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_app_frame/disclosure-ack
new file mode 100644
index 000000000..154033b79
--- /dev/null
+++ b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_app_frame/disclosure-ack
@@ -0,0 +1 @@
+{"schema":"auths.gateway-disclosure-acknowledge/1","operation":"fixture-1","acknowledgement_b64":"eyJyZXF1ZXN0Ijp7ImFja25vd2xlZGdlZF9ieXRlcyI6NDIsImNoYW5uZWwiOlsxMiwxMiwxMiwxMiwxMiwxMiwxMiwxMiwxMiwxMiwxMiwxMiwxMiwxMiwxMiwxMiwxMiwxMiwxMiwxMiwxMiwxMiwxMiwxMiwxMiwxMiwxMiwxMiwxMiwxMiwxMiwxMl0sImRlbGl2ZXJ5X2luc3RhbmNlIjpbMTMsMTMsMTMsMTMsMTMsMTMsMTMsMTMsMTMsMTMsMTMsMTMsMTMsMTMsMTMsMTMsMTMsMTMsMTMsMTMsMTMsMTMsMTMsMTMsMTMsMTMsMTMsMTMsMTMsMTMsMTMsMTNdLCJlZmZlY3QiOlsxODAsMTkwLDIzNSwzNSwyMzcsMTI1LDI1NCw5MSwyMiwyMzUsMjQxLDE4NywxMzksMTExLDExLDUsMjM0LDIyNiwxMTMsMTY0LDEwNywxMTEsMjM5LDg2LDIyNSw2OSwzNSwyMzIsMjQ2LDIxMywyNDcsMTY5XSwibWF4aW11bV9ieXRlcyI6MTAwLCJwcm9qZWN0aW9uIjpbMTQsMTQsMTQsMTQsMTQsMTQsMTQsMTQsMTQsMTQsMTQsMTQsMTQsMTQsMTQsMTQsMTQsMTQsMTQsMTQsMTQsMTQsMTQsMTQsMTQsMTQsMTQsMTQsMTQsMTQsMTQsMTRdLCJwcm90ZWN0ZWRfY29udGVudCI6WzExLDExLDExLDExLDExLDExLDExLDExLDExLDExLDExLDExLDExLDExLDExLDExLDExLDExLDExLDExLDExLDExLDExLDExLDExLDExLDExLDExLDExLDExLDExLDExXSwicmVjaXBpZW50Ijoia2V5OnNoYTI1NjprZUNTNXVtdjRnZkpUejJpMDFvcGZIaDZXTWo5RjZrYzF0c3lleXFvaTdnIiwicmVjaXBpZW50X2tleSI6WzY3LDQsMTA3LDI1NCw2NCwxNDYsMTc5LDIzMyw3MywxNDgsMjM0LDIxOCwyMSwyMjAsMTk0LDEzLDEzOCwxNzAsNywxODIsODgsMjUzLDU3LDg0LDIzNSwxNDIsMTQsMjUxLDEzOSwyMjAsMTY1LDIyMl0sInNjaGVtYSI6ImF1dGhzLmdhdGV3YXktZGlzY2xvc3VyZS1hY2svMSIsInNvdXJjZV9lZmZlY3QiOlsxMCwxMCwxMCwxMCwxMCwxMCwxMCwxMCwxMCwxMCwxMCwxMCwxMCwxMCwxMCwxMCwxMCwxMCwxMCwxMCwxMCwxMCwxMCwxMCwxMCwxMCwxMCwxMCwxMCwxMCwxMCwxMF19LCJzaWduYXR1cmVfYjY0IjoiMlZOR1Z6NmVjd3NDeUM0bGdiZkFxcl94b1FkUDhFM0UySEFmYmJJV29tNWNLTlpPSVdLSzFmeTBXaDZsRWozRWduaTN6NFVRT3N3SVE2bmZ5U1RmQmcifQ"}
\ No newline at end of file
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_app_frame/observe-outcome b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_app_frame/observe-outcome
new file mode 100644
index 000000000..051425019
--- /dev/null
+++ b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_app_frame/observe-outcome
@@ -0,0 +1 @@
+{"schema":"auths.gateway-observe/4","request":{"kind":"outcome","operation_id":"refund-1"}}
\ No newline at end of file
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_app_frame/observe-pre-entry b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_app_frame/observe-pre-entry
index 0629d67fd..7a7a6ff47 100644
--- a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_app_frame/observe-pre-entry
+++ b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_app_frame/observe-pre-entry
@@ -1 +1 @@
-{"schema": "auths.gateway-observe/2", "request": {"kind": "pre-entry", "operation_id": "refund-1"}}
\ No newline at end of file
+{"schema":"auths.gateway-observe/4","request":{"kind":"pre-entry","operation_id":"refund-1"}}
\ No newline at end of file
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_app_frame/admin-reobserve b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_app_frame/obsolete-admin-reobserve
similarity index 100%
rename from product/runtime/auths-gateway/fuzz/corpus/target_gateway_app_frame/admin-reobserve
rename to product/runtime/auths-gateway/fuzz/corpus/target_gateway_app_frame/obsolete-admin-reobserve
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_app_frame/obsolete-observe-v3 b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_app_frame/obsolete-observe-v3
new file mode 100644
index 000000000..440d2527a
--- /dev/null
+++ b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_app_frame/obsolete-observe-v3
@@ -0,0 +1 @@
+{"schema": "auths.gateway-observe/3", "request": {"kind": "pre-entry", "operation_id": "refund-1"}}
\ No newline at end of file
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_app_frame/obsolete-read-back b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_app_frame/obsolete-read-back
new file mode 100644
index 000000000..1bac3a815
--- /dev/null
+++ b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_app_frame/obsolete-read-back
@@ -0,0 +1 @@
+{"schema":"auths.gateway-observe/4","request":{"kind":"read-back","arguments":{"record_id":"recTEST0000000001"}}}
\ No newline at end of file
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_app_frame/obsolete-submit b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_app_frame/obsolete-submit
new file mode 100644
index 000000000..5eada08fd
--- /dev/null
+++ b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_app_frame/obsolete-submit
@@ -0,0 +1 @@
+{"schema": "auths.gateway-submit/1", "proof_b64": "AA", "action_b64": "AA"}
\ No newline at end of file
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_app_frame/recover b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_app_frame/recover
new file mode 100644
index 000000000..cfddcfd6b
--- /dev/null
+++ b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_app_frame/recover
@@ -0,0 +1 @@
+{"schema":"auths.gateway-recover/1","source_operation":"source-1","submission":{"schema":"auths.gateway-submit/2","proof_b64":"AA","action_b64":"AA","approval_b64":"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"}}
\ No newline at end of file
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_app_frame/submit b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_app_frame/submit
index 5eada08fd..733f8be8a 100644
--- a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_app_frame/submit
+++ b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_app_frame/submit
@@ -1 +1 @@
-{"schema": "auths.gateway-submit/1", "proof_b64": "AA", "action_b64": "AA"}
\ No newline at end of file
+{"schema":"auths.gateway-submit/2","proof_b64":"AA","action_b64":"AA","approval_b64":"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"}
\ No newline at end of file
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/record-not-entered-refund-5-window b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/obsolete-record-not-entered-refund-5-window
similarity index 100%
rename from product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/record-not-entered-refund-5-window
rename to product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/obsolete-record-not-entered-refund-5-window
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/record-response-recorded-refund-1 b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/obsolete-record-response-recorded-refund-1
similarity index 100%
rename from product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/record-response-recorded-refund-1
rename to product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/obsolete-record-response-recorded-refund-1
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/record-response-recorded-refund-4 b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/obsolete-record-response-recorded-refund-4
similarity index 100%
rename from product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/record-response-recorded-refund-4
rename to product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/obsolete-record-response-recorded-refund-4
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/transition-attempting-to-not-entered-refund-5-window b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/obsolete-transition-attempting-to-not-entered-refund-5-window
similarity index 100%
rename from product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/transition-attempting-to-not-entered-refund-5-window
rename to product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/obsolete-transition-attempting-to-not-entered-refund-5-window
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/transition-attempting-to-response-recorded-refund-1 b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/obsolete-transition-attempting-to-response-recorded-refund-1
similarity index 100%
rename from product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/transition-attempting-to-response-recorded-refund-1
rename to product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/obsolete-transition-attempting-to-response-recorded-refund-1
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/transition-attempting-to-response-recorded-refund-4 b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/obsolete-transition-attempting-to-response-recorded-refund-4
similarity index 100%
rename from product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/transition-attempting-to-response-recorded-refund-4
rename to product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/obsolete-transition-attempting-to-response-recorded-refund-4
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/record-entry-granted-synthetic-entry b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/record-entry-granted-synthetic-entry
new file mode 100644
index 000000000..1b8507450
--- /dev/null
+++ b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/record-entry-granted-synthetic-entry
@@ -0,0 +1 @@
+{"revision":2,"schema":"auths.gateway-attempt/16","binding_canonical":"{\"action\":[11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11],\"approval_policy\":{\"approver\":[232,218,99,164,12,166,135,200,124,252,224,92,178,74,120,108,126,117,204,73,199,13,181,87,63,2,111,28,106,134,206,170],\"kind\":\"exact\"},\"compiler\":\"auths.gateway-effect-compiler/8\",\"configuration\":[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],\"connection\":[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],\"credential_generation\":1,\"epoch\":1,\"namespace\":\"airtable-demo\",\"native_authority\":[18,235,131,109,161,47,186,164,207,117,158,84,83,81,226,82,123,165,214,134,91,192,31,155,84,121,98,93,248,158,220,213],\"operation\":\"synthetic-entry\",\"partition\":\"test-partition\",\"protocol\":{\"kind\":\"mutation\",\"request\":[133,246,13,83,160,52,144,78,110,249,236,145,242,255,85,229,203,206,104,215,97,230,81,91,198,238,222,43,194,199,159,214]},\"qualification_policy\":\"optional\",\"recipe\":[100,255,80,4,185,117,142,47,179,116,233,214,180,67,239,27,53,170,0,196,0,219,225,53,139,144,239,137,197,201,237,169],\"reservations\":[],\"schema\":\"auths.bound-effect/3\",\"scope\":{\"kind\":\"unscoped\"}}","authority_canonical_b64":"sgAAAQQCogAAAWphdXRob3JpemVkA1ggAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAEWCAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAVYIAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAABvYHWCAKhJq0RdktVd4THsBIP8vGv9Mb8AiIH854B9rXBwvtyAiACYAKgAunAAABAAIAAwAEAAUABgAMWCAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA32DlggAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAPBBCAEYA=","approval_canonical":"{\"request\":{\"approver\":[232,218,99,164,12,166,135,200,124,252,224,92,178,74,120,108,126,117,204,73,199,13,181,87,63,2,111,28,106,134,206,170],\"effect\":[218,111,158,133,196,21,24,92,131,17,221,40,251,90,102,151,153,204,156,92,154,223,40,214,213,222,12,92,245,160,150,252],\"expires_at\":1790003600,\"not_before\":1790000000,\"schema\":\"auths.exact-effect-approval/1\"},\"signature_b64\":\"qLYR1rtS2s4I4BRH3O2W6D6LAwOm8GDI6o1XBGsQtDBRnh4BZcKt4KakVAq5x2gcIigaqdkof557aY0Via0JCQ\"}","namespace":"airtable-demo","operation_id":"synthetic-entry","action_commitment":"0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b","recipe_digest":"64ff5004b9758e2fb374e9d6b443ef1b35aa00c400dbe1358b90ef89c5c9eda9","nonce":"00112233445566778899aabbccddeeff","evaluated_at":1790000000,"entry":{"revision":2,"authorization_started_at":1790000000,"approval_canonical":"{\"request\":{\"approver\":[232,218,99,164,12,166,135,200,124,252,224,92,178,74,120,108,126,117,204,73,199,13,181,87,63,2,111,28,106,134,206,170],\"effect\":[218,111,158,133,196,21,24,92,131,17,221,40,251,90,102,151,153,204,156,92,154,223,40,214,213,222,12,92,245,160,150,252],\"expires_at\":1790003600,\"not_before\":1790000000,\"schema\":\"auths.exact-effect-approval/1\"},\"signature_b64\":\"qLYR1rtS2s4I4BRH3O2W6D6LAwOm8GDI6o1XBGsQtDBRnh4BZcKt4KakVAq5x2gcIigaqdkof557aY0Via0JCQ\"}","pre_entry":null,"schema":"auths.effect-entry/3","granted_at":1790000000,"effect_id":"da6f9e85c415185c8311dd28fb5a669799cc9c5c9adf28d6d5de0c5cf5a096fc","dependencies":[]},"state":{"protocol":"mutation","facts":{"stage":"entry-granted","observation_plan":{"origin":"https://api.airtable.com","segments":[{"kind":"resolved","value":"v0"},{"kind":"resolved","value":"appTEST0000000001"},{"kind":"resolved","value":"tblTEST0000000001"},{"kind":"resolved","value":"recTEST0000000001"}],"headers":[],"json_pointer":"/fields/DemoStatus","expected":"\"Approved\"","maximum_response_bytes":16384,"echo_pointer":"/fields/auths_echo"}}}}
\ No newline at end of file
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/record-not-entered-synthetic-entry b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/record-not-entered-synthetic-entry
new file mode 100644
index 000000000..bc0cbd7ff
--- /dev/null
+++ b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/record-not-entered-synthetic-entry
@@ -0,0 +1 @@
+{"revision":2,"schema":"auths.gateway-attempt/16","binding_canonical":"{\"action\":[11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11],\"approval_policy\":{\"approver\":[232,218,99,164,12,166,135,200,124,252,224,92,178,74,120,108,126,117,204,73,199,13,181,87,63,2,111,28,106,134,206,170],\"kind\":\"exact\"},\"compiler\":\"auths.gateway-effect-compiler/8\",\"configuration\":[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],\"connection\":[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],\"credential_generation\":1,\"epoch\":1,\"namespace\":\"airtable-demo\",\"native_authority\":[18,235,131,109,161,47,186,164,207,117,158,84,83,81,226,82,123,165,214,134,91,192,31,155,84,121,98,93,248,158,220,213],\"operation\":\"synthetic-entry\",\"partition\":\"test-partition\",\"protocol\":{\"kind\":\"mutation\",\"request\":[133,246,13,83,160,52,144,78,110,249,236,145,242,255,85,229,203,206,104,215,97,230,81,91,198,238,222,43,194,199,159,214]},\"qualification_policy\":\"optional\",\"recipe\":[100,255,80,4,185,117,142,47,179,116,233,214,180,67,239,27,53,170,0,196,0,219,225,53,139,144,239,137,197,201,237,169],\"reservations\":[],\"schema\":\"auths.bound-effect/3\",\"scope\":{\"kind\":\"unscoped\"}}","authority_canonical_b64":"sgAAAQQCogAAAWphdXRob3JpemVkA1ggAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAEWCAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAVYIAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAABvYHWCAKhJq0RdktVd4THsBIP8vGv9Mb8AiIH854B9rXBwvtyAiACYAKgAunAAABAAIAAwAEAAUABgAMWCAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA32DlggAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAPBBCAEYA=","approval_canonical":"{\"request\":{\"approver\":[232,218,99,164,12,166,135,200,124,252,224,92,178,74,120,108,126,117,204,73,199,13,181,87,63,2,111,28,106,134,206,170],\"effect\":[218,111,158,133,196,21,24,92,131,17,221,40,251,90,102,151,153,204,156,92,154,223,40,214,213,222,12,92,245,160,150,252],\"expires_at\":1790003600,\"not_before\":1790000000,\"schema\":\"auths.exact-effect-approval/1\"},\"signature_b64\":\"qLYR1rtS2s4I4BRH3O2W6D6LAwOm8GDI6o1XBGsQtDBRnh4BZcKt4KakVAq5x2gcIigaqdkof557aY0Via0JCQ\"}","namespace":"airtable-demo","operation_id":"synthetic-entry","action_commitment":"0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b","recipe_digest":"64ff5004b9758e2fb374e9d6b443ef1b35aa00c400dbe1358b90ef89c5c9eda9","nonce":"00112233445566778899aabbccddeeff","evaluated_at":1790000000,"entry":null,"state":{"protocol":"mutation","facts":{"stage":"not-entered","refusal":"test.before-grant","observation_plan":{"origin":"https://api.airtable.com","segments":[{"kind":"resolved","value":"v0"},{"kind":"resolved","value":"appTEST0000000001"},{"kind":"resolved","value":"tblTEST0000000001"},{"kind":"resolved","value":"recTEST0000000001"}],"headers":[],"json_pointer":"/fields/DemoStatus","expected":"\"Approved\"","maximum_response_bytes":16384,"echo_pointer":"/fields/auths_echo"}}}}
\ No newline at end of file
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/record-reserved-synthetic-entry b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/record-reserved-synthetic-entry
new file mode 100644
index 000000000..6ed0371f9
--- /dev/null
+++ b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/record-reserved-synthetic-entry
@@ -0,0 +1 @@
+{"revision":1,"schema":"auths.gateway-attempt/16","binding_canonical":"{\"action\":[11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11],\"approval_policy\":{\"approver\":[232,218,99,164,12,166,135,200,124,252,224,92,178,74,120,108,126,117,204,73,199,13,181,87,63,2,111,28,106,134,206,170],\"kind\":\"exact\"},\"compiler\":\"auths.gateway-effect-compiler/8\",\"configuration\":[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],\"connection\":[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],\"credential_generation\":1,\"epoch\":1,\"namespace\":\"airtable-demo\",\"native_authority\":[18,235,131,109,161,47,186,164,207,117,158,84,83,81,226,82,123,165,214,134,91,192,31,155,84,121,98,93,248,158,220,213],\"operation\":\"synthetic-entry\",\"partition\":\"test-partition\",\"protocol\":{\"kind\":\"mutation\",\"request\":[133,246,13,83,160,52,144,78,110,249,236,145,242,255,85,229,203,206,104,215,97,230,81,91,198,238,222,43,194,199,159,214]},\"qualification_policy\":\"optional\",\"recipe\":[100,255,80,4,185,117,142,47,179,116,233,214,180,67,239,27,53,170,0,196,0,219,225,53,139,144,239,137,197,201,237,169],\"reservations\":[],\"schema\":\"auths.bound-effect/3\",\"scope\":{\"kind\":\"unscoped\"}}","authority_canonical_b64":"sgAAAQQCogAAAWphdXRob3JpemVkA1ggAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAEWCAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAVYIAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAABvYHWCAKhJq0RdktVd4THsBIP8vGv9Mb8AiIH854B9rXBwvtyAiACYAKgAunAAABAAIAAwAEAAUABgAMWCAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA32DlggAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAPBBCAEYA=","approval_canonical":"{\"request\":{\"approver\":[232,218,99,164,12,166,135,200,124,252,224,92,178,74,120,108,126,117,204,73,199,13,181,87,63,2,111,28,106,134,206,170],\"effect\":[218,111,158,133,196,21,24,92,131,17,221,40,251,90,102,151,153,204,156,92,154,223,40,214,213,222,12,92,245,160,150,252],\"expires_at\":1790003600,\"not_before\":1790000000,\"schema\":\"auths.exact-effect-approval/1\"},\"signature_b64\":\"qLYR1rtS2s4I4BRH3O2W6D6LAwOm8GDI6o1XBGsQtDBRnh4BZcKt4KakVAq5x2gcIigaqdkof557aY0Via0JCQ\"}","namespace":"airtable-demo","operation_id":"synthetic-entry","action_commitment":"0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b","recipe_digest":"64ff5004b9758e2fb374e9d6b443ef1b35aa00c400dbe1358b90ef89c5c9eda9","nonce":"00112233445566778899aabbccddeeff","evaluated_at":1790000000,"entry":null,"state":{"protocol":"mutation","facts":{"stage":"reserved","observation_plan":{"origin":"https://api.airtable.com","segments":[{"kind":"resolved","value":"v0"},{"kind":"resolved","value":"appTEST0000000001"},{"kind":"resolved","value":"tblTEST0000000001"},{"kind":"resolved","value":"recTEST0000000001"}],"headers":[],"json_pointer":"/fields/DemoStatus","expected":"\"Approved\"","maximum_response_bytes":16384,"echo_pointer":"/fields/auths_echo"}}}}
\ No newline at end of file
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/record-response-recorded-synthetic-entry b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/record-response-recorded-synthetic-entry
new file mode 100644
index 000000000..764783e08
--- /dev/null
+++ b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/record-response-recorded-synthetic-entry
@@ -0,0 +1 @@
+{"revision":3,"schema":"auths.gateway-attempt/16","binding_canonical":"{\"action\":[11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11],\"approval_policy\":{\"approver\":[232,218,99,164,12,166,135,200,124,252,224,92,178,74,120,108,126,117,204,73,199,13,181,87,63,2,111,28,106,134,206,170],\"kind\":\"exact\"},\"compiler\":\"auths.gateway-effect-compiler/8\",\"configuration\":[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],\"connection\":[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],\"credential_generation\":1,\"epoch\":1,\"namespace\":\"airtable-demo\",\"native_authority\":[18,235,131,109,161,47,186,164,207,117,158,84,83,81,226,82,123,165,214,134,91,192,31,155,84,121,98,93,248,158,220,213],\"operation\":\"synthetic-entry\",\"partition\":\"test-partition\",\"protocol\":{\"kind\":\"mutation\",\"request\":[133,246,13,83,160,52,144,78,110,249,236,145,242,255,85,229,203,206,104,215,97,230,81,91,198,238,222,43,194,199,159,214]},\"qualification_policy\":\"optional\",\"recipe\":[100,255,80,4,185,117,142,47,179,116,233,214,180,67,239,27,53,170,0,196,0,219,225,53,139,144,239,137,197,201,237,169],\"reservations\":[],\"schema\":\"auths.bound-effect/3\",\"scope\":{\"kind\":\"unscoped\"}}","authority_canonical_b64":"sgAAAQQCogAAAWphdXRob3JpemVkA1ggAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAEWCAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAVYIAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAABvYHWCAKhJq0RdktVd4THsBIP8vGv9Mb8AiIH854B9rXBwvtyAiACYAKgAunAAABAAIAAwAEAAUABgAMWCAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA32DlggAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAPBBCAEYA=","approval_canonical":"{\"request\":{\"approver\":[232,218,99,164,12,166,135,200,124,252,224,92,178,74,120,108,126,117,204,73,199,13,181,87,63,2,111,28,106,134,206,170],\"effect\":[218,111,158,133,196,21,24,92,131,17,221,40,251,90,102,151,153,204,156,92,154,223,40,214,213,222,12,92,245,160,150,252],\"expires_at\":1790003600,\"not_before\":1790000000,\"schema\":\"auths.exact-effect-approval/1\"},\"signature_b64\":\"qLYR1rtS2s4I4BRH3O2W6D6LAwOm8GDI6o1XBGsQtDBRnh4BZcKt4KakVAq5x2gcIigaqdkof557aY0Via0JCQ\"}","namespace":"airtable-demo","operation_id":"synthetic-entry","action_commitment":"0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b","recipe_digest":"64ff5004b9758e2fb374e9d6b443ef1b35aa00c400dbe1358b90ef89c5c9eda9","nonce":"00112233445566778899aabbccddeeff","evaluated_at":1790000000,"entry":{"revision":2,"authorization_started_at":1790000000,"approval_canonical":"{\"request\":{\"approver\":[232,218,99,164,12,166,135,200,124,252,224,92,178,74,120,108,126,117,204,73,199,13,181,87,63,2,111,28,106,134,206,170],\"effect\":[218,111,158,133,196,21,24,92,131,17,221,40,251,90,102,151,153,204,156,92,154,223,40,214,213,222,12,92,245,160,150,252],\"expires_at\":1790003600,\"not_before\":1790000000,\"schema\":\"auths.exact-effect-approval/1\"},\"signature_b64\":\"qLYR1rtS2s4I4BRH3O2W6D6LAwOm8GDI6o1XBGsQtDBRnh4BZcKt4KakVAq5x2gcIigaqdkof557aY0Via0JCQ\"}","pre_entry":null,"schema":"auths.effect-entry/3","granted_at":1790000000,"effect_id":"da6f9e85c415185c8311dd28fb5a669799cc9c5c9adf28d6d5de0c5cf5a096fc","dependencies":[]},"state":{"protocol":"mutation","facts":{"stage":"response-recorded","response_status":200,"response_digest":"0909090909090909090909090909090909090909090909090909090909090909","observation_plan":{"origin":"https://api.airtable.com","segments":[{"kind":"resolved","value":"v0"},{"kind":"resolved","value":"appTEST0000000001"},{"kind":"resolved","value":"tblTEST0000000001"},{"kind":"resolved","value":"recTEST0000000001"}],"headers":[],"json_pointer":"/fields/DemoStatus","expected":"\"Approved\"","maximum_response_bytes":16384,"echo_pointer":"/fields/auths_echo"}}}}
\ No newline at end of file
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/record-unknown-synthetic-entry b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/record-unknown-synthetic-entry
new file mode 100644
index 000000000..ee7c96f6f
--- /dev/null
+++ b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/record-unknown-synthetic-entry
@@ -0,0 +1 @@
+{"revision":3,"schema":"auths.gateway-attempt/16","binding_canonical":"{\"action\":[11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11],\"approval_policy\":{\"approver\":[232,218,99,164,12,166,135,200,124,252,224,92,178,74,120,108,126,117,204,73,199,13,181,87,63,2,111,28,106,134,206,170],\"kind\":\"exact\"},\"compiler\":\"auths.gateway-effect-compiler/8\",\"configuration\":[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],\"connection\":[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],\"credential_generation\":1,\"epoch\":1,\"namespace\":\"airtable-demo\",\"native_authority\":[18,235,131,109,161,47,186,164,207,117,158,84,83,81,226,82,123,165,214,134,91,192,31,155,84,121,98,93,248,158,220,213],\"operation\":\"synthetic-entry\",\"partition\":\"test-partition\",\"protocol\":{\"kind\":\"mutation\",\"request\":[133,246,13,83,160,52,144,78,110,249,236,145,242,255,85,229,203,206,104,215,97,230,81,91,198,238,222,43,194,199,159,214]},\"qualification_policy\":\"optional\",\"recipe\":[100,255,80,4,185,117,142,47,179,116,233,214,180,67,239,27,53,170,0,196,0,219,225,53,139,144,239,137,197,201,237,169],\"reservations\":[],\"schema\":\"auths.bound-effect/3\",\"scope\":{\"kind\":\"unscoped\"}}","authority_canonical_b64":"sgAAAQQCogAAAWphdXRob3JpemVkA1ggAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAEWCAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAVYIAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAABvYHWCAKhJq0RdktVd4THsBIP8vGv9Mb8AiIH854B9rXBwvtyAiACYAKgAunAAABAAIAAwAEAAUABgAMWCAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA32DlggAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAPBBCAEYA=","approval_canonical":"{\"request\":{\"approver\":[232,218,99,164,12,166,135,200,124,252,224,92,178,74,120,108,126,117,204,73,199,13,181,87,63,2,111,28,106,134,206,170],\"effect\":[218,111,158,133,196,21,24,92,131,17,221,40,251,90,102,151,153,204,156,92,154,223,40,214,213,222,12,92,245,160,150,252],\"expires_at\":1790003600,\"not_before\":1790000000,\"schema\":\"auths.exact-effect-approval/1\"},\"signature_b64\":\"qLYR1rtS2s4I4BRH3O2W6D6LAwOm8GDI6o1XBGsQtDBRnh4BZcKt4KakVAq5x2gcIigaqdkof557aY0Via0JCQ\"}","namespace":"airtable-demo","operation_id":"synthetic-entry","action_commitment":"0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b","recipe_digest":"64ff5004b9758e2fb374e9d6b443ef1b35aa00c400dbe1358b90ef89c5c9eda9","nonce":"00112233445566778899aabbccddeeff","evaluated_at":1790000000,"entry":{"revision":2,"authorization_started_at":1790000000,"approval_canonical":"{\"request\":{\"approver\":[232,218,99,164,12,166,135,200,124,252,224,92,178,74,120,108,126,117,204,73,199,13,181,87,63,2,111,28,106,134,206,170],\"effect\":[218,111,158,133,196,21,24,92,131,17,221,40,251,90,102,151,153,204,156,92,154,223,40,214,213,222,12,92,245,160,150,252],\"expires_at\":1790003600,\"not_before\":1790000000,\"schema\":\"auths.exact-effect-approval/1\"},\"signature_b64\":\"qLYR1rtS2s4I4BRH3O2W6D6LAwOm8GDI6o1XBGsQtDBRnh4BZcKt4KakVAq5x2gcIigaqdkof557aY0Via0JCQ\"}","pre_entry":null,"schema":"auths.effect-entry/3","granted_at":1790000000,"effect_id":"da6f9e85c415185c8311dd28fb5a669799cc9c5c9adf28d6d5de0c5cf5a096fc","dependencies":[]},"state":{"protocol":"mutation","facts":{"stage":"unknown","observation_plan":{"origin":"https://api.airtable.com","segments":[{"kind":"resolved","value":"v0"},{"kind":"resolved","value":"appTEST0000000001"},{"kind":"resolved","value":"tblTEST0000000001"},{"kind":"resolved","value":"recTEST0000000001"}],"headers":[],"json_pointer":"/fields/DemoStatus","expected":"\"Approved\"","maximum_response_bytes":16384,"echo_pointer":"/fields/auths_echo"}}}}
\ No newline at end of file
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/transition-entry-granted-to-response-recorded-synthetic-entry b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/transition-entry-granted-to-response-recorded-synthetic-entry
new file mode 100644
index 000000000..2b853f75f
Binary files /dev/null and b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/transition-entry-granted-to-response-recorded-synthetic-entry differ
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/transition-entry-granted-to-unknown-synthetic-entry b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/transition-entry-granted-to-unknown-synthetic-entry
new file mode 100644
index 000000000..3f640003f
Binary files /dev/null and b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/transition-entry-granted-to-unknown-synthetic-entry differ
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/transition-reserved-to-entry-granted-synthetic-entry b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/transition-reserved-to-entry-granted-synthetic-entry
new file mode 100644
index 000000000..bfbdac7ca
Binary files /dev/null and b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/transition-reserved-to-entry-granted-synthetic-entry differ
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/transition-reserved-to-not-entered-synthetic-entry b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/transition-reserved-to-not-entered-synthetic-entry
new file mode 100644
index 000000000..49a8c4f8b
Binary files /dev/null and b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_attempt/transition-reserved-to-not-entered-synthetic-entry differ
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-not-entered-after-claim b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-not-entered-after-claim
new file mode 100644
index 000000000..c5137204d
Binary files /dev/null and b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-not-entered-after-claim differ
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-not-entered-exhausted b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-not-entered-exhausted
new file mode 100644
index 000000000..c2e215013
Binary files /dev/null and b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-not-entered-exhausted differ
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-not-entered-pre-entry-condition-false b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-not-entered-pre-entry-condition-false
new file mode 100644
index 000000000..938d83a36
Binary files /dev/null and b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-not-entered-pre-entry-condition-false differ
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-not-entered-relative-ceiling-above b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-not-entered-relative-ceiling-above
new file mode 100644
index 000000000..6150f66c1
Binary files /dev/null and b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-not-entered-relative-ceiling-above differ
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-observed-by-provider-after-response b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-observed-by-provider-after-response
new file mode 100644
index 000000000..59adada01
Binary files /dev/null and b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-observed-by-provider-after-response differ
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-observed-by-provider-after-unknown b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-observed-by-provider-after-unknown
new file mode 100644
index 000000000..300d8f92f
Binary files /dev/null and b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-observed-by-provider-after-unknown differ
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-observed-by-provider-with-pre-entry b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-observed-by-provider-with-pre-entry
new file mode 100644
index 000000000..4a4f02b7e
Binary files /dev/null and b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-observed-by-provider-with-pre-entry differ
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-observed-echo-mismatch b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-observed-echo-mismatch
new file mode 100644
index 000000000..411257190
Binary files /dev/null and b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-observed-echo-mismatch differ
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-observed-match b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-observed-match
new file mode 100644
index 000000000..f7a8fb36c
Binary files /dev/null and b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-observed-match differ
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-observed-mismatch b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-observed-mismatch
new file mode 100644
index 000000000..1871aa798
Binary files /dev/null and b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-observed-mismatch differ
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-response-recorded b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-response-recorded
new file mode 100644
index 000000000..611f1dd55
Binary files /dev/null and b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-response-recorded differ
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-response-recorded-provider-rejected b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-response-recorded-provider-rejected
new file mode 100644
index 000000000..be6f2d698
Binary files /dev/null and b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-response-recorded-provider-rejected differ
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-unknown b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-unknown
new file mode 100644
index 000000000..e65b55924
Binary files /dev/null and b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-unknown differ
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-unknown-without-counters b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-unknown-without-counters
new file mode 100644
index 000000000..53daad577
Binary files /dev/null and b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/accepted-unknown-without-counters differ
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/outcome-v1-schema b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/obsolete-outcome-v1-schema
similarity index 100%
rename from product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/outcome-v1-schema
rename to product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/obsolete-outcome-v1-schema
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/observed-by-provider-after-response b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/obsolete-outcome-v2-observed-by-provider-after-response
similarity index 100%
rename from product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/observed-by-provider-after-response
rename to product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/obsolete-outcome-v2-observed-by-provider-after-response
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/refused-missing-record-commitment b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/refused-missing-record-commitment
new file mode 100644
index 000000000..71194835f
Binary files /dev/null and b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/refused-missing-record-commitment differ
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/refused-other-signer b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/refused-other-signer
new file mode 100644
index 000000000..6867b389b
Binary files /dev/null and b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/refused-other-signer differ
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/refused-outcome-v1-schema b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/refused-outcome-v1-schema
new file mode 100644
index 000000000..d4b10c246
Binary files /dev/null and b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/refused-outcome-v1-schema differ
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/refused-outcome-v2-schema b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/refused-outcome-v2-schema
new file mode 100644
index 000000000..da2afd471
Binary files /dev/null and b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/refused-outcome-v2-schema differ
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/refused-private-evidence-digest b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/refused-private-evidence-digest
new file mode 100644
index 000000000..f73880a54
Binary files /dev/null and b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/refused-private-evidence-digest differ
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/refused-private-outcome-schema b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/refused-private-outcome-schema
new file mode 100644
index 000000000..a0ef2aaa5
Binary files /dev/null and b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/refused-private-outcome-schema differ
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/refused-private-relative-basis b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/refused-private-relative-basis
new file mode 100644
index 000000000..938391133
Binary files /dev/null and b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/refused-private-relative-basis differ
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/refused-raw-action-commitment b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/refused-raw-action-commitment
new file mode 100644
index 000000000..9135f0004
Binary files /dev/null and b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_outcome/refused-raw-action-commitment differ
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_recipe/airtable b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_recipe/airtable
index aace62f96..3c23d8206 100644
Binary files a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_recipe/airtable and b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_recipe/airtable differ
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_recipe/content-only b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_recipe/content-only
new file mode 100644
index 000000000..1c805b1e1
Binary files /dev/null and b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_recipe/content-only differ
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_recipe/github b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_recipe/github
index 4cdccaee4..c44126ea6 100644
Binary files a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_recipe/github and b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_recipe/github differ
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_recipe/stripe-refund-approval b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_recipe/stripe-refund-approval
index d489d59c2..214fca7ac 100644
Binary files a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_recipe/stripe-refund-approval and b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_recipe/stripe-refund-approval differ
diff --git a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_recipe/todoist b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_recipe/todoist
index 3324c59b2..cb4735529 100644
Binary files a/product/runtime/auths-gateway/fuzz/corpus/target_gateway_recipe/todoist and b/product/runtime/auths-gateway/fuzz/corpus/target_gateway_recipe/todoist differ
diff --git a/product/runtime/auths-gateway/fuzz/fuzz_targets/target_gateway_app_frame.rs b/product/runtime/auths-gateway/fuzz/fuzz_targets/target_gateway_app_frame.rs
index 1a5208e15..9cef4ba13 100644
--- a/product/runtime/auths-gateway/fuzz/fuzz_targets/target_gateway_app_frame.rs
+++ b/product/runtime/auths-gateway/fuzz/fuzz_targets/target_gateway_app_frame.rs
@@ -4,7 +4,10 @@
#![no_main]
use auths_gateway::admin::{ADMIN_REQUEST_SCHEMA, parse_admin_request};
-use auths_gateway::app::{APP_OBSERVE_SCHEMA, APP_REQUEST_SCHEMA};
+use auths_gateway::app::{
+ APP_CONTENT_HISTORY_SCHEMA, APP_CONTENT_SCHEMA, APP_DISCLOSURE_ACK_SCHEMA,
+ APP_DISCLOSURE_SCHEMA, APP_OBSERVE_SCHEMA, APP_RECOVERY_SCHEMA, APP_REQUEST_SCHEMA,
+};
use auths_gateway::fuzzing;
use libfuzzer_sys::fuzz_target;
use serde_json::Value;
@@ -23,7 +26,13 @@ fuzz_target!(|data: &[u8]| {
if let Some(kind) = fuzzing::app_frame_kind(data) {
let expected = match kind {
"submit" => APP_REQUEST_SCHEMA,
- _ => APP_OBSERVE_SCHEMA,
+ "observe" => APP_OBSERVE_SCHEMA,
+ "recover" => APP_RECOVERY_SCHEMA,
+ "content" => APP_CONTENT_SCHEMA,
+ "disclose" => APP_DISCLOSURE_SCHEMA,
+ "content-history" => APP_CONTENT_HISTORY_SCHEMA,
+ "disclosure-ack" => APP_DISCLOSURE_ACK_SCHEMA,
+ _ => panic!("unknown closed application frame"),
};
assert_eq!(schema().as_deref(), Some(expected));
}
diff --git a/product/runtime/auths-gateway/fuzz/fuzz_targets/target_gateway_attempt.rs b/product/runtime/auths-gateway/fuzz/fuzz_targets/target_gateway_attempt.rs
index 9c186ae29..3ecf13f18 100644
--- a/product/runtime/auths-gateway/fuzz/fuzz_targets/target_gateway_attempt.rs
+++ b/product/runtime/auths-gateway/fuzz/fuzz_targets/target_gateway_attempt.rs
@@ -23,29 +23,31 @@ const fn listed(from: Stage, to: Stage) -> bool {
matches!(
(from, to),
(
- Stage::Attempting,
- Stage::Attempting
- | Stage::NotEntered
- | Stage::Unknown
- | Stage::ResponseRecorded
- | Stage::ObservedByProvider
- ) | (Stage::Unknown, Stage::ObservedByProvider)
- | (
- Stage::ResponseRecorded,
- Stage::Observed | Stage::ObservedByProvider
- )
+ Stage::Reserved,
+ Stage::Reserved | Stage::EntryGranted | Stage::NotEntered
+ ) | (
+ Stage::EntryGranted,
+ Stage::Unknown | Stage::ResponseRecorded | Stage::ObservedByProvider
+ ) | (
+ Stage::Unknown,
+ Stage::ResponseRecorded | Stage::ObservedByProvider
+ ) | (
+ Stage::ResponseRecorded,
+ Stage::Observed | Stage::ObservedByProvider
+ )
)
}
-const FIXED: [&str; 8] = [
+const FIXED: [&str; 9] = [
"schema",
+ "binding_canonical",
+ "approval_canonical",
"namespace",
"operation_id",
"action_commitment",
"recipe_digest",
"nonce",
"evaluated_at",
- "counters",
];
fuzz_target!(|data: &[u8]| {
@@ -68,6 +70,17 @@ fuzz_target!(|data: &[u8]| {
) else {
panic!("valid records are JSON");
};
+ let revision = old["revision"].as_u64().expect("old revision");
+ assert!(revision < 9_007_199_254_740_991);
+ assert_eq!(new["revision"].as_u64(), Some(revision + 1));
+ if from != Stage::Reserved || to != Stage::EntryGranted {
+ assert_eq!(old.get("entry"), new.get("entry"), "entry grant changed");
+ }
+ assert_eq!(old["state"]["protocol"], new["state"]["protocol"]);
+ assert_eq!(
+ old["state"]["facts"]["counters"],
+ new["state"]["facts"]["counters"]
+ );
for field in FIXED {
assert_eq!(old.get(field), new.get(field), "{field} changed");
}
diff --git a/product/runtime/auths-gateway/fuzz/fuzz_targets/target_gateway_recipe.rs b/product/runtime/auths-gateway/fuzz/fuzz_targets/target_gateway_recipe.rs
index 147d8ae15..22af7b374 100644
--- a/product/runtime/auths-gateway/fuzz/fuzz_targets/target_gateway_recipe.rs
+++ b/product/runtime/auths-gateway/fuzz/fuzz_targets/target_gateway_recipe.rs
@@ -28,7 +28,11 @@ fuzz_target!(|data: &[u8]| {
let review = recipe.review();
let _ = recipe.review_document();
assert_eq!(review.recovery(), recipe.recovery());
- assert!(!recipe.recovery().write_is_conditional);
+ assert!(
+ recipe
+ .recovery()
+ .is_none_or(|capability| !capability.write_is_conditional)
+ );
let Ok(value) = serde_json::from_slice::(source) else {
panic!("an accepted source is JSON");
};
diff --git a/product/runtime/auths-gateway/fuzz/tests/corpus.rs b/product/runtime/auths-gateway/fuzz/tests/corpus.rs
index ca4e00b5d..8e28cd5e6 100644
--- a/product/runtime/auths-gateway/fuzz/tests/corpus.rs
+++ b/product/runtime/auths-gateway/fuzz/tests/corpus.rs
@@ -97,11 +97,26 @@ fn request_seeds_build_a_closed_request() {
#[test]
fn attempt_seeds_are_records_and_valid_transitions() {
for (name, data) in seeds("target_gateway_attempt") {
+ if name.starts_with("obsolete-transition-") {
+ let (old, new) = split(&data);
+ assert!(fuzzing::attempt_stage(old).is_none(), "obsolete {name}");
+ assert!(fuzzing::attempt_stage(new).is_none(), "obsolete {name}");
+ continue;
+ }
+ if name.starts_with("obsolete-record-") {
+ assert!(fuzzing::attempt_stage(&data).is_none(), "obsolete {name}");
+ continue;
+ }
+
if name.starts_with("transition-") {
let (old, new) = split(&data);
assert_eq!(
fuzzing::attempt_stage(old),
- Some(GatewayAttemptStage::Attempting),
+ Some(if name.starts_with("transition-reserved-") {
+ GatewayAttemptStage::Reserved
+ } else {
+ GatewayAttemptStage::EntryGranted
+ }),
"{name}"
);
assert_eq!(fuzzing::attempt_transition(old, new), Some(true), "{name}");
@@ -114,10 +129,56 @@ fn attempt_seeds_are_records_and_valid_transitions() {
#[test]
fn outcome_seeds_verify_or_are_refused_as_named() {
let observer = fuzzing::test_observer(0x5a);
+ let vectors: Value = serde_json::from_slice(
+ &fs::read(
+ PathBuf::from(env!("CARGO_MANIFEST_DIR"))
+ .join("../../../../bindings/fixtures/gateway/outcome-v3.json"),
+ )
+ .expect("canonical public outcome vectors"),
+ )
+ .expect("vector JSON");
+ let mut accepted = 0;
+ let mut refused = 0;
for (name, data) in seeds("target_gateway_outcome") {
let verified = fuzzing::verify_outcome(&data, &observer);
- assert_eq!(verified.is_some(), name != "outcome-v1-schema", "{name}");
+ if name.starts_with("obsolete-") {
+ assert!(verified.is_none(), "obsolete {name}");
+ continue;
+ }
+ let (prefix, id) = name.split_once('-').expect("closed seed name");
+ let expected = match prefix {
+ "accepted" => {
+ accepted += 1;
+ true
+ }
+ "refused" => {
+ refused += 1;
+ false
+ }
+ _ => panic!("unknown outcome seed class: {name}"),
+ };
+ assert!(
+ vectors[prefix]
+ .as_array()
+ .expect("cases")
+ .iter()
+ .any(|v| v["id"] == id),
+ "unknown canonical case: {name}"
+ );
+ assert_eq!(verified.is_some(), expected, "{name}");
}
+ assert!(
+ accepted > 0 && refused > 0,
+ "both decoder paths require seeds"
+ );
+ assert_eq!(
+ accepted,
+ vectors["accepted"].as_array().expect("accepted").len()
+ );
+ assert_eq!(
+ refused,
+ vectors["refused"].as_array().expect("refused").len()
+ );
}
#[cfg(unix)]
@@ -126,6 +187,6 @@ fn frame_seeds_parse_as_their_closed_schemas() {
for (name, data) in seeds("target_gateway_app_frame") {
let parsed = fuzzing::app_frame_kind(&data).is_some()
|| auths_gateway::admin::parse_admin_request(&data).is_ok();
- assert!(parsed, "{name}");
+ assert_eq!(parsed, !name.starts_with("obsolete-"), "{name}");
}
}
diff --git a/product/runtime/auths-gateway/semantic-closure.json b/product/runtime/auths-gateway/semantic-closure.json
index b5d7f2261..b880a158f 100644
--- a/product/runtime/auths-gateway/semantic-closure.json
+++ b/product/runtime/auths-gateway/semantic-closure.json
@@ -1 +1 @@
-{"files":[{"path":"Cargo.lock","sha256":"0e75e2fdef44e79e57797ac6661295ae58ddb8de4a1990a6fd22ce6c54f78604"},{"path":"Cargo.toml","sha256":"6e0240aa432bcdb5b4f855618994582d2e428afe5ee4e4f6a3f8d21a518f8d24"},{"path":"core/adapters/auths-did-keri/Cargo.toml","sha256":"86ba3e3f9a0a70d970568bfb3113b7ba7bbdd81c3eebbb52743dd574a0c0ecaa"},{"path":"core/adapters/auths-did-keri/src/lib.rs","sha256":"a75674842f8a7142004369ccd206ab43e476459293f0ff4bed4d4e659031b3e3"},{"path":"core/adapters/auths-did-key/Cargo.toml","sha256":"d5a3ae07ecde83f91eed4c3785051dd0f44eab6c891754f3309b6b40596707d7"},{"path":"core/adapters/auths-did-key/src/lib.rs","sha256":"b76a3f4528a725df53cfa9242e6e414e1d26caa48468aaed497e28b121f4497c"},{"path":"core/adapters/auths-raw-key/Cargo.toml","sha256":"f49616bc18c19e6aaa7e6a02ccebcdb9f5c1422c1278ef9869e0dbf22ceda7be"},{"path":"core/adapters/auths-raw-key/src/lib.rs","sha256":"45eec1b295dac3bed65cd79c942870b926b06b3f5d38eaebaeb6140ac69315fa"},{"path":"core/crates/auths-algebra-kernel/Cargo.toml","sha256":"0cb078915287b8c47d20fb97b6fa14135ed1c61c4985c4de2753855563c27bba"},{"path":"core/crates/auths-algebra-kernel/src/generated.rs","sha256":"ae4bcbef42a408c7ffa0c6a74a42f7de177a6f14479e51d205acf830bda05960"},{"path":"core/crates/auths-algebra-kernel/src/lib.rs","sha256":"6654667eeb9b353d6648954c9fd2e8838462a82ce6b2803eb49e1366892c68c6"},{"path":"core/crates/auths-assurance/Cargo.toml","sha256":"afcb12a97ddd09d7a264b9ecd3fcaefc865ff2c9b4539ee85b00883420ba333a"},{"path":"core/crates/auths-assurance/src/lib.rs","sha256":"0fb2ec844d674b25a9e20b1601452c89c2454901fa506adfa9bb38ef9a7fa213"},{"path":"core/crates/auths-author/Cargo.toml","sha256":"1e6f0e3bc079ddb558decf0db26b291c6e72fc47327b6df1b7650d41689b215f"},{"path":"core/crates/auths-author/src/lib.rs","sha256":"e14475f2f49715a1869d5fa6b11ff1fbb72497e39097e09f833b0bae7e5b3f2b"},{"path":"core/crates/auths-author/src/observations.rs","sha256":"dda7d6bfa6f4bd7e95a480ca283c2c0e996f69c5c15cb0c7197af328131a7066"},{"path":"core/crates/auths-authority/Cargo.toml","sha256":"4f68212938029f79ebd857b4f3038494eb85d0163d49f07e51376715cf959815"},{"path":"core/crates/auths-authority/src/lib.rs","sha256":"da88d5194de4e17449e69972c0fb042afa3c39798a8f570f30771e77c384f1eb"},{"path":"core/crates/auths-codec/Cargo.toml","sha256":"e20e51826e17814effd9d8925d18d5fbdd6c0d37bf50231e743ec2611314722c"},{"path":"core/crates/auths-codec/src/approval.rs","sha256":"fbff43bf4d57f69f8b7738a8647751cde75c5a932dee65d924a1fe9c7202f91f"},{"path":"core/crates/auths-codec/src/bounded_policy.rs","sha256":"d21f6728be2d8a04b6d94921f2b683f7e61cbb42e0ac6866f1c88a49caa534c3"},{"path":"core/crates/auths-codec/src/decode.rs","sha256":"ba2999584cdf53b1b517ecb9c7ea50d39efee17c666a9a38e2ad1ac94379f778"},{"path":"core/crates/auths-codec/src/encode.rs","sha256":"8491192053e2e15d62b4d2a491de22405b7487ead69ab6f6bca40669ce7ca5f7"},{"path":"core/crates/auths-codec/src/error.rs","sha256":"fecb0737c16deb42d859f1dc7727c698438d9ac12064fd99e0024629642c390c"},{"path":"core/crates/auths-codec/src/hash.rs","sha256":"f88e111f1dbb2cef6b7b5e9c948cb45ff92e366710c8a5f5c636138aa748479e"},{"path":"core/crates/auths-codec/src/lib.rs","sha256":"3b0e5c94c6f34f750b88a9333e1d5c8145746d845d3a23e0d2f61807cf592ea7"},{"path":"core/crates/auths-codec/src/observation.rs","sha256":"38328b0c71ff97a4196eeade8b4faa889527ff1f80bb168468bcd49082a3987e"},{"path":"core/crates/auths-composition/Cargo.toml","sha256":"331e13025eb82588b5269ed3975929eb41c823d588c49e4c296273d85c6fe80e"},{"path":"core/crates/auths-composition/src/lib.rs","sha256":"e42e3b0b6a9985b73c533aeb23799c638979aa27a9ce7d0f10a8d60340570fb1"},{"path":"core/crates/auths-model/Cargo.toml","sha256":"565c15302b0cc9270db1e28efa3862804846850ba933cc17aea8abcec4a65987"},{"path":"core/crates/auths-model/src/approval.rs","sha256":"54cf44a9ef747ec21546a4086aecfb130bc7803b0d646430e6082df51e6e305f"},{"path":"core/crates/auths-model/src/bounded.rs","sha256":"f90496abc0a06c028b4d850af6ddab7d850bfafeaa3d042f1b1246a41d8bbbc9"},{"path":"core/crates/auths-model/src/bounded_policy.rs","sha256":"306a8f049cc7cfc7635fe0e4de47eb7813eed63d3c14c3be2b51586aa4eea450"},{"path":"core/crates/auths-model/src/lib.rs","sha256":"e7cfe6bb9574c6b016aa63e70658c31e0b90903ec6479a49f4a2ceb64f5bad97"},{"path":"core/crates/auths-model/src/observation.rs","sha256":"b3ed3e3cc090afca2896a7d3de7988afccc1ee1ea0ebd91b51a8bc0f20ad8dbb"},{"path":"core/crates/auths-multikey/Cargo.toml","sha256":"858756c1197befcada06134742bffe8d58cf674cadd6c046bf710570e96efaf2"},{"path":"core/crates/auths-multikey/src/lib.rs","sha256":"f5177fffb11e40eeaaa665ba886c541fc1f24d5eacdd6983fb03dc20148733f0"},{"path":"core/crates/auths-ports/Cargo.toml","sha256":"596a9b0f5af3717c16a6d13f7054d8a106bcd9f01f5f72c2235dd2afcf0ab16c"},{"path":"core/crates/auths-ports/src/binding.rs","sha256":"a16c012d40aa4249e3ec54372b9611006d0bfe8c7dc3c741911c2194ad17736d"},{"path":"core/crates/auths-ports/src/diagnostics.rs","sha256":"aa596572503b5a65a918bb2ff38281c144ff53821ee914e9325c852828d852df"},{"path":"core/crates/auths-ports/src/lib.rs","sha256":"8da1debf86f9532df0660155d04298beb1ba3a1204508f8586e2e67cd2fb3449"},{"path":"core/crates/auths-ports/src/path.rs","sha256":"b68fc547b4e4d97a60cbbc1cc1939fc6b34a8233b81f840bd396914a707e3ea1"},{"path":"core/crates/auths-raw-key-core/Cargo.toml","sha256":"98c0bc8527a537512e43a74dd6ba9658c6f0cdc16610cf938e3f9b43785d836d"},{"path":"core/crates/auths-raw-key-core/src/lib.rs","sha256":"98c0500c41697aae28708ad0bf970b9a97583d96a3d2c18d23e7fa351e0f7d12"},{"path":"core/crates/auths-registries/Cargo.toml","sha256":"590e49e58b0622d9d6274fea46796dcdb07dec9550eff295cce6a6427ffa51e2"},{"path":"core/crates/auths-registries/src/lib.rs","sha256":"450d3886d85ee333b7e046317fb5b47cde62483d5f3e41d46265e2caf4c373ec"},{"path":"core/crates/auths-registries/src/template.rs","sha256":"1224039dd2b3ce6fa8357613d8cafbae40400493da7176da202b9ade62fa05e4"},{"path":"core/crates/auths-signature-core/Cargo.toml","sha256":"6bee2230eb30af1be0bd5c2d57bd3e1903497a70028d47290d4ac6e9e0e8bca0"},{"path":"core/crates/auths-signature-core/src/lib.rs","sha256":"6e3f9311ded5e49ea50d445eb21e6d8fa12442aa4362a73e0cb6fda15b7cd865"},{"path":"core/crates/auths-signature/Cargo.toml","sha256":"7f7b5682466ba439d8c59579022cb61068cad9fb4bd2d73e1cc21327c902b450"},{"path":"core/crates/auths-signature/src/lib.rs","sha256":"b3d9ac4cc3ba0797b9e1c09f2d6ac05089b25dac4e285a5af28c0165b33d925b"},{"path":"core/crates/auths-verifier/Cargo.toml","sha256":"c43358e06a33324b41a8b93e5cc7279c55f7d21f1f6cfaa539a3fc7b4c27a38f"},{"path":"core/crates/auths-verifier/src/approval.rs","sha256":"c3cae6a49e169051eb555bb092b1069a4d422516d6a2114623a89e5f31ff923f"},{"path":"core/crates/auths-verifier/src/causal.rs","sha256":"9635230f33b13a263573d319ac8a1976addbcc07677514a43e3c7bf28dc04587"},{"path":"core/crates/auths-verifier/src/lib.rs","sha256":"f22dec38dfa47ced8da4f605390da8453273bee1da010f57fd1fdfcb10e0b389"},{"path":"core/crates/auths-verifier/src/observation.rs","sha256":"0bb07f06a0f9524c88427df1c972b1de8ba038a9eeeb15d3134c25125ae84b1e"},{"path":"core/crates/auths-verifier/src/trace.rs","sha256":"2213d72023a1d9e0413a2cf2c51c3dad9e04775d86033a0caaa9d29a8f0036f8"},{"path":"exchange/crates/auths-proof-exchange-model/Cargo.toml","sha256":"34b2db64f93a288e942b5a3918d28cdcbef906e0dc9074ce848018d3d5f7a1fa"},{"path":"exchange/crates/auths-proof-exchange-model/src/lib.rs","sha256":"49fac5aff2339d3a57c41cb8bb2a82852538e8f77ecc4b157c88e264ff31fb55"},{"path":"exchange/crates/auths-proof-exchange-port/Cargo.toml","sha256":"7ff660f1b7579dc570c791090aeb450404fbbf44c92781e289f5ccdde746d739"},{"path":"exchange/crates/auths-proof-exchange-port/src/lib.rs","sha256":"b408fb40c6432edc1facb3d8b3e9433a033f8dc73fe2e0ba5a88c972ca675021"},{"path":"product/config/auths-config/Cargo.toml","sha256":"e4b568ab9f6d84f7d7e6d66443fe825f360db9c27bc7c31d6ddc0647ba3ab304"},{"path":"product/config/auths-config/src/lib.rs","sha256":"fecf00ea65cef9018fda9af2063326d7065b869266d0d9e73a4f06816f1102cc"},{"path":"product/errors/auths-errors/Cargo.toml","sha256":"3b3a199b6460ba75bda84475bb5accd35e1485448a3981038db85bf4772dd350"},{"path":"product/errors/auths-errors/src/lib.rs","sha256":"7feeb9069fa4003f94f1f07c1c94619cf1c883e0dbaaf2187f04d0f742264bdd"},{"path":"product/integrations/auths-credentials-aws-secrets-manager/Cargo.toml","sha256":"716f1f2a117d6d69d8effe9ee9057e24ba3502f20af6a602b06facb638cdabed"},{"path":"product/integrations/auths-credentials-aws-secrets-manager/src/api.rs","sha256":"9ea61dd7ddd93282b073b98e453e6b75b26dc49689c04e9782435d06a3a09e48"},{"path":"product/integrations/auths-credentials-aws-secrets-manager/src/clock.rs","sha256":"73f0a6e0f5903883329fa6683f5888288f759512ac69e5c87efd2b6b87f7d79f"},{"path":"product/integrations/auths-credentials-aws-secrets-manager/src/http.rs","sha256":"14c1fce14e24c984b2a275563a29b6326fb08367922b6e8d37ba38b9cb81aa2b"},{"path":"product/integrations/auths-credentials-aws-secrets-manager/src/identity.rs","sha256":"186de11230c8b33375951c4368609f8976bf9795f171af3fd7fac26207a94361"},{"path":"product/integrations/auths-credentials-aws-secrets-manager/src/lib.rs","sha256":"0b4617d044f5254a5ab22f5a2578f91cc025206a8c1845be39c1c56dcdf7a5df"},{"path":"product/integrations/auths-credentials-aws-secrets-manager/src/names.rs","sha256":"41bb23572d6002e23ee103d6fe70ecd68764861c6f2e48a1c53af998270a4afa"},{"path":"product/integrations/auths-credentials-aws-secrets-manager/src/sigv4.rs","sha256":"39fed883f4802e68d3b41a68fea0aba0737810996b33d13577de70faba004c29"},{"path":"product/integrations/auths-credentials-aws-secrets-manager/src/store.rs","sha256":"fc60ee9c49b2f4f063c449b4c78100c5c4872779e94ca168f796603b380ef6e4"},{"path":"product/integrations/auths-custody/Cargo.toml","sha256":"1e14034185ad1ec24a977fd5f41d0abee219606245c046ce2a82287fadef448b"},{"path":"product/integrations/auths-custody/src/conformance.rs","sha256":"8ed1afd58f18fc6becc268908aa6419c9aa404c79da0197963eb2599df44acf6"},{"path":"product/integrations/auths-custody/src/key.rs","sha256":"70cbb0fd21b6f5652157845a22c35be2d0f6c05b3d138fc4b9bb6dbdc3d81c40"},{"path":"product/integrations/auths-custody/src/lib.rs","sha256":"8ee593c1f59b0fd2126bdda9138026f5bc531beb64d4489a99785842e7b98b43"},{"path":"product/operations/auths-operations/Cargo.toml","sha256":"73dde1e914ba5e1386f8d49c0d730423cc72ae77ec44c6554a6a75cc8e8922b1"},{"path":"product/operations/auths-operations/src/explanation.rs","sha256":"b4a39bcbd5712cb9e735ee7deeb281f3fa6af5ddca973735bf3d87dce427ef11"},{"path":"product/operations/auths-operations/src/lib.rs","sha256":"3f401a56555113b98f57bca172a2660eaecedf76247b83a3fdbb089c133a29cd"},{"path":"product/operations/auths-operations/src/render/json.rs","sha256":"2ae042639960afee123730e945d373bc5dda1d2b3af20ae16c939a7e99b9e418"},{"path":"product/operations/auths-operations/src/render/mod.rs","sha256":"b3b7f6fe473f49227ace5a68bfc3d8dc14b5db8dd9888fd9e207070337b1b8bf"},{"path":"product/operations/auths-operations/src/render/text.rs","sha256":"f216cf3bdb00e264523271b16e7059535ac6b1aabf271cebd7935af155e4bfec"},{"path":"product/policy/auths-bounded-policy/Cargo.toml","sha256":"babd5d200923137ead71d8c2528f847cea323ce98da942a1320b054436612c47"},{"path":"product/policy/auths-bounded-policy/src/arithmetic.rs","sha256":"75e43c5d4eab8d1f71c1daf53839f5f9fe45a400afc1f63a30c6f55ce3632627"},{"path":"product/policy/auths-bounded-policy/src/commitment.rs","sha256":"c5706f2d583454e1a7f570f7ada138d4d864bef0205dfae7c1296774f396bb3f"},{"path":"product/policy/auths-bounded-policy/src/eligibility.rs","sha256":"c9acdfb793074e687fc36efad506be27dbd99ffbf40387bc66c8de39ef238432"},{"path":"product/policy/auths-bounded-policy/src/identifier.rs","sha256":"0006a32bae9eb16c4178cf797df418d2b0a781f2bc6474b741dc1a908141831c"},{"path":"product/policy/auths-bounded-policy/src/kernel.rs","sha256":"99187c28b161ac70bee1afed60711f7f645adb4d1671a01863f9849b369811d5"},{"path":"product/policy/auths-bounded-policy/src/lib.rs","sha256":"76a18638b810e62e594d6e951504679305ce21841388e0c282ad51cd7841b91c"},{"path":"product/policy/auths-bounded-policy/src/receipt.rs","sha256":"3b9eae82f349956d3e7f29b57b254bff5ce67cc703e6cf43445ce2827676d1ad"},{"path":"product/policy/auths-bounded-policy/src/registry.rs","sha256":"4e251ea5833559841f8ee0bd9b66c0c01c39fa5e3a21720a19c4e8adbfcb99d0"},{"path":"product/profiles/auths-profile-api/Cargo.toml","sha256":"7ae85761f58625b5d678f961fd052eebf3e000ef565c7312e3ac857763c0017f"},{"path":"product/profiles/auths-profile-api/src/lib.rs","sha256":"a194b82b20bf8aa864c2334ef6c6aec30b2154ded972862252d5dcefa488ab95"},{"path":"product/profiles/auths-profile-mcp/Cargo.toml","sha256":"db264fc25242f7b7e4133cbe5782545b3107b406c87168ae072683cfb59eac66"},{"path":"product/profiles/auths-profile-mcp/src/arguments.rs","sha256":"a5de696586d507e2aee4a49fee99ef1ef4c51a2e46603f8838f1af6d9009833f"},{"path":"product/profiles/auths-profile-mcp/src/lib.rs","sha256":"d0f8b733a09ef8408336cdebc8948d6a5861012ce6878598b11be7084dfd4280"},{"path":"product/profiles/auths-profile-mcp/src/session.rs","sha256":"166c281fa66333388e4760d6098dad02d3359e1d5de4b8f93312c014f97abb03"},{"path":"product/qualification/auths-recipe-qualification/Cargo.toml","sha256":"b7b15b30b9a3d96d4f089d514266d3a892de6608d30707aa44be31cca89ca9fe"},{"path":"product/qualification/auths-recipe-qualification/src/canonical.rs","sha256":"30a9b50777f0dcb4b606abfee2467d5a975d9f6ad941ab6c13ca3df5b359f667"},{"path":"product/qualification/auths-recipe-qualification/src/closure.rs","sha256":"0ad6aee9bd981e2863fa725f209ec84031db2ff12f4c7ac3aaf5120f0bb86039"},{"path":"product/qualification/auths-recipe-qualification/src/commissioning.rs","sha256":"db135c55e4e3079a73253ebae275536523eb36b545d1c152b4d7dce3c6b76828"},{"path":"product/qualification/auths-recipe-qualification/src/commissioning/tests.rs","sha256":"4dde3736d206099b1bea4c14e0092c6c1287d1af9065958777d7c54331dd8686"},{"path":"product/qualification/auths-recipe-qualification/src/error.rs","sha256":"52d5cf720a1e4579008fdda90a3ac2c31f0b8db57985f121f65e54b4da2b7dd1"},{"path":"product/qualification/auths-recipe-qualification/src/evidence.rs","sha256":"ba45a7409d67a8af26ab811af306412014557477a101ab2a8eb0faf8f2bd4b91"},{"path":"product/qualification/auths-recipe-qualification/src/ids.rs","sha256":"e81a906d571497ab2da753e1532d4443706410e0b75fdd59a7d2d788c7aec403"},{"path":"product/qualification/auths-recipe-qualification/src/launch.rs","sha256":"805154ce931e3977f2a2a2c62909d90f7c0a22dc167bf4f1554dcf19316ffef8"},{"path":"product/qualification/auths-recipe-qualification/src/lib.rs","sha256":"0e24b0f81191df0d078c7f4d73ced784ba9472372bfcdff3c7d0e2fb1687b323"},{"path":"product/qualification/auths-recipe-qualification/src/model.rs","sha256":"3484eb0ec287b23d4682ade7b1b0f8d1335985b3fb180f06fc28716076957d5f"},{"path":"product/qualification/auths-recipe-qualification/src/release.rs","sha256":"ec478e53f70052e566aac7e15be2f9e1b1cd46837ac6990ae955fa864d311586"},{"path":"product/qualification/auths-recipe-qualification/src/vectors/mod.rs","sha256":"fa23a962fd90331152eed5edbcdf271955f7aed79b8da1e065bfe69a20b487a9"},{"path":"product/qualification/auths-recipe-qualification/src/vectors/schemas.rs","sha256":"c59feec2805621f341a90f81fc005e2c12f7a70f9d264fdeacccb2be4fe1c046"},{"path":"product/qualification/auths-recipe-qualification/src/vectors/verification.rs","sha256":"bdf5746fed4d10a7f577f7878619a917d29aa8b7d522f00d9f774e53eff4c8c1"},{"path":"product/qualification/auths-recipe-qualification/src/verify.rs","sha256":"646ff0e52e915174ad183768706e35e5ce2118b83475455310b930a77a7877f6"},{"path":"product/receipts/auths-receipts/Cargo.toml","sha256":"6001fa1375899953aafcb07c23a6da0c8f57ea7ec08bb9c950b12a60a8cc6f0f"},{"path":"product/receipts/auths-receipts/src/disclosure.rs","sha256":"ba9ed439be0a320c8a01f5b6e9b51361c1a37e076947490d8e4a484ba6bb119c"},{"path":"product/receipts/auths-receipts/src/lib.rs","sha256":"547a57c4160256719cafb7d0ab972d9fc80b0da14fbbce4148c70c414070ff36"},{"path":"product/runtime/auths-connections/Cargo.toml","sha256":"babeb6e1dda3dfb66291713ce1e43dce5c349833d0b50ea4dccbf986030026b2"},{"path":"product/runtime/auths-connections/src/credential.rs","sha256":"82921413bb053f0f6c7931fbd4e6a86c8cc1d8c6b51debd9c2cef33a8d5b1c51"},{"path":"product/runtime/auths-connections/src/kernel.rs","sha256":"a456af19526ef7b19278551fd8d97606b6e7aa5a195e70bdd85d1e0786090ce8"},{"path":"product/runtime/auths-connections/src/lib.rs","sha256":"79209249a0acba0e854d9937255ea0e30a9a6f8e1c2baad77ab83646d392de22"},{"path":"product/runtime/auths-connections/src/model.rs","sha256":"de34a089cac9702acb889b7602b3e4732306c9e61029ec892582b7e02b73fc77"},{"path":"product/runtime/auths-connections/src/registry.rs","sha256":"e57640e003e72ad3235c567c2bb944a262bd64e673f023b62ea091094cc7e946"},{"path":"product/runtime/auths-gateway-kernel/Cargo.toml","sha256":"34f864e67656af4ac7e5c7a347a23f01fc8ab31365b8e23c0ef474483b14a183"},{"path":"product/runtime/auths-gateway-kernel/src/construct.rs","sha256":"e9d65744bae273d65face4285e5d2da75c06a92ab48f7b8333d496604e31c63c"},{"path":"product/runtime/auths-gateway-kernel/src/lib.rs","sha256":"cfda6f8453426564c085475f183b6f790b438768a19668f1b65d33ba59dc8632"},{"path":"product/runtime/auths-gateway-kernel/src/order.rs","sha256":"e0045f877c4b7fe3860c291039853885064ddcdebba7c9b345ab65e81e646dda"},{"path":"product/runtime/auths-gateway-kernel/src/outcome.rs","sha256":"e9785b1afe70788c89b51b18655bcd5acff7b0c8e3d12fb79c6f05a202e67e6b"},{"path":"product/runtime/auths-gateway-kernel/src/ratio.rs","sha256":"7ace22264812e6fc1ef3a62afee402cf43494bb71ae0d3de219ad1b39d690bcb"},{"path":"product/runtime/auths-gateway-kernel/src/recovery.rs","sha256":"9f46dcd1685ab6e4d1ba04a3fc286a9f3afee8fb47829be1fbc670a7926a2dc6"},{"path":"product/runtime/auths-gateway-kernel/src/transition.rs","sha256":"cefb8f01d915880f6921f2501ff881ad94b9de61072c34d9cd19220992f97374"},{"path":"product/runtime/auths-gateway/Cargo.toml","sha256":"00b28be86921dcdc0ad90cce0c62c2961cafefe96ea6c2edf0e926c3d405ccad"},{"path":"product/runtime/auths-gateway/src/admin.rs","sha256":"0dee40a9fb67c5de10142fc37a7bbabdd8979fb0199fb26caf4bdb024e0a5a40"},{"path":"product/runtime/auths-gateway/src/app.rs","sha256":"6f80f63833ea0815bb44fbf3b0cc54624a08e9fb578b035819f2980dc04e8256"},{"path":"product/runtime/auths-gateway/src/audit.rs","sha256":"0edb2037eb7b5d6f13acfce85c54f62a82bf7fdd31bc7e626c4fec183cef61dd"},{"path":"product/runtime/auths-gateway/src/audit_tests.rs","sha256":"990969384a52d75bc2489d9d412a9e7dd814bca1c8f65c701d6205fae62a15d5"},{"path":"product/runtime/auths-gateway/src/bin/auths-gateway-harness.rs","sha256":"12b0c24d36b2e6a3be392d3c081a783b355fdd67dc5b54bacc0410e803116b8c"},{"path":"product/runtime/auths-gateway/src/bin/auths-gateway.rs","sha256":"3dba485deeb4d60f3f5603ad4ce9a0edf49a85c68c5c4ca072df73d2910da832"},{"path":"product/runtime/auths-gateway/src/binding.rs","sha256":"663b0ce878319511aa7d05722ab4ab2fa9be18817884ffb2ee9f131ff1e382d1"},{"path":"product/runtime/auths-gateway/src/bounds.rs","sha256":"38251ad6f22f9d07ef9a219c79fa265ca36f1526eca5f147d7efa46b52ad609d"},{"path":"product/runtime/auths-gateway/src/bounds_aggregate_tests.rs","sha256":"ab34f227b7c8692ba1da88ccb32bb3687a66adf8353d8282fae905d4a88eb612"},{"path":"product/runtime/auths-gateway/src/bounds_tests.rs","sha256":"74136bc570da7b9338b5a1a1ed9d3fd892c55135160964f9d3ac0e2cf5741375"},{"path":"product/runtime/auths-gateway/src/commissioning_budget.rs","sha256":"fab29da3269438cc62330f3eab7b5bd4dc482b4db8af243aae47df5d3aa1deb8"},{"path":"product/runtime/auths-gateway/src/commissioning_budget/tests.rs","sha256":"da104a1643c37515deec6397bca6db2054f2c9158ec3d230a1b1ec9c7d0f659f"},{"path":"product/runtime/auths-gateway/src/commissioning_floor.rs","sha256":"7fa6991c5ccd5ae1403be84b3da390323fab09f36e15ac01fd7a38ea350c6fdb"},{"path":"product/runtime/auths-gateway/src/commissioning_floor/tests.rs","sha256":"6b663d6e556adc0b455b81845bc4394d82c6c4de0c5f43e1310f3ddc6e4e40f7"},{"path":"product/runtime/auths-gateway/src/commissioning_session.rs","sha256":"24b48d5eaeaaa3c620e2dfc7ee1f9255fc4519a1eb404b8608cea7af6a910ee2"},{"path":"product/runtime/auths-gateway/src/commissioning_session/tests.rs","sha256":"37a097fc8a9261d3fffdaef3f905f40b22e3e6601bb5101fadd9414f963c0824"},{"path":"product/runtime/auths-gateway/src/connection.rs","sha256":"e8d24277d7e4a53610e091fe78b373b841245d8e59a2bc788716b76e7d912441"},{"path":"product/runtime/auths-gateway/src/credential_journal.rs","sha256":"7297fccaee5cd8622928e8e0a42bba50da7941970a6f682ecd300c41ee6509da"},{"path":"product/runtime/auths-gateway/src/echo_verify.rs","sha256":"6d453051665f823abd7354d5c50031d9524cf47d6092d2f09080ae814e3b9575"},{"path":"product/runtime/auths-gateway/src/engine.rs","sha256":"692ff372ffab10f27fcc45b7aa28457ca3fd6f9ca0ca06212f20f34c4431b491"},{"path":"product/runtime/auths-gateway/src/execution_witness.rs","sha256":"e8b1a8e41d5e3fa309760067849d390ef065feeb46701a83391832645a2f000c"},{"path":"product/runtime/auths-gateway/src/fuzzing.rs","sha256":"ed3b0028487eb546a3b0a91d115aa8f18bcea324b61620bfe8e35a9e6ab18758"},{"path":"product/runtime/auths-gateway/src/generation_floor.rs","sha256":"fe1510ec96ce0f67e9a78edbfa94b33eb7df64c0ac4d0a67901124dcd0164777"},{"path":"product/runtime/auths-gateway/src/harness.rs","sha256":"c87e479f3eb281960f6c9c80d43c7560d7ccb654b0154ec823d3fcec4884ac05"},{"path":"product/runtime/auths-gateway/src/lib.rs","sha256":"fad95d031f88072dd024c78facf0bb0267562b1ac743ddd7219fce47e64024f0"},{"path":"product/runtime/auths-gateway/src/listener.rs","sha256":"aa08239bcbc1fcb55e8a212fbdf6a84f12985b0d436dcb51dc58b06e59e435f2"},{"path":"product/runtime/auths-gateway/src/observed_tests.rs","sha256":"a22be071695f89ca2578a4d82fd80d9cb82289213c9bdb8f0470f6b2dc3d9201"},{"path":"product/runtime/auths-gateway/src/observer.rs","sha256":"562ffcaea071c3856e0708dc2fe33516917e145d2438aabb75af0a2b6ba01046"},{"path":"product/runtime/auths-gateway/src/onboarding.rs","sha256":"0dfd7d67e7f02f18a28d1c86ca1b7751d7463f5efcdbe3825de0ad8cbccb3891"},{"path":"product/runtime/auths-gateway/src/operator.rs","sha256":"c3c36954a3e86723df035b47f43b85fc8ab3a6292983522b6258d5c6b4e6c4e4"},{"path":"product/runtime/auths-gateway/src/pending_vectors/attempts.rs","sha256":"47ef3c8cb1c5504d47ae9a0088705ce6715fe86bac1671a34a56fb38a64d0dd3"},{"path":"product/runtime/auths-gateway/src/pending_vectors/bounds.rs","sha256":"e612217307e1da5b0eb5ee6582a7458bb064b8aa8a529fc6555b1b334127ab68"},{"path":"product/runtime/auths-gateway/src/pending_vectors/codes.rs","sha256":"e00670a5ef24eb856cc2c6dcda8e5a263842f55c76bb939c0c27bd9df37e5522"},{"path":"product/runtime/auths-gateway/src/pending_vectors/keys.rs","sha256":"f84c56f2ea9cb34b2b79c72d13fbf2c6585010ec1ed450af58a78c80c1ea1edd"},{"path":"product/runtime/auths-gateway/src/pending_vectors/mod.rs","sha256":"5de8340f0ef71a0b57a09e323fefe473d3160c6ae952b9ba46f714825fbacafc"},{"path":"product/runtime/auths-gateway/src/pending_vectors/outcomes.rs","sha256":"8ec07b9511b48913af36ac999d2aa0e82924289f6acb23ddaeb9f00f99815166"},{"path":"product/runtime/auths-gateway/src/pending_vectors/production.rs","sha256":"772d032e0a0d9c924fb6d6c188528db85a97e71c9e466ddc19eee277baf83614"},{"path":"product/runtime/auths-gateway/src/pending_vectors/recipes.rs","sha256":"51693208270231c25c3aef2a32151d5962114c71e86b88cb9df83181c7af83d9"},{"path":"product/runtime/auths-gateway/src/pre_entry.rs","sha256":"06ea78ca140468bfaba17d9ab04912ad68673954c3203c98fed28272ff2a00c0"},{"path":"product/runtime/auths-gateway/src/property_tests.rs","sha256":"939c081e2e154aaf10186754044bdd65608932c899b970a04a71d0afaa999a93"},{"path":"product/runtime/auths-gateway/src/qualification.rs","sha256":"c597c83de2b94f0374ae54e40a8b57c2a64ec6e254704dca32185f448a908aca"},{"path":"product/runtime/auths-gateway/src/qualification_simulation.rs","sha256":"4ee7cadfb5f46a92713c80f3e55f499cb074d04d8b21ea89d57762af38642bfa"},{"path":"product/runtime/auths-gateway/src/qualification_tests.rs","sha256":"a9178a2e58eea459d5e5dafc53a3999ea7c84dbbffefe6e5eb1f3d098827f028"},{"path":"product/runtime/auths-gateway/src/quorum_tests.rs","sha256":"a78380b019104d688c11a4410082c7e08ded3106aab099954c4f8a1b4ae0fd89"},{"path":"product/runtime/auths-gateway/src/readiness.rs","sha256":"911c3f11892f94641574375e2da4c671c802738f71082f12c39832331052d43d"},{"path":"product/runtime/auths-gateway/src/recipe.rs","sha256":"fecf9af67e7afc48d080e9f8f55f8ce28e49d0e08988717a27908bbda461d6b6"},{"path":"product/runtime/auths-gateway/src/recipe/lower.rs","sha256":"8088b8751ed831dbb48247af953c6290ac95e2a438ccd5bd65eff54be06a1fd1"},{"path":"product/runtime/auths-gateway/src/recipe/review.rs","sha256":"c3eb7d9fdd98dacf52750b14b67ad7b7faaecbbdeafff1e0b049f67d0a2d0139"},{"path":"product/runtime/auths-gateway/src/recipe/runtime.rs","sha256":"c286f0d15f974402822c0810d8e97f98411dc1c557755cc23e540d85e2bb42d9"},{"path":"product/runtime/auths-gateway/src/recipe/source.rs","sha256":"1af50298fb1461e1b24cbdf560d43124d6b1c4c529586c911b1c4797214a69ff"},{"path":"product/runtime/auths-gateway/src/recipe/tests.rs","sha256":"aac551a2481bbc60864c64079bc667dcb232e45a0f41a04073ed84ab5dbe1fce"},{"path":"product/runtime/auths-gateway/src/recipe/validate.rs","sha256":"3711b17605cb7b0d639bc1b180449231f5882a306cab20e2c7782155cc63b0d4"},{"path":"product/runtime/auths-gateway/src/scenario_tests.rs","sha256":"025b53473b3af6900e03bc2fee91aa63ed26a8778bc8c8296fe08f5540794c39"},{"path":"product/runtime/auths-gateway/src/separation.rs","sha256":"be9306848ca5d801767cf6aa14247c8618312ef0e202f00ce0c24c4844491061"},{"path":"product/runtime/auths-gateway/src/simulation_attestation.rs","sha256":"88705e5d45415f760f1e3c3a065c35c48f4f4065bf7b3feb7882246983814d6f"},{"path":"product/runtime/auths-gateway/src/store.rs","sha256":"884bcad4d70d53e12eae9c8517f6ecd0643068ee0be0370015119d04d6cc2516"},{"path":"product/runtime/auths-gateway/src/store_testkit.rs","sha256":"88d82fee42f1b8ef521e8deef0216e938ebeaad822eb2f5020e70b89d145ea70"},{"path":"product/runtime/auths-gateway/src/submit.rs","sha256":"fe2eacb5ad8152eeecbd232c47a356fc72db0098e34fda99532bdf6ef0cf895c"},{"path":"product/runtime/auths-gateway/src/support.rs","sha256":"9fd979029793a22b15d9776a7fc172476fd7c126a80d4080c65d720845becbcb"},{"path":"product/runtime/auths-gateway/src/transport.rs","sha256":"978fc9fb34a33b3db265b5d1c3ce11696a5b9f08529e46dd2a1e7bb62afb9220"},{"path":"product/runtime/auths-kernel-runtime/Cargo.toml","sha256":"926bcd0053e5edd703be8f0027ad009e2982db013c8174c1366edbf38b48087d"},{"path":"product/runtime/auths-kernel-runtime/src/lib.rs","sha256":"7108faae9ca4bc32dc093ddaf885ac8f74f3c9b8d317b7c7d0ed00d1d0ab5972"},{"path":"product/runtime/auths-lifecycle/Cargo.toml","sha256":"c5053c705f8d5554b44e432fe75287a2faf11d644e5a2552e45ad6cd94176af5"},{"path":"product/runtime/auths-lifecycle/src/codec.rs","sha256":"51c4b746749150e8bfcd8ff972b3f6c1ef40bec15583448ec1b471d9ad737dc0"},{"path":"product/runtime/auths-lifecycle/src/digest.rs","sha256":"55282cd7d8aba37713c2dcb5357c9877819edb72a30eadefb860d28d7d3a3eb7"},{"path":"product/runtime/auths-lifecycle/src/identifier.rs","sha256":"31620e8c16de68d3d2d3b84d6353c1a5b9d4151d8247f91fa389b906a8e660f1"},{"path":"product/runtime/auths-lifecycle/src/kernel.rs","sha256":"2abb8c094b36d35b364e74d78091d14c5d6f3d1eb6552cca3b5538dbd017ce23"},{"path":"product/runtime/auths-lifecycle/src/lib.rs","sha256":"afe5a345cd309b5a4f08ffba3170a6d7f16ff914892b3f2006d5208ed24ec736"},{"path":"product/runtime/auths-lifecycle/src/model.rs","sha256":"1af12bbeeeaa65ec02ddcebd7d2c8c92ec1e339fa40ae85b45ab9b01a722a4fa"},{"path":"product/runtime/auths-lifecycle/src/operation.rs","sha256":"78e6d05fcf5d7ea4f148032a3fa26fdfcab77eb346019effe927b08e0d64d732"},{"path":"product/runtime/auths-lifecycle/src/registry.rs","sha256":"53495fe6f7a77422593a41623e28716f1c140dcc1e00f87f6b0ffbfb93a5ef90"},{"path":"product/runtime/auths-lifecycle/src/sealed.rs","sha256":"2e0c03b5813c926c159598bb7890d1523825edbe57cd687d42a5dfb3374a0c19"},{"path":"product/runtime/auths-lifecycle/src/test_support.rs","sha256":"221d9a5c5ef8e1f9f8fd66927484eaf813571d0568470278ecd8fa08a5eec8cd"},{"path":"product/runtime/auths-lifecycle/src/transition.rs","sha256":"011805a28efdd541f44ccea54dd0e34a1e58f4004447ee58013703b4fb6b2822"},{"path":"product/runtime/auths-runtime/Cargo.toml","sha256":"dd29d80a8b4bb56c869f036a6c9998f6e6d1c0383ac5dc03fa42b24abe2a7ff7"},{"path":"product/runtime/auths-runtime/src/lib.rs","sha256":"132ca03d58c14905d3a822a4972f42a22a5ad33a45f2ba9ca44f3a57d810b4b0"},{"path":"product/runtime/auths-runtime/src/production.rs","sha256":"fd4eff4270c571c10fcff6cf27df28961ebf074de8a1a028f500fda1ca376820"},{"path":"product/sdk/auths-approval-quorum/Cargo.toml","sha256":"df904971fca87a24194c5816738a331588fe8f164acb5d70839a35dca5f57114"},{"path":"product/sdk/auths-approval-quorum/src/lib.rs","sha256":"72e4086088a1c0912e4e1a9e980f26a5c280bf5ab16ea39c9f0f58d5096ac2ce"},{"path":"product/sdk/auths-approval-quorum/src/remote.rs","sha256":"9fa029c7c63003adf260f8aec55160c545681c60c1bace0d934e7525862aac7b"},{"path":"product/sdk/auths-approval-quorum/src/remote_tests.rs","sha256":"8a09ce17cfee82a035a74edd535f3ebeb3db0e078784740d3997b17dbdb10197"},{"path":"product/sdk/auths-approval-quorum/src/tests.rs","sha256":"0acbf12be3e497dcc6c2d8f29c9af3b1a7ed8e11b75f6399570d4f217a785060"},{"path":"product/stores/auths-stores/Cargo.toml","sha256":"a32c4b76d17b264be5d40d24497ff237e1ed87822e8bab86229b3db93106dbe3"},{"path":"product/stores/auths-stores/src/gateway_attempt.rs","sha256":"218841602d3d9b23b44625f50e5b54cf275ccedc2d5c89f02d2482b8524472af"},{"path":"product/stores/auths-stores/src/lib.rs","sha256":"e721262edf8a9ab5568ddc27639176312d4a366203b5f73e2d7c11fde2c554e9"},{"path":"product/stores/auths-stores/src/lifecycle.rs","sha256":"3f2cecfda13fdcd7c4a944518646bcc8abd1237f900c0332d5822281f8f9f1cd"}],"schema":"auths.gateway-semantic-closure/1"}
\ No newline at end of file
+{"files":[{"path":"Cargo.lock","sha256":"1b9e6f8e02fd41f6becc2158fca15854138b3bc162ad203854bc2251391da8fc"},{"path":"Cargo.toml","sha256":"ccc17deac72da945940a9ab4be2ce27628c056b977d19429bc1679b19fae30f3"},{"path":"core/adapters/auths-did-keri/Cargo.toml","sha256":"86ba3e3f9a0a70d970568bfb3113b7ba7bbdd81c3eebbb52743dd574a0c0ecaa"},{"path":"core/adapters/auths-did-keri/src/lib.rs","sha256":"a75674842f8a7142004369ccd206ab43e476459293f0ff4bed4d4e659031b3e3"},{"path":"core/adapters/auths-did-key/Cargo.toml","sha256":"d5a3ae07ecde83f91eed4c3785051dd0f44eab6c891754f3309b6b40596707d7"},{"path":"core/adapters/auths-did-key/src/lib.rs","sha256":"b76a3f4528a725df53cfa9242e6e414e1d26caa48468aaed497e28b121f4497c"},{"path":"core/adapters/auths-raw-key/Cargo.toml","sha256":"f49616bc18c19e6aaa7e6a02ccebcdb9f5c1422c1278ef9869e0dbf22ceda7be"},{"path":"core/adapters/auths-raw-key/src/lib.rs","sha256":"45eec1b295dac3bed65cd79c942870b926b06b3f5d38eaebaeb6140ac69315fa"},{"path":"core/crates/auths-algebra-kernel/Cargo.toml","sha256":"0cb078915287b8c47d20fb97b6fa14135ed1c61c4985c4de2753855563c27bba"},{"path":"core/crates/auths-algebra-kernel/src/generated.rs","sha256":"ae4bcbef42a408c7ffa0c6a74a42f7de177a6f14479e51d205acf830bda05960"},{"path":"core/crates/auths-algebra-kernel/src/lib.rs","sha256":"6654667eeb9b353d6648954c9fd2e8838462a82ce6b2803eb49e1366892c68c6"},{"path":"core/crates/auths-assurance/Cargo.toml","sha256":"afcb12a97ddd09d7a264b9ecd3fcaefc865ff2c9b4539ee85b00883420ba333a"},{"path":"core/crates/auths-assurance/src/lib.rs","sha256":"0fb2ec844d674b25a9e20b1601452c89c2454901fa506adfa9bb38ef9a7fa213"},{"path":"core/crates/auths-author/Cargo.toml","sha256":"1e6f0e3bc079ddb558decf0db26b291c6e72fc47327b6df1b7650d41689b215f"},{"path":"core/crates/auths-author/src/lib.rs","sha256":"e14475f2f49715a1869d5fa6b11ff1fbb72497e39097e09f833b0bae7e5b3f2b"},{"path":"core/crates/auths-author/src/observations.rs","sha256":"dda7d6bfa6f4bd7e95a480ca283c2c0e996f69c5c15cb0c7197af328131a7066"},{"path":"core/crates/auths-authority/Cargo.toml","sha256":"4f68212938029f79ebd857b4f3038494eb85d0163d49f07e51376715cf959815"},{"path":"core/crates/auths-authority/src/lib.rs","sha256":"da88d5194de4e17449e69972c0fb042afa3c39798a8f570f30771e77c384f1eb"},{"path":"core/crates/auths-codec/Cargo.toml","sha256":"e20e51826e17814effd9d8925d18d5fbdd6c0d37bf50231e743ec2611314722c"},{"path":"core/crates/auths-codec/src/approval.rs","sha256":"fbff43bf4d57f69f8b7738a8647751cde75c5a932dee65d924a1fe9c7202f91f"},{"path":"core/crates/auths-codec/src/bounded_policy.rs","sha256":"d21f6728be2d8a04b6d94921f2b683f7e61cbb42e0ac6866f1c88a49caa534c3"},{"path":"core/crates/auths-codec/src/decode.rs","sha256":"ba2999584cdf53b1b517ecb9c7ea50d39efee17c666a9a38e2ad1ac94379f778"},{"path":"core/crates/auths-codec/src/encode.rs","sha256":"8491192053e2e15d62b4d2a491de22405b7487ead69ab6f6bca40669ce7ca5f7"},{"path":"core/crates/auths-codec/src/error.rs","sha256":"fecb0737c16deb42d859f1dc7727c698438d9ac12064fd99e0024629642c390c"},{"path":"core/crates/auths-codec/src/hash.rs","sha256":"f88e111f1dbb2cef6b7b5e9c948cb45ff92e366710c8a5f5c636138aa748479e"},{"path":"core/crates/auths-codec/src/lib.rs","sha256":"3b0e5c94c6f34f750b88a9333e1d5c8145746d845d3a23e0d2f61807cf592ea7"},{"path":"core/crates/auths-codec/src/observation.rs","sha256":"38328b0c71ff97a4196eeade8b4faa889527ff1f80bb168468bcd49082a3987e"},{"path":"core/crates/auths-composition/Cargo.toml","sha256":"331e13025eb82588b5269ed3975929eb41c823d588c49e4c296273d85c6fe80e"},{"path":"core/crates/auths-composition/src/lib.rs","sha256":"e42e3b0b6a9985b73c533aeb23799c638979aa27a9ce7d0f10a8d60340570fb1"},{"path":"core/crates/auths-model/Cargo.toml","sha256":"565c15302b0cc9270db1e28efa3862804846850ba933cc17aea8abcec4a65987"},{"path":"core/crates/auths-model/src/approval.rs","sha256":"54cf44a9ef747ec21546a4086aecfb130bc7803b0d646430e6082df51e6e305f"},{"path":"core/crates/auths-model/src/bounded.rs","sha256":"f90496abc0a06c028b4d850af6ddab7d850bfafeaa3d042f1b1246a41d8bbbc9"},{"path":"core/crates/auths-model/src/bounded_policy.rs","sha256":"306a8f049cc7cfc7635fe0e4de47eb7813eed63d3c14c3be2b51586aa4eea450"},{"path":"core/crates/auths-model/src/lib.rs","sha256":"e7cfe6bb9574c6b016aa63e70658c31e0b90903ec6479a49f4a2ceb64f5bad97"},{"path":"core/crates/auths-model/src/observation.rs","sha256":"b3ed3e3cc090afca2896a7d3de7988afccc1ee1ea0ebd91b51a8bc0f20ad8dbb"},{"path":"core/crates/auths-multikey/Cargo.toml","sha256":"858756c1197befcada06134742bffe8d58cf674cadd6c046bf710570e96efaf2"},{"path":"core/crates/auths-multikey/src/lib.rs","sha256":"f5177fffb11e40eeaaa665ba886c541fc1f24d5eacdd6983fb03dc20148733f0"},{"path":"core/crates/auths-ports/Cargo.toml","sha256":"596a9b0f5af3717c16a6d13f7054d8a106bcd9f01f5f72c2235dd2afcf0ab16c"},{"path":"core/crates/auths-ports/src/binding.rs","sha256":"a16c012d40aa4249e3ec54372b9611006d0bfe8c7dc3c741911c2194ad17736d"},{"path":"core/crates/auths-ports/src/diagnostics.rs","sha256":"aa596572503b5a65a918bb2ff38281c144ff53821ee914e9325c852828d852df"},{"path":"core/crates/auths-ports/src/lib.rs","sha256":"8da1debf86f9532df0660155d04298beb1ba3a1204508f8586e2e67cd2fb3449"},{"path":"core/crates/auths-ports/src/path.rs","sha256":"b68fc547b4e4d97a60cbbc1cc1939fc6b34a8233b81f840bd396914a707e3ea1"},{"path":"core/crates/auths-raw-key-core/Cargo.toml","sha256":"98c0bc8527a537512e43a74dd6ba9658c6f0cdc16610cf938e3f9b43785d836d"},{"path":"core/crates/auths-raw-key-core/src/lib.rs","sha256":"98c0500c41697aae28708ad0bf970b9a97583d96a3d2c18d23e7fa351e0f7d12"},{"path":"core/crates/auths-registries/Cargo.toml","sha256":"590e49e58b0622d9d6274fea46796dcdb07dec9550eff295cce6a6427ffa51e2"},{"path":"core/crates/auths-registries/src/lib.rs","sha256":"450d3886d85ee333b7e046317fb5b47cde62483d5f3e41d46265e2caf4c373ec"},{"path":"core/crates/auths-registries/src/template.rs","sha256":"1224039dd2b3ce6fa8357613d8cafbae40400493da7176da202b9ade62fa05e4"},{"path":"core/crates/auths-signature-core/Cargo.toml","sha256":"6bee2230eb30af1be0bd5c2d57bd3e1903497a70028d47290d4ac6e9e0e8bca0"},{"path":"core/crates/auths-signature-core/src/lib.rs","sha256":"db97ecb45a82faa6f96a92738d3af1da92d1ba2f241f2c29114f4b8107dc4ab0"},{"path":"core/crates/auths-signature/Cargo.toml","sha256":"7f7b5682466ba439d8c59579022cb61068cad9fb4bd2d73e1cc21327c902b450"},{"path":"core/crates/auths-signature/src/lib.rs","sha256":"b3d9ac4cc3ba0797b9e1c09f2d6ac05089b25dac4e285a5af28c0165b33d925b"},{"path":"core/crates/auths-verifier/Cargo.toml","sha256":"c43358e06a33324b41a8b93e5cc7279c55f7d21f1f6cfaa539a3fc7b4c27a38f"},{"path":"core/crates/auths-verifier/src/approval.rs","sha256":"c3cae6a49e169051eb555bb092b1069a4d422516d6a2114623a89e5f31ff923f"},{"path":"core/crates/auths-verifier/src/causal.rs","sha256":"9635230f33b13a263573d319ac8a1976addbcc07677514a43e3c7bf28dc04587"},{"path":"core/crates/auths-verifier/src/lib.rs","sha256":"f22dec38dfa47ced8da4f605390da8453273bee1da010f57fd1fdfcb10e0b389"},{"path":"core/crates/auths-verifier/src/observation.rs","sha256":"0bb07f06a0f9524c88427df1c972b1de8ba038a9eeeb15d3134c25125ae84b1e"},{"path":"core/crates/auths-verifier/src/trace.rs","sha256":"2213d72023a1d9e0413a2cf2c51c3dad9e04775d86033a0caaa9d29a8f0036f8"},{"path":"exchange/crates/auths-proof-exchange-model/Cargo.toml","sha256":"34b2db64f93a288e942b5a3918d28cdcbef906e0dc9074ce848018d3d5f7a1fa"},{"path":"exchange/crates/auths-proof-exchange-model/src/lib.rs","sha256":"49fac5aff2339d3a57c41cb8bb2a82852538e8f77ecc4b157c88e264ff31fb55"},{"path":"exchange/crates/auths-proof-exchange-port/Cargo.toml","sha256":"7ff660f1b7579dc570c791090aeb450404fbbf44c92781e289f5ccdde746d739"},{"path":"exchange/crates/auths-proof-exchange-port/src/lib.rs","sha256":"b408fb40c6432edc1facb3d8b3e9433a033f8dc73fe2e0ba5a88c972ca675021"},{"path":"product/config/auths-config/Cargo.toml","sha256":"e4b568ab9f6d84f7d7e6d66443fe825f360db9c27bc7c31d6ddc0647ba3ab304"},{"path":"product/config/auths-config/src/lib.rs","sha256":"fecf00ea65cef9018fda9af2063326d7065b869266d0d9e73a4f06816f1102cc"},{"path":"product/errors/auths-errors/Cargo.toml","sha256":"3b3a199b6460ba75bda84475bb5accd35e1485448a3981038db85bf4772dd350"},{"path":"product/errors/auths-errors/src/lib.rs","sha256":"7feeb9069fa4003f94f1f07c1c94619cf1c883e0dbaaf2187f04d0f742264bdd"},{"path":"product/integrations/auths-credentials-aws-secrets-manager/Cargo.toml","sha256":"716f1f2a117d6d69d8effe9ee9057e24ba3502f20af6a602b06facb638cdabed"},{"path":"product/integrations/auths-credentials-aws-secrets-manager/src/api.rs","sha256":"9ea61dd7ddd93282b073b98e453e6b75b26dc49689c04e9782435d06a3a09e48"},{"path":"product/integrations/auths-credentials-aws-secrets-manager/src/clock.rs","sha256":"73f0a6e0f5903883329fa6683f5888288f759512ac69e5c87efd2b6b87f7d79f"},{"path":"product/integrations/auths-credentials-aws-secrets-manager/src/http.rs","sha256":"14c1fce14e24c984b2a275563a29b6326fb08367922b6e8d37ba38b9cb81aa2b"},{"path":"product/integrations/auths-credentials-aws-secrets-manager/src/identity.rs","sha256":"186de11230c8b33375951c4368609f8976bf9795f171af3fd7fac26207a94361"},{"path":"product/integrations/auths-credentials-aws-secrets-manager/src/lib.rs","sha256":"0b4617d044f5254a5ab22f5a2578f91cc025206a8c1845be39c1c56dcdf7a5df"},{"path":"product/integrations/auths-credentials-aws-secrets-manager/src/names.rs","sha256":"41bb23572d6002e23ee103d6fe70ecd68764861c6f2e48a1c53af998270a4afa"},{"path":"product/integrations/auths-credentials-aws-secrets-manager/src/sigv4.rs","sha256":"39fed883f4802e68d3b41a68fea0aba0737810996b33d13577de70faba004c29"},{"path":"product/integrations/auths-credentials-aws-secrets-manager/src/store.rs","sha256":"fc60ee9c49b2f4f063c449b4c78100c5c4872779e94ca168f796603b380ef6e4"},{"path":"product/integrations/auths-custody/Cargo.toml","sha256":"1e14034185ad1ec24a977fd5f41d0abee219606245c046ce2a82287fadef448b"},{"path":"product/integrations/auths-custody/src/conformance.rs","sha256":"13a2b28bba4a7af5b00bb96d16ecf027d35d5487b314e48a666fb8ed09c2f2e9"},{"path":"product/integrations/auths-custody/src/key.rs","sha256":"c3b1468b8538421a796aeb65a0cd94b942e3788e78741f20dde07b4838e0c3da"},{"path":"product/integrations/auths-custody/src/lib.rs","sha256":"836ad8044440fd250457a46f5c4a980339534a8af6a4daac2643560050b62d39"},{"path":"product/operations/auths-operations/Cargo.toml","sha256":"73dde1e914ba5e1386f8d49c0d730423cc72ae77ec44c6554a6a75cc8e8922b1"},{"path":"product/operations/auths-operations/src/explanation.rs","sha256":"b4a39bcbd5712cb9e735ee7deeb281f3fa6af5ddca973735bf3d87dce427ef11"},{"path":"product/operations/auths-operations/src/lib.rs","sha256":"3f401a56555113b98f57bca172a2660eaecedf76247b83a3fdbb089c133a29cd"},{"path":"product/operations/auths-operations/src/render/json.rs","sha256":"2ae042639960afee123730e945d373bc5dda1d2b3af20ae16c939a7e99b9e418"},{"path":"product/operations/auths-operations/src/render/mod.rs","sha256":"b3b7f6fe473f49227ace5a68bfc3d8dc14b5db8dd9888fd9e207070337b1b8bf"},{"path":"product/operations/auths-operations/src/render/text.rs","sha256":"f216cf3bdb00e264523271b16e7059535ac6b1aabf271cebd7935af155e4bfec"},{"path":"product/policy/auths-bounded-policy/Cargo.toml","sha256":"babd5d200923137ead71d8c2528f847cea323ce98da942a1320b054436612c47"},{"path":"product/policy/auths-bounded-policy/src/arithmetic.rs","sha256":"75e43c5d4eab8d1f71c1daf53839f5f9fe45a400afc1f63a30c6f55ce3632627"},{"path":"product/policy/auths-bounded-policy/src/commitment.rs","sha256":"c5706f2d583454e1a7f570f7ada138d4d864bef0205dfae7c1296774f396bb3f"},{"path":"product/policy/auths-bounded-policy/src/eligibility.rs","sha256":"c9acdfb793074e687fc36efad506be27dbd99ffbf40387bc66c8de39ef238432"},{"path":"product/policy/auths-bounded-policy/src/identifier.rs","sha256":"0006a32bae9eb16c4178cf797df418d2b0a781f2bc6474b741dc1a908141831c"},{"path":"product/policy/auths-bounded-policy/src/kernel.rs","sha256":"99187c28b161ac70bee1afed60711f7f645adb4d1671a01863f9849b369811d5"},{"path":"product/policy/auths-bounded-policy/src/lib.rs","sha256":"76a18638b810e62e594d6e951504679305ce21841388e0c282ad51cd7841b91c"},{"path":"product/policy/auths-bounded-policy/src/receipt.rs","sha256":"3b9eae82f349956d3e7f29b57b254bff5ce67cc703e6cf43445ce2827676d1ad"},{"path":"product/policy/auths-bounded-policy/src/registry.rs","sha256":"4e251ea5833559841f8ee0bd9b66c0c01c39fa5e3a21720a19c4e8adbfcb99d0"},{"path":"product/profiles/auths-profile-api/Cargo.toml","sha256":"7ae85761f58625b5d678f961fd052eebf3e000ef565c7312e3ac857763c0017f"},{"path":"product/profiles/auths-profile-api/src/lib.rs","sha256":"a194b82b20bf8aa864c2334ef6c6aec30b2154ded972862252d5dcefa488ab95"},{"path":"product/profiles/auths-profile-mcp/Cargo.toml","sha256":"db264fc25242f7b7e4133cbe5782545b3107b406c87168ae072683cfb59eac66"},{"path":"product/profiles/auths-profile-mcp/src/arguments.rs","sha256":"a5de696586d507e2aee4a49fee99ef1ef4c51a2e46603f8838f1af6d9009833f"},{"path":"product/profiles/auths-profile-mcp/src/lib.rs","sha256":"d0f8b733a09ef8408336cdebc8948d6a5861012ce6878598b11be7084dfd4280"},{"path":"product/profiles/auths-profile-mcp/src/session.rs","sha256":"166c281fa66333388e4760d6098dad02d3359e1d5de4b8f93312c014f97abb03"},{"path":"product/qualification/auths-recipe-qualification/Cargo.toml","sha256":"b7b15b30b9a3d96d4f089d514266d3a892de6608d30707aa44be31cca89ca9fe"},{"path":"product/qualification/auths-recipe-qualification/src/canonical.rs","sha256":"30a9b50777f0dcb4b606abfee2467d5a975d9f6ad941ab6c13ca3df5b359f667"},{"path":"product/qualification/auths-recipe-qualification/src/closure.rs","sha256":"0ad6aee9bd981e2863fa725f209ec84031db2ff12f4c7ac3aaf5120f0bb86039"},{"path":"product/qualification/auths-recipe-qualification/src/commissioning.rs","sha256":"db135c55e4e3079a73253ebae275536523eb36b545d1c152b4d7dce3c6b76828"},{"path":"product/qualification/auths-recipe-qualification/src/commissioning/tests.rs","sha256":"4dde3736d206099b1bea4c14e0092c6c1287d1af9065958777d7c54331dd8686"},{"path":"product/qualification/auths-recipe-qualification/src/error.rs","sha256":"52d5cf720a1e4579008fdda90a3ac2c31f0b8db57985f121f65e54b4da2b7dd1"},{"path":"product/qualification/auths-recipe-qualification/src/evidence.rs","sha256":"ba45a7409d67a8af26ab811af306412014557477a101ab2a8eb0faf8f2bd4b91"},{"path":"product/qualification/auths-recipe-qualification/src/ids.rs","sha256":"e81a906d571497ab2da753e1532d4443706410e0b75fdd59a7d2d788c7aec403"},{"path":"product/qualification/auths-recipe-qualification/src/launch.rs","sha256":"805154ce931e3977f2a2a2c62909d90f7c0a22dc167bf4f1554dcf19316ffef8"},{"path":"product/qualification/auths-recipe-qualification/src/lib.rs","sha256":"0e24b0f81191df0d078c7f4d73ced784ba9472372bfcdff3c7d0e2fb1687b323"},{"path":"product/qualification/auths-recipe-qualification/src/model.rs","sha256":"3484eb0ec287b23d4682ade7b1b0f8d1335985b3fb180f06fc28716076957d5f"},{"path":"product/qualification/auths-recipe-qualification/src/release.rs","sha256":"ec478e53f70052e566aac7e15be2f9e1b1cd46837ac6990ae955fa864d311586"},{"path":"product/qualification/auths-recipe-qualification/src/vectors/mod.rs","sha256":"1e4e5edc12d921e6b95475b3c1850f58f622c088d24a1fc9398d34019fb2f288"},{"path":"product/qualification/auths-recipe-qualification/src/vectors/schemas.rs","sha256":"c59feec2805621f341a90f81fc005e2c12f7a70f9d264fdeacccb2be4fe1c046"},{"path":"product/qualification/auths-recipe-qualification/src/vectors/verification.rs","sha256":"bdf5746fed4d10a7f577f7878619a917d29aa8b7d522f00d9f774e53eff4c8c1"},{"path":"product/qualification/auths-recipe-qualification/src/verify.rs","sha256":"646ff0e52e915174ad183768706e35e5ce2118b83475455310b930a77a7877f6"},{"path":"product/receipts/auths-receipts/Cargo.toml","sha256":"6001fa1375899953aafcb07c23a6da0c8f57ea7ec08bb9c950b12a60a8cc6f0f"},{"path":"product/receipts/auths-receipts/src/disclosure.rs","sha256":"ba9ed439be0a320c8a01f5b6e9b51361c1a37e076947490d8e4a484ba6bb119c"},{"path":"product/receipts/auths-receipts/src/lib.rs","sha256":"547a57c4160256719cafb7d0ab972d9fc80b0da14fbbce4148c70c414070ff36"},{"path":"product/runtime/auths-connections/Cargo.toml","sha256":"babeb6e1dda3dfb66291713ce1e43dce5c349833d0b50ea4dccbf986030026b2"},{"path":"product/runtime/auths-connections/src/credential.rs","sha256":"7e6fbca8911d5c0c1baddf35572236d419566dd6e3fcbbdfda77c8dd114dbcad"},{"path":"product/runtime/auths-connections/src/kernel.rs","sha256":"a456af19526ef7b19278551fd8d97606b6e7aa5a195e70bdd85d1e0786090ce8"},{"path":"product/runtime/auths-connections/src/lib.rs","sha256":"79209249a0acba0e854d9937255ea0e30a9a6f8e1c2baad77ab83646d392de22"},{"path":"product/runtime/auths-connections/src/model.rs","sha256":"de34a089cac9702acb889b7602b3e4732306c9e61029ec892582b7e02b73fc77"},{"path":"product/runtime/auths-connections/src/registry.rs","sha256":"e57640e003e72ad3235c567c2bb944a262bd64e673f023b62ea091094cc7e946"},{"path":"product/runtime/auths-gateway-kernel/Cargo.toml","sha256":"34f864e67656af4ac7e5c7a347a23f01fc8ab31365b8e23c0ef474483b14a183"},{"path":"product/runtime/auths-gateway-kernel/src/construct.rs","sha256":"e9d65744bae273d65face4285e5d2da75c06a92ab48f7b8333d496604e31c63c"},{"path":"product/runtime/auths-gateway-kernel/src/lib.rs","sha256":"cfda6f8453426564c085475f183b6f790b438768a19668f1b65d33ba59dc8632"},{"path":"product/runtime/auths-gateway-kernel/src/order.rs","sha256":"efc2f2eef68232fe3082e1607e6f1e87d4a19cbb2f492bdc6e71916a2461b6a8"},{"path":"product/runtime/auths-gateway-kernel/src/outcome.rs","sha256":"ac0245699457c802a70cdd6f52a01dcb4777c487b58ed7dc9cfc99345465ccaa"},{"path":"product/runtime/auths-gateway-kernel/src/ratio.rs","sha256":"7ace22264812e6fc1ef3a62afee402cf43494bb71ae0d3de219ad1b39d690bcb"},{"path":"product/runtime/auths-gateway-kernel/src/recovery.rs","sha256":"9f46dcd1685ab6e4d1ba04a3fc286a9f3afee8fb47829be1fbc670a7926a2dc6"},{"path":"product/runtime/auths-gateway-kernel/src/transition.rs","sha256":"9c9c292397aa640c2f002ffb5befe110f86d0aa246f3ffe99440718e5174af3c"},{"path":"product/runtime/auths-gateway/Cargo.toml","sha256":"713a55ac7dfbc9c0d89b2d9d370278ed5f7a7a2d76460d69248db0da5c11ebdf"},{"path":"product/runtime/auths-gateway/src/admin.rs","sha256":"5d0f1b583a4cacdcbc62521ed1a5730ed6b57d42bf826708a0c5e591272c4a12"},{"path":"product/runtime/auths-gateway/src/admin_boundary_tests.rs","sha256":"7ef23fdaf512cfa6315aed72577ccf0c4d04725f5b4fa4b5f0482e27e5e7ba1c"},{"path":"product/runtime/auths-gateway/src/app.rs","sha256":"04c7f98ea3a9506fa4a2da8039455dccc4eebfb2452e7fb3c612aa2aa9bc405f"},{"path":"product/runtime/auths-gateway/src/app/content.rs","sha256":"5de081c7f31f0b1af466d500372e123f2927918d63ed2da0adc48d5c4178e8f8"},{"path":"product/runtime/auths-gateway/src/app/content_tests.rs","sha256":"6a239a15732dd279abb35a72721eb2034bf5aac79439edc7c34aa8d4fdfd5212"},{"path":"product/runtime/auths-gateway/src/audit.rs","sha256":"aebb55cfe1305c6de26fda03989e138a6731ec57d1cfebf4114886bfb6b93d26"},{"path":"product/runtime/auths-gateway/src/audit_export.rs","sha256":"6cc3d70ef4cc64026de4ec4e094979a95fee6fadb86707b0e09515ad754632be"},{"path":"product/runtime/auths-gateway/src/audit_tests.rs","sha256":"8cf1ab4cd2789973791805b07216d423690645aa4f28307f87dbcb803bd2ad22"},{"path":"product/runtime/auths-gateway/src/bin/auths-gateway-harness.rs","sha256":"12b0c24d36b2e6a3be392d3c081a783b355fdd67dc5b54bacc0410e803116b8c"},{"path":"product/runtime/auths-gateway/src/bin/auths-gateway.rs","sha256":"97c4388a3d95307e083c55152b02aaf288b3cb7b7c11ac863cd1932f274fc029"},{"path":"product/runtime/auths-gateway/src/binding.rs","sha256":"663b0ce878319511aa7d05722ab4ab2fa9be18817884ffb2ee9f131ff1e382d1"},{"path":"product/runtime/auths-gateway/src/bounds.rs","sha256":"d0658585765d4d8f9865f65d23d850ec196eeca23fe65a66a55f1f1bc9b2c9b9"},{"path":"product/runtime/auths-gateway/src/bounds_aggregate_tests.rs","sha256":"ab34f227b7c8692ba1da88ccb32bb3687a66adf8353d8282fae905d4a88eb612"},{"path":"product/runtime/auths-gateway/src/bounds_tests.rs","sha256":"80b116ccf65bda2d7f2fbcd3b26fcf82b0a4bd3dce3a5df7842828cb271d7741"},{"path":"product/runtime/auths-gateway/src/candidate_runtime.rs","sha256":"5160b5a26668d2a177cadb472f95b10a2c97d2307ac444e9be8046f01b41909d"},{"path":"product/runtime/auths-gateway/src/commissioning_budget.rs","sha256":"f5c38e9e07df3cb99b6bd664f33eae414b9f3dfef4932d69c19a9b174906dc1b"},{"path":"product/runtime/auths-gateway/src/commissioning_budget/tests.rs","sha256":"da104a1643c37515deec6397bca6db2054f2c9158ec3d230a1b1ec9c7d0f659f"},{"path":"product/runtime/auths-gateway/src/commissioning_floor.rs","sha256":"7fa6991c5ccd5ae1403be84b3da390323fab09f36e15ac01fd7a38ea350c6fdb"},{"path":"product/runtime/auths-gateway/src/commissioning_floor/tests.rs","sha256":"6b663d6e556adc0b455b81845bc4394d82c6c4de0c5f43e1310f3ddc6e4e40f7"},{"path":"product/runtime/auths-gateway/src/commissioning_session.rs","sha256":"919fe57fafa563f770402be66b4e32f1f0d652f09fe1276a62e41b50d66f8696"},{"path":"product/runtime/auths-gateway/src/commissioning_session/tests.rs","sha256":"48d44a5fccea737930f24c49a397428edc439525330c83a868b5113a48f122ab"},{"path":"product/runtime/auths-gateway/src/configuration.rs","sha256":"f6b35e3f1257d25628244d0908b9c29afea7875c4bc9ddca832a2e8cd1418f9e"},{"path":"product/runtime/auths-gateway/src/connection.rs","sha256":"5d5cc58bec7665e44c625b7a6195a35e513593681ada669294cd6bf47921143f"},{"path":"product/runtime/auths-gateway/src/credential_journal.rs","sha256":"7297fccaee5cd8622928e8e0a42bba50da7941970a6f682ecd300c41ee6509da"},{"path":"product/runtime/auths-gateway/src/disclosure_transport.rs","sha256":"614c975a4b9810f40092a2f9cdd022308b75e649167dbc486c9907f7570d8d45"},{"path":"product/runtime/auths-gateway/src/echo_verify.rs","sha256":"6d453051665f823abd7354d5c50031d9524cf47d6092d2f09080ae814e3b9575"},{"path":"product/runtime/auths-gateway/src/effect.rs","sha256":"4c67f575f7597652dbe1e27ccfedd6f9c3d09e4d146756be9b8df098b4fd3f2f"},{"path":"product/runtime/auths-gateway/src/effect/acquisition.rs","sha256":"24d73218c07db78b79fe6ad96da6acacee7aea3df68a453f39a06c5cf8298e7f"},{"path":"product/runtime/auths-gateway/src/effect/authority.rs","sha256":"2490494c3a1636decbfd04b2dbac71c47e49be26b61cc6cf41319a13450bb4a0"},{"path":"product/runtime/auths-gateway/src/effect/candidate.rs","sha256":"d1efc3ec7e1fa642af26f4e4c6a25924c60821dc2e96bd0b458d9292d3866678"},{"path":"product/runtime/auths-gateway/src/effect/candidate/proposal.rs","sha256":"d78574109f41300edfcd7ad38155439ffc3adb9a799554d75d5484738ef174f1"},{"path":"product/runtime/auths-gateway/src/effect/capacity.rs","sha256":"300fa32e09dab7c679b8c0a69bb4e3c1e924ed2ce1416090a2aa477a7eec9f0f"},{"path":"product/runtime/auths-gateway/src/effect/content.rs","sha256":"27b83afdc6b5612299f527348a50dd16ea6dce98c3f2523b596a56004850c994"},{"path":"product/runtime/auths-gateway/src/effect/disclosure.rs","sha256":"0acb5ac805e91c952c8f0028db03ae607c3753cbd08a76035f2d1b4fa6fe6244"},{"path":"product/runtime/auths-gateway/src/effect/disclosure_auth.rs","sha256":"32fa6f040ae46f3323df6e153c6ae810eccf32d22b9945e5565e7189dc877eac"},{"path":"product/runtime/auths-gateway/src/effect/prerequisite.rs","sha256":"fbeaab5f5422becbfeeff5c94d41a3bb99a2dd9c571c23895a594c2a3b4cd1fa"},{"path":"product/runtime/auths-gateway/src/effect/release.rs","sha256":"fd841cbce131136acd888bf35f08d0dc6ba524f98a24bdeb9ee5c20bcb81fe4d"},{"path":"product/runtime/auths-gateway/src/effect_binding_boundary_tests.rs","sha256":"83d8e0edf524a2be0d2c494e4b9897cab0ddcf61184828d716bae10cbaf904b4"},{"path":"product/runtime/auths-gateway/src/effect_candidate_staging_tests.rs","sha256":"12d16a9d2752a1791c8814be1857db9f6703d8e368da3bad1d6fd128be01cb28"},{"path":"product/runtime/auths-gateway/src/effect_candidate_tests.rs","sha256":"55f85891a8559fe3dea618b09f6e790749b406c8fee501306a57837af6319ab6"},{"path":"product/runtime/auths-gateway/src/effect_content_tests.rs","sha256":"59d10b48daa531bb7f46b0d30e077476c83ecdf4bddfdf3c3bda0621ca176f6f"},{"path":"product/runtime/auths-gateway/src/effect_disclosure_delivery_tests.rs","sha256":"9586b84b045561edbcf8276cd9769b412240f04248e240cd29a5068d4cd03a00"},{"path":"product/runtime/auths-gateway/src/effect_disclosure_journal_tests.rs","sha256":"a044537d6ee65bf3a3c3e650199d1c9ac18955ef6b6e6ef0d5793f4a5c3a253a"},{"path":"product/runtime/auths-gateway/src/effect_disclosure_tests.rs","sha256":"5389b03d7d0cd5b23abaa3c396450fa4018102c942b27af35aef81e65b9bb841"},{"path":"product/runtime/auths-gateway/src/effect_native_tests.rs","sha256":"cf7263f6b3d5f5749a76d970f0c73d3a0410832ac8f47d8770b4a4754928f21a"},{"path":"product/runtime/auths-gateway/src/effect_prerequisite_tests.rs","sha256":"0e76695c5d00c6806de7f1edc9d873d036d837c1fd0989530b820a41b1beb472"},{"path":"product/runtime/auths-gateway/src/effect_read_budget_tests.rs","sha256":"9bf8491f0da471e07b5d96dbfdbdf5fbe5a47fec11700dca55a6ac05a1ebb68b"},{"path":"product/runtime/auths-gateway/src/effect_recovery_prerequisite_tests.rs","sha256":"e6235b83ac159d9446d79e6e072dd4aa9d0447a68055be3b32de6450a2300699"},{"path":"product/runtime/auths-gateway/src/effect_review_approval_tests.rs","sha256":"6e5c5a72ec93255d7d8f219ae8cf4401431644ff03d624164645a1e9d2b2e87e"},{"path":"product/runtime/auths-gateway/src/engine.rs","sha256":"58525d87087da11ac8c86d4765e2af51f5176898f244e189a48188c61c578851"},{"path":"product/runtime/auths-gateway/src/entry.rs","sha256":"053e4cce4f63d4406eea47b536af4d344bcee073844827390aa6b4eb43c69638"},{"path":"product/runtime/auths-gateway/src/entry_tests.rs","sha256":"6fdd7a0373b3eca51d3062682f1187ee1318add7d2a1eb0e6b0a05ea36dadaae"},{"path":"product/runtime/auths-gateway/src/execution_witness.rs","sha256":"e8b1a8e41d5e3fa309760067849d390ef065feeb46701a83391832645a2f000c"},{"path":"product/runtime/auths-gateway/src/fuzzing.rs","sha256":"2dcf0a15e407a9180c7fecb7eecfaf4674698aa997b1f5de9a3649a49baebd1b"},{"path":"product/runtime/auths-gateway/src/generation_floor.rs","sha256":"3e82c3832685c43b546ebd12857ebecbfa6be29d8749654ab65e3e520ad552fb"},{"path":"product/runtime/auths-gateway/src/harness.rs","sha256":"94469e437df7b296bb12a4f311cf4bbfdbaaed31bf72983e83136f4b3707b34d"},{"path":"product/runtime/auths-gateway/src/harness/authoring.rs","sha256":"de511d89bf6bb72e443558caf0c4835dad2908b5fe5c13a3a02483e85dd7d354"},{"path":"product/runtime/auths-gateway/src/lib.rs","sha256":"c03c3a73276ace55b9b8aa1dd3dc6ae56b3f01850ff2a4526d70ab38797b40ee"},{"path":"product/runtime/auths-gateway/src/listener.rs","sha256":"aa08239bcbc1fcb55e8a212fbdf6a84f12985b0d436dcb51dc58b06e59e435f2"},{"path":"product/runtime/auths-gateway/src/native_budget_tests.rs","sha256":"d6de965afc7f794e611545744fa515f833abf315157519ed8390c195a7dac8c0"},{"path":"product/runtime/auths-gateway/src/native_cost_store.rs","sha256":"d1acdd7e4076e3a32f312dce74a01e5dcbccdb1f8912758524b7841d61377eed"},{"path":"product/runtime/auths-gateway/src/native_cost_tests.rs","sha256":"45bcb4e9174d4c23495ac734a42b5f8b7735c14d4599cd191d50f6547aaf374c"},{"path":"product/runtime/auths-gateway/src/native_crash_tests.rs","sha256":"46998a973388884d902a6b07e12fa342fb08ea77363a43a41b0b0a7a5ccbcde9"},{"path":"product/runtime/auths-gateway/src/native_dependency_tests.rs","sha256":"a8aafce830175c7aa26b3831076703265af8dcbec9ad7b7f34a851c9bc4ebc96"},{"path":"product/runtime/auths-gateway/src/native_differential_tests.rs","sha256":"4c9ef638d4eb38c564ed94d0c6cdc7cf0ede0bff669face453fa58809c1174f1"},{"path":"product/runtime/auths-gateway/src/native_disclosure_budget_tests.rs","sha256":"bb2a6f76f06a6d42d69698f7d22d0b4aeb4dc857e442042d65582097567009fb"},{"path":"product/runtime/auths-gateway/src/native_disclosure_crash_testkit.rs","sha256":"4beb946a288e4fe082cc0917f4200b18b8154eff98eca1ea3d0c830ffdc3625d"},{"path":"product/runtime/auths-gateway/src/native_disclosure_crash_tests.rs","sha256":"425e423f04d92794656d4da915125e4d0a8be25ec85fb88244ee0c27ece8ec44"},{"path":"product/runtime/auths-gateway/src/native_disclosure_entry_tests.rs","sha256":"50f5dc35a9a8ef462b8833296cb313b2acec81d544b7e317c7f188741d0786db"},{"path":"product/runtime/auths-gateway/src/native_disclosure_file_fault_tests.rs","sha256":"72d67bb0d3fd77fb93c4f5d3ab0afa5a939742b5a47acf006a511d99d72cee9f"},{"path":"product/runtime/auths-gateway/src/native_disclosure_pg_batch_tests.rs","sha256":"a8b8a90210d9f7f0012e40701c09c524527f93f9e56933d89b58a38464cba702"},{"path":"product/runtime/auths-gateway/src/native_disclosure_pg_commit_tests.rs","sha256":"1e14503140805c12883e1c288b75029aa7fd9ebe8c5d490a4c503161c4a27a4b"},{"path":"product/runtime/auths-gateway/src/native_epoch_tests.rs","sha256":"e37d34e699baa2b84578bd4c378975dbdceb407de51861c2217afdcf7b48e3b9"},{"path":"product/runtime/auths-gateway/src/native_file_fault_tests.rs","sha256":"e1077a0059aa69918fb7e0a3481edd19af45a7343ce876583fbdbc96c4291b21"},{"path":"product/runtime/auths-gateway/src/native_generated_history_tests.rs","sha256":"5a5e308c1e4bc32cac31badf91e9453bed62e3e90f1b6585fc522bb4038990f9"},{"path":"product/runtime/auths-gateway/src/native_guarded_mutation_crash_tests.rs","sha256":"5329f72e7f22399a17627b058c2035c243e1b6ba60c9dfcd1c834258244cc440"},{"path":"product/runtime/auths-gateway/src/native_information_journal_tests.rs","sha256":"dd0821e8337e56189708496ca85cc2f2da2d6a4ea8b53db5aa0412beead69864"},{"path":"product/runtime/auths-gateway/src/native_information_replay_tests.rs","sha256":"71edb47c63d194950d9898107487bd36b3f3eeba8405ff81d85cec2af2d2ecf5"},{"path":"product/runtime/auths-gateway/src/native_journal_tests.rs","sha256":"d803ff53f90ce2a8c29808b4f874cd81954a3a5c52e19b519ab22f0543808fe1"},{"path":"product/runtime/auths-gateway/src/native_mutation_budget_tests.rs","sha256":"41d7f560d6b04e750acb2f8add2544745743ea9fea760a8ffd5ec5aab307e94c"},{"path":"product/runtime/auths-gateway/src/native_mutation_crash_tests.rs","sha256":"3e1befd35e7264024666331b52ca2245540ffa8f9927ed62e3aa45c7c8cc8950"},{"path":"product/runtime/auths-gateway/src/native_mutation_entry_tests.rs","sha256":"219a1986cf7440d31734db9db3c25f7997486a2fc29fee0a1ac8c7f0ee9ce803"},{"path":"product/runtime/auths-gateway/src/native_mutation_file_fault_tests.rs","sha256":"252f6ca9455b291b0873d983c8fccc75254d947b89a3fe9a1de0155c8ee53930"},{"path":"product/runtime/auths-gateway/src/native_mutation_journal_tests.rs","sha256":"181bf0a196805e50944cff4fe2315459eba877e7d7b5b90bf3c99e3353a8cad7"},{"path":"product/runtime/auths-gateway/src/native_mutation_pg_batch_tests.rs","sha256":"137f0e2f0e8d950cea3b38a7a03113c61177e54083f11eecc34b994ce41728f1"},{"path":"product/runtime/auths-gateway/src/native_mutation_pg_commit_tests.rs","sha256":"5d9a3b8a1601c83c66db3aefd904c16233ce2700d070ba5c9c4f81f4051c5fad"},{"path":"product/runtime/auths-gateway/src/native_mutation_replay_tests.rs","sha256":"3a874e1416cbb29cce70930f9955a70b4a5ad46e2cbca8d55f1acf070215cfbd"},{"path":"product/runtime/auths-gateway/src/native_partial_upload_tests.rs","sha256":"5fa867219d61b7c8a563e7cf5111291f15b8a11464623859be8aa9ea858170ea"},{"path":"product/runtime/auths-gateway/src/native_postgres_commit_proxy.rs","sha256":"1c95b3f8af48b78dc4f1c1baa3b26c1aa7bbe868c23aefbb32f1f0ae592f0045"},{"path":"product/runtime/auths-gateway/src/native_postgres_commit_tests.rs","sha256":"f021b110991ba3ae497705d0aaa0350a210f09b6f102fd71ac2f8eb4ff8f30c0"},{"path":"product/runtime/auths-gateway/src/native_postgres_fault_tests.rs","sha256":"0c6edea869a20765b322fa69c48e580aa0664e726b6cef483d4eba49803135ba"},{"path":"product/runtime/auths-gateway/src/native_process_tests.rs","sha256":"44c41da202499180fb9ab8d8f10bf5efc5c12f7a2c1fb9397f31e7eaf08c3425"},{"path":"product/runtime/auths-gateway/src/native_public_evidence_tests.rs","sha256":"dac3a70f4535a14aab49ceed5a6b5fd7002e199270bb5a120f5bc8edfa5398bf"},{"path":"product/runtime/auths-gateway/src/native_recovery_actor_tests.rs","sha256":"bfdcdbaa0ee1d6822c57e5ab9ca524e731c393f9289c5508b61a0fc1d0b065e5"},{"path":"product/runtime/auths-gateway/src/native_recovery_authoring.rs","sha256":"9f4e8008160b349a52267eacd6a1c1e5c31c3c1cdfd71b574162ccaa9f7430b6"},{"path":"product/runtime/auths-gateway/src/native_recovery_history_tests.rs","sha256":"0c6ec8fe18fd4cc8496f04c3bc38552c3602d7254cae979fac76aeb7c7e4f41b"},{"path":"product/runtime/auths-gateway/src/native_recovery_limits_tests.rs","sha256":"f147c2e78cbe364fb3fd0c97bbbacdad86d77ca8c0b8575d9fa108968f82379a"},{"path":"product/runtime/auths-gateway/src/native_result_wire_cases.rs","sha256":"2b93c8ed96d7092216410bb99435411fd57497284639f56e0120b16639a98eb2"},{"path":"product/runtime/auths-gateway/src/native_retention_tests.rs","sha256":"93844bfe86b2983f9b26134154d9148c641ad9120bd294ebdaf648ead5ba2666"},{"path":"product/runtime/auths-gateway/src/native_scope_tests.rs","sha256":"8d3f362df9fadc10c22f0483f3a16d238d0f82f393a628727adf0b728707673e"},{"path":"product/runtime/auths-gateway/src/native_threshold_testkit.rs","sha256":"74e6b6ed1b77b463effff132fd8af9b3b59aeee64704a8fe8a19ed3ca9d79bf7"},{"path":"product/runtime/auths-gateway/src/observed_tests.rs","sha256":"91223493fefd3efc3fec58db225f00786b539e3b24027419e919d8bd7310059c"},{"path":"product/runtime/auths-gateway/src/observer.rs","sha256":"1c8988dd8807566d5a152e1894aa0f4c80c3ba7302ac910a4a6978d4f31b323d"},{"path":"product/runtime/auths-gateway/src/onboarding.rs","sha256":"6180e79d7698bcd4dd6ffdf32dcfbe14427f96de92f4ae4ceac4dda651f3671b"},{"path":"product/runtime/auths-gateway/src/operator.rs","sha256":"fe3b11fcf8e2394ab955b09a5b9b1f979c46f9e07a4f16e1a3ca160d675f1ace"},{"path":"product/runtime/auths-gateway/src/pending_vectors/attempts.rs","sha256":"e4c08fda8ea764701ab7b07285260c770da5f585c1e2adc406216eca79095120"},{"path":"product/runtime/auths-gateway/src/pending_vectors/bounds.rs","sha256":"e612217307e1da5b0eb5ee6582a7458bb064b8aa8a529fc6555b1b334127ab68"},{"path":"product/runtime/auths-gateway/src/pending_vectors/codes.rs","sha256":"6fb3f1679b0fd2ca7a8ac6aaee65f586b9eb809d9d7c9e26d3dd8ae2bfd10307"},{"path":"product/runtime/auths-gateway/src/pending_vectors/keys.rs","sha256":"b0b1f79d2fadc06e91378a8250223a429163138e4b86e90e09bc207d80fabf86"},{"path":"product/runtime/auths-gateway/src/pending_vectors/mod.rs","sha256":"5de8340f0ef71a0b57a09e323fefe473d3160c6ae952b9ba46f714825fbacafc"},{"path":"product/runtime/auths-gateway/src/pending_vectors/outcomes.rs","sha256":"a8960d8e055ac31827e379513f57f00ed868b559079fa2a4687e1b055d48e3ca"},{"path":"product/runtime/auths-gateway/src/pending_vectors/production.rs","sha256":"95d8e5f51383fcc6e9a1209b89fe1d9c5bd1dc2b9c8424d1f151ab94976db381"},{"path":"product/runtime/auths-gateway/src/pending_vectors/recipes.rs","sha256":"5d550e3708ea41a9547ae2207699bde5ff4435f5a1ed7afa4378fb8defb51bb4"},{"path":"product/runtime/auths-gateway/src/pre_entry.rs","sha256":"3f2554feda72fa047229d427504676a1fb14894bdd377055a3108d7b3e3587cd"},{"path":"product/runtime/auths-gateway/src/prerequisite_race_tests.rs","sha256":"683cc46b3718d12dbb42668dae4bc6ec7e7508dcbb6b6232fead2e3e4f510d0d"},{"path":"product/runtime/auths-gateway/src/property_tests.rs","sha256":"85a77aaf01d9be51e1dd7c72461653a099c51b0ca9c818800d991abe262b198a"},{"path":"product/runtime/auths-gateway/src/qualification.rs","sha256":"30686268fee3d3ddc49301e58b779368f6cb6d3b38e6277d86728f8b7102af3e"},{"path":"product/runtime/auths-gateway/src/qualification_simulation.rs","sha256":"4b6961472fb825e535a64f0b6eb92278cb50036755dfa03e8a63ce8dd00827ae"},{"path":"product/runtime/auths-gateway/src/qualification_tests.rs","sha256":"2712cac071c15dd20ed28159a63f8c6488f9d8221d8c34c3f54577f1a6831913"},{"path":"product/runtime/auths-gateway/src/quorum_tests.rs","sha256":"a6533de9bc2689ec632f009404daf363494e224e7b89b1c65bb2e2cd4fc6421d"},{"path":"product/runtime/auths-gateway/src/readiness.rs","sha256":"911c3f11892f94641574375e2da4c671c802738f71082f12c39832331052d43d"},{"path":"product/runtime/auths-gateway/src/recipe.rs","sha256":"fdfc6a1eda078f3c1bbc35bf019a34c78395611d8e3c4ece5f382aa795024a35"},{"path":"product/runtime/auths-gateway/src/recipe/content.rs","sha256":"e3f8aa8cf2dd9ef5777d9d1e76331052c67be87873db0c2cd6a8f48ff940d9d6"},{"path":"product/runtime/auths-gateway/src/recipe/effect_review.rs","sha256":"8f94896f0d124162d3dbd080c44582bb23ad181bb582ff5cd3d049d75fcb6d77"},{"path":"product/runtime/auths-gateway/src/recipe/lower.rs","sha256":"81118ea7ae6c2d9984b88fe9991f91283070859d56923ee4fdfe9a50eee9ccb1"},{"path":"product/runtime/auths-gateway/src/recipe/response.rs","sha256":"0b4a074eb05b98685697094ed4775c5fe588b3e34f954aa1844135fe14c94487"},{"path":"product/runtime/auths-gateway/src/recipe/review.rs","sha256":"efb557741edc08d7c53770299afdb9c13a07de28426290030210c3da26826c0b"},{"path":"product/runtime/auths-gateway/src/recipe/runtime.rs","sha256":"0853711dc8549b2ed6949e09faca15163f9ed708e3129d1ca55e0c450cab76cd"},{"path":"product/runtime/auths-gateway/src/recipe/source.rs","sha256":"2fa3605998a959ccb61e1042dda81e5bc0d2e553af38e15101cff109621b0f4e"},{"path":"product/runtime/auths-gateway/src/recipe/tests.rs","sha256":"c4cf62a9f882071f9cec205d074953f758c6492121911a62e0dad19f4aed5348"},{"path":"product/runtime/auths-gateway/src/recipe/validate.rs","sha256":"c1a0393c0afbadbad01f119fb9257020abea686916c992c7e8a18dc92627f4f4"},{"path":"product/runtime/auths-gateway/src/scenario_tests.rs","sha256":"cefbaefaffb3147ac43acae67c2b2c9f30ed2bb1c014af29083d02e43e482312"},{"path":"product/runtime/auths-gateway/src/separation.rs","sha256":"be9306848ca5d801767cf6aa14247c8618312ef0e202f00ce0c24c4844491061"},{"path":"product/runtime/auths-gateway/src/simulation_attestation.rs","sha256":"88705e5d45415f760f1e3c3a065c35c48f4f4065bf7b3feb7882246983814d6f"},{"path":"product/runtime/auths-gateway/src/store.rs","sha256":"fdf746572649f554f24089d0dd6322d964be1c12d1695dd5d9f8d86079b23a82"},{"path":"product/runtime/auths-gateway/src/store/acquisition.rs","sha256":"56ee776420bbd00312f732a97a4f342c1b953605290c7c5a1a10497669786e05"},{"path":"product/runtime/auths-gateway/src/store/acquisition/candidate.rs","sha256":"4300b9ab31986d1150ef93050cf5b827e84d9a34c60fd19ea474d7d5a01fc984"},{"path":"product/runtime/auths-gateway/src/store/acquisition/content.rs","sha256":"92f87e4556335af8e1c4157e60e6936ca02b271381569e14510bc2c35e697ece"},{"path":"product/runtime/auths-gateway/src/store/acquisition/recovery.rs","sha256":"87363b666c16cef5f4566db25b3cbec93f3083b2bf8508a5c55ce78667bc4a6c"},{"path":"product/runtime/auths-gateway/src/store/capacity.rs","sha256":"d25c4a0d753317270cb352246c70642bf0659e6448d14061356e4e9ea7ed3811"},{"path":"product/runtime/auths-gateway/src/store/disclosure.rs","sha256":"6c8f0b788501dbacf1d51ac5adc78e35c40e0bd51b7fe84c97f1609f1008744a"},{"path":"product/runtime/auths-gateway/src/store_retention_tests.rs","sha256":"a9e300fde3c45999764ee05c4c6ea46cfb9ffcaa51b82337a322925ad50f4f50"},{"path":"product/runtime/auths-gateway/src/store_testkit.rs","sha256":"5bcdabc1c67c4d44f124786edab0d0a97c6dbd22a25df0ba576cdc30e88f5b50"},{"path":"product/runtime/auths-gateway/src/submit.rs","sha256":"06d554107f389a1e445ef0505d12ee75d557132a25be17bdda0da0071fe73bfd"},{"path":"product/runtime/auths-gateway/src/support.rs","sha256":"d4865f95a7a2fdfe297f3f73687c05ab64a6b911814daf5f2b6c703ce58262ce"},{"path":"product/runtime/auths-gateway/src/transport.rs","sha256":"0ad16c63c19cc41d11c485625c8fc0617aede21401d16ec814401f963538bb46"},{"path":"product/runtime/auths-kernel-runtime/Cargo.toml","sha256":"926bcd0053e5edd703be8f0027ad009e2982db013c8174c1366edbf38b48087d"},{"path":"product/runtime/auths-kernel-runtime/src/lib.rs","sha256":"7108faae9ca4bc32dc093ddaf885ac8f74f3c9b8d317b7c7d0ed00d1d0ab5972"},{"path":"product/runtime/auths-lifecycle/Cargo.toml","sha256":"c5053c705f8d5554b44e432fe75287a2faf11d644e5a2552e45ad6cd94176af5"},{"path":"product/runtime/auths-lifecycle/src/codec.rs","sha256":"51c4b746749150e8bfcd8ff972b3f6c1ef40bec15583448ec1b471d9ad737dc0"},{"path":"product/runtime/auths-lifecycle/src/digest.rs","sha256":"55282cd7d8aba37713c2dcb5357c9877819edb72a30eadefb860d28d7d3a3eb7"},{"path":"product/runtime/auths-lifecycle/src/identifier.rs","sha256":"31620e8c16de68d3d2d3b84d6353c1a5b9d4151d8247f91fa389b906a8e660f1"},{"path":"product/runtime/auths-lifecycle/src/kernel.rs","sha256":"2abb8c094b36d35b364e74d78091d14c5d6f3d1eb6552cca3b5538dbd017ce23"},{"path":"product/runtime/auths-lifecycle/src/lib.rs","sha256":"afe5a345cd309b5a4f08ffba3170a6d7f16ff914892b3f2006d5208ed24ec736"},{"path":"product/runtime/auths-lifecycle/src/model.rs","sha256":"1af12bbeeeaa65ec02ddcebd7d2c8c92ec1e339fa40ae85b45ab9b01a722a4fa"},{"path":"product/runtime/auths-lifecycle/src/operation.rs","sha256":"78e6d05fcf5d7ea4f148032a3fa26fdfcab77eb346019effe927b08e0d64d732"},{"path":"product/runtime/auths-lifecycle/src/registry.rs","sha256":"53495fe6f7a77422593a41623e28716f1c140dcc1e00f87f6b0ffbfb93a5ef90"},{"path":"product/runtime/auths-lifecycle/src/sealed.rs","sha256":"2e0c03b5813c926c159598bb7890d1523825edbe57cd687d42a5dfb3374a0c19"},{"path":"product/runtime/auths-lifecycle/src/test_support.rs","sha256":"221d9a5c5ef8e1f9f8fd66927484eaf813571d0568470278ecd8fa08a5eec8cd"},{"path":"product/runtime/auths-lifecycle/src/transition.rs","sha256":"011805a28efdd541f44ccea54dd0e34a1e58f4004447ee58013703b4fb6b2822"},{"path":"product/runtime/auths-runtime/Cargo.toml","sha256":"dd29d80a8b4bb56c869f036a6c9998f6e6d1c0383ac5dc03fa42b24abe2a7ff7"},{"path":"product/runtime/auths-runtime/src/lib.rs","sha256":"132ca03d58c14905d3a822a4972f42a22a5ad33a45f2ba9ca44f3a57d810b4b0"},{"path":"product/runtime/auths-runtime/src/production.rs","sha256":"fd4eff4270c571c10fcff6cf27df28961ebf074de8a1a028f500fda1ca376820"},{"path":"product/sdk/auths-approval-quorum/Cargo.toml","sha256":"df904971fca87a24194c5816738a331588fe8f164acb5d70839a35dca5f57114"},{"path":"product/sdk/auths-approval-quorum/src/lib.rs","sha256":"72e4086088a1c0912e4e1a9e980f26a5c280bf5ab16ea39c9f0f58d5096ac2ce"},{"path":"product/sdk/auths-approval-quorum/src/remote.rs","sha256":"9fa029c7c63003adf260f8aec55160c545681c60c1bace0d934e7525862aac7b"},{"path":"product/sdk/auths-approval-quorum/src/remote_tests.rs","sha256":"8a09ce17cfee82a035a74edd535f3ebeb3db0e078784740d3997b17dbdb10197"},{"path":"product/sdk/auths-approval-quorum/src/tests.rs","sha256":"0acbf12be3e497dcc6c2d8f29c9af3b1a7ed8e11b75f6399570d4f217a785060"},{"path":"product/stores/auths-stores/Cargo.toml","sha256":"a32c4b76d17b264be5d40d24497ff237e1ed87822e8bab86229b3db93106dbe3"},{"path":"product/stores/auths-stores/src/gateway_attempt.rs","sha256":"f39dc84a238db0ebc70ad1ac495da194d2c24b5e1f9905552a9b1143278725a9"},{"path":"product/stores/auths-stores/src/lib.rs","sha256":"40c280a5ff79bfc1cafcf0bb415e775e205637c526973eb94b43a30f86d4e6fe"},{"path":"product/stores/auths-stores/src/lifecycle.rs","sha256":"4033f69a8d36de9285b87ab86af363fc85f785918ec2042d0ef1bdcb64d6f73a"}],"schema":"auths.gateway-semantic-closure/1"}
\ No newline at end of file
diff --git a/product/runtime/auths-gateway/src/admin.rs b/product/runtime/auths-gateway/src/admin.rs
index bb67d56e2..4c3887c12 100644
--- a/product/runtime/auths-gateway/src/admin.rs
+++ b/product/runtime/auths-gateway/src/admin.rs
@@ -1,61 +1,131 @@
//! The operator socket's request frame: one JSON object with exactly the
-//! schema, the command, and that command's arguments. A secret, which only
-//! `rotate` and `rotate-prepare` carry, travels in a second frame and never
-//! in this one.
+//! schema, the command, and that command's arguments. Rotation secrets and
+//! recovery proof/action/approval inputs travel in a second bounded frame.
use serde::Deserialize;
+#[cfg(all(test, unix))]
+#[path = "admin_boundary_tests.rs"]
+mod boundary_tests;
+
/// Schema of an admin request frame.
-pub const ADMIN_REQUEST_SCHEMA: &str = "auths.gateway-admin-request/1";
+pub const ADMIN_REQUEST_SCHEMA: &str = "auths.gateway-admin-request/2";
/// Schema of an admin response frame.
-pub const ADMIN_RESPONSE_SCHEMA: &str = "auths.gateway-admin-response/1";
+pub const ADMIN_RESPONSE_SCHEMA: &str = "auths.gateway-admin-response/2";
+
+/// Largest operator command object; secrets and native submissions use their
+/// own second frames. Five bounded string fields need substantially less space.
+pub const MAX_ADMIN_REQUEST_BYTES: usize = 4096;
-/// One admin command and its arguments.
-#[derive(Clone, Debug, Deserialize, Eq, PartialEq)]
-#[serde(tag = "command", rename_all = "kebab-case", deny_unknown_fields)]
-pub enum AdminRequestCommand {
+macro_rules! admin_commands {
+ ($( $(#[$variant_doc:meta])* $variant:ident($name:literal) {
+ $( $(#[$field_doc:meta])* $field:ident: $field_type:ty ),* $(,)?
+ } ),* $(,)?) => {
+ /// One closed operator command and its arguments.
+ #[derive(Clone, Debug, Deserialize, Eq, PartialEq)]
+ #[serde(tag = "command", deny_unknown_fields)]
+ pub enum AdminRequestCommand {
+ $( $(#[$variant_doc])* #[serde(rename = $name)] $variant {
+ $( $(#[$field_doc])* $field: $field_type ),*
+ } ),*
+ }
+ /// Complete operator command catalog, generated with the native enum.
+ pub const ADMIN_REQUEST_COMMANDS: &[&str] = &[$($name),*];
+ const ARGUMENT_NAMES: &[&str] = &[$($(stringify!($field),)*)*];
+ };
+}
+
+admin_commands! {
// Struct variants, so an argument the command does not take is
// refused: an internally tagged unit variant ignores extra members.
/// Stop new entries in every process sharing the store.
- Disable {},
+ Disable("disable") {},
/// Re-enable a disabled connection.
- Enable {},
+ Enable("enable") {},
/// Revoke the connection permanently.
- Revoke {},
+ Revoke("revoke") {},
/// Rotate the credential; the secret follows in a second frame.
- Rotate {},
+ Rotate("rotate") {},
/// Store a successor credential without publishing it; the secret
/// follows in a second frame. Answers with its reference commitment.
- RotatePrepare {},
+ RotatePrepare("rotate-prepare") {},
/// Publish the prepared successor the commitment names.
- RotateCommit {
+ RotateCommit("rotate-commit") {
/// The reference commitment `rotate-prepare` answered with, as 64
/// lowercase hexadecimal characters.
commitment: String,
},
/// Report connection state.
- Status {},
+ Status("status") {},
/// Read process-local custody and transport measurements; grants no authority.
- ExecutionWitness {},
+ ExecutionWitness("execution-witness") {},
/// Read the qualification inputs the operator placed on this host again
/// and report the resulting state.
- QualificationReload {},
- /// Re-observe one logical operation, read-only.
- Reobserve {
- /// The logical operation ID.
+ QualificationReload("qualification-reload") {},
+ /// Export protected audit evidence to an authenticated operator.
+ AuditExport("audit-export") {
+ /// Logical operation whose protected evidence to export.
operation_id: String,
},
+ /// Acquire/refine one recovery read; a native submission follows in frame 2.
+ Recover("recover") {
+ /// Retained source mutation whose contract the read must bind.
+ source_operation: String,
+ },
+ /// Record evidence from an existing acquisition without provider/custody I/O.
+ RecoveryRecord("recovery-record") {
+ /// Retained source mutation.
+ source_operation: String,
+ /// Separately authorized read whose response is already retained.
+ acquisition_operation: String,
+ },
}
-/// Parses one `auths.gateway-admin-request/1` frame.
+struct AdminFields(serde_json::Map);
+
+impl<'de> Deserialize<'de> for AdminFields {
+ fn deserialize>(deserializer: D) -> Result {
+ struct Fields;
+ impl<'de> serde::de::Visitor<'de> for Fields {
+ type Value = AdminFields;
+ fn expecting(&self, formatter: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+ formatter.write_str("unique operator command string fields")
+ }
+ fn visit_map>(
+ self,
+ mut map: M,
+ ) -> Result {
+ let mut fields = serde_json::Map::new();
+ while let Some(key) = map.next_key::()? {
+ if fields.contains_key(&key)
+ || (key != "schema"
+ && key != "command"
+ && !ARGUMENT_NAMES.contains(&key.as_str()))
+ {
+ return Err(serde::de::Error::custom("invalid operator command field"));
+ }
+ let value = map.next_value::()?;
+ fields.insert(key, serde_json::Value::String(value));
+ }
+ Ok(AdminFields(fields))
+ }
+ }
+ deserializer.deserialize_map(Fields)
+ }
+}
+
+/// Parses one `auths.gateway-admin-request/2` frame.
///
/// # Errors
/// Returns `gateway.admin.invalid-frame` for anything but a JSON object with
-/// exactly the schema, a known command, and that command's arguments.
+/// exactly the schema, a known command, and that command's arguments. Duplicate
+/// members, non-string values and objects over `MAX_ADMIN_REQUEST_BYTES` are refused.
pub fn parse_admin_request(bytes: &[u8]) -> Result {
let invalid = "gateway.admin.invalid-frame";
- let mut request: serde_json::Map =
- serde_json::from_slice(bytes).map_err(|_| invalid)?;
+ if bytes.len() > MAX_ADMIN_REQUEST_BYTES {
+ return Err(invalid);
+ }
+ let AdminFields(mut request) = serde_json::from_slice(bytes).map_err(|_| invalid)?;
if request.remove("schema") != Some(serde_json::Value::String(ADMIN_REQUEST_SCHEMA.into())) {
return Err(invalid);
}
diff --git a/product/runtime/auths-gateway/src/admin_boundary_tests.rs b/product/runtime/auths-gateway/src/admin_boundary_tests.rs
new file mode 100644
index 000000000..b8953d0bd
--- /dev/null
+++ b/product/runtime/auths-gateway/src/admin_boundary_tests.rs
@@ -0,0 +1,147 @@
+//! Native operator frame boundary and generated interface catalog artifact.
+use super::*;
+use serde_json::{Value, json};
+use std::{collections::BTreeSet, path::PathBuf};
+
+fn request(command: &str) -> Value {
+ let mut value = json!({"schema":ADMIN_REQUEST_SCHEMA,"command":command});
+ match command {
+ "disable"
+ | "enable"
+ | "revoke"
+ | "rotate"
+ | "rotate-prepare"
+ | "status"
+ | "execution-witness"
+ | "qualification-reload" => {}
+ "rotate-commit" => value["commitment"] = json!("00".repeat(32)),
+ "audit-export" => value["operation_id"] = json!("original"),
+ "recover" => value["source_operation"] = json!("original"),
+ "recovery-record" => {
+ value["source_operation"] = json!("original");
+ value["acquisition_operation"] = json!("independent-read");
+ }
+ _ => panic!("new native command requires a boundary example"),
+ }
+ value
+}
+
+fn record(cases: &mut Vec, name: &str, bytes: Vec, accepted: bool) {
+ let result = parse_admin_request(&bytes);
+ assert_eq!(result.is_ok(), accepted, "{name}");
+ if !accepted {
+ assert_eq!(result, Err("gateway.admin.invalid-frame"));
+ }
+ cases.push(json!({"case":name,"input":String::from_utf8(bytes).expect("JSON fixture"),
+ "accepted":accepted,"refusal":if accepted {None} else {Some("gateway.admin.invalid-frame")}}));
+}
+
+#[test]
+#[allow(
+ clippy::too_many_lines,
+ reason = "one finite native parser corpus constructs its boundary artifact"
+)]
+fn admin_request_boundary_fixture_is_current() {
+ let commands: BTreeSet<_> = ADMIN_REQUEST_COMMANDS.iter().copied().collect();
+ assert_eq!(commands.len(), ADMIN_REQUEST_COMMANDS.len());
+ let mut cases = Vec::new();
+ for command in ADMIN_REQUEST_COMMANDS {
+ let value = request(command);
+ record(
+ &mut cases,
+ &format!("{command}:valid"),
+ serde_json::to_vec(&value).expect("JSON"),
+ true,
+ );
+ for (key, field) in value.as_object().expect("request") {
+ for replacement in [field.clone(), json!("altered-public-fixture")] {
+ let mut bytes = serde_json::to_vec(&value).expect("JSON");
+ bytes.pop();
+ bytes.extend_from_slice(
+ format!(
+ ",{}:{replacement}}}",
+ serde_json::to_string(key).expect("key")
+ )
+ .as_bytes(),
+ );
+ record(
+ &mut cases,
+ &format!(
+ "{command}:duplicate-{key}-{}",
+ if &replacement == field {
+ "same"
+ } else {
+ "different"
+ }
+ ),
+ bytes,
+ false,
+ );
+ }
+ }
+ let mut unknown = value.clone();
+ unknown["administrative_resolution"] = json!("not-entered-and-refunded");
+ record(
+ &mut cases,
+ &format!("{command}:unknown-argument"),
+ serde_json::to_vec(&unknown).expect("JSON"),
+ false,
+ );
+ let mut wrong_type = value;
+ wrong_type["command"] = json!({"name":command});
+ record(
+ &mut cases,
+ &format!("{command}:non-string-command"),
+ serde_json::to_vec(&wrong_type).expect("JSON"),
+ false,
+ );
+ }
+ for (name, bytes) in [
+ ("escaped-duplicate-schema", format!(r#"{{"schema":"{ADMIN_REQUEST_SCHEMA}","\u0073chema":"{ADMIN_REQUEST_SCHEMA}","command":"status"}}"#).into_bytes()),
+ ("duplicate-command-first-invalid", format!(r#"{{"schema":"{ADMIN_REQUEST_SCHEMA}","command":"unsupported","command":"status"}}"#).into_bytes()),
+ ("argument-for-other-command", format!(r#"{{"schema":"{ADMIN_REQUEST_SCHEMA}","command":"disable","commitment":"public"}}"#).into_bytes()),
+ ("nested-source", format!(r#"{{"schema":"{ADMIN_REQUEST_SCHEMA}","command":"recover","source_operation":{{"id":"original"}}}}"#).into_bytes()),
+ ("obsolete-command", format!(r#"{{"schema":"{ADMIN_REQUEST_SCHEMA}","command":"reobserve","operation_id":"original"}}"#).into_bytes()),
+ ("non-object", b"[]".to_vec()),
+ ("trailing-data", format!(r#"{{"schema":"{ADMIN_REQUEST_SCHEMA}","command":"status"}} null"#).into_bytes()),
+ ] {
+ record(&mut cases, name, bytes, false);
+ }
+ let mut boundary = serde_json::to_vec(&request("status")).expect("JSON");
+ boundary.resize(MAX_ADMIN_REQUEST_BYTES, b' ');
+ record(&mut cases, "byte-limit", boundary.clone(), true);
+ boundary.push(b' ');
+ record(&mut cases, "byte-limit-plus-one", boundary, false);
+ tokio::runtime::Runtime::new()
+ .expect("runtime")
+ .block_on(async {
+ use tokio::io::AsyncWriteExt as _;
+ let (mut sender, mut receiver) = tokio::net::UnixStream::pair().expect("socket pair");
+ sender
+ .write_all(
+ &u32::try_from(MAX_ADMIN_REQUEST_BYTES + 1)
+ .expect("length")
+ .to_be_bytes(),
+ )
+ .await
+ .expect("advertise length without a body");
+ let clock = crate::app::SessionClock::start(crate::app::APP_SESSION_LIMITS);
+ let result = tokio::time::timeout(
+ std::time::Duration::from_secs(2),
+ clock.read_frame_limited(&mut receiver, MAX_ADMIN_REQUEST_BYTES),
+ )
+ .await
+ .expect("must refuse length before waiting for body bytes");
+ assert_eq!(result, Err("gateway.ipc.invalid-size"));
+ });
+ let artifact = serde_json::to_vec_pretty(&json!({"schema":"auths.admin-request-boundary/1",
+ "scope":"native closed decoder and enum-generated interface catalog; decoding grants no operator, provider or recovery authority",
+ "maximum_command_bytes":MAX_ADMIN_REQUEST_BYTES,"oversized_advertised_length_refused_before_body":true,"application_schemas":crate::app::APPLICATION_REQUEST_SCHEMAS,
+ "operator_commands":ADMIN_REQUEST_COMMANDS,"cases":cases})).expect("artifact");
+ let path = PathBuf::from(env!("CARGO_MANIFEST_DIR"))
+ .join("../../../bindings/fixtures/gateway/admin-request-boundary.json");
+ if std::env::var_os("AUTHS_UPDATE_FIXTURES").is_some() {
+ std::fs::write(&path, &artifact).expect("fixture");
+ }
+ assert_eq!(std::fs::read(path).expect("artifact"), artifact);
+}
diff --git a/product/runtime/auths-gateway/src/app.rs b/product/runtime/auths-gateway/src/app.rs
index 967e1c4fa..a05f5a9f2 100644
--- a/product/runtime/auths-gateway/src/app.rs
+++ b/product/runtime/auths-gateway/src/app.rs
@@ -1,8 +1,8 @@
-//! The application-socket protocol: one length-prefixed JSON frame in, one
-//! out. A frame is exactly one of two closed schemas, a proof/action
-//! submission or an observation request; anything else is refused without
-//! reaching the application.
+//! Closed application protocol: one control frame and native result, except
+//! disclosure, which explicitly switches to bounded recipient authentication,
+//! content and ACK frames. Native owners decide every effect and history change.
+use crate::effect::{ExactEffectApproval, MAX_EFFECT_APPROVAL_BYTES};
use crate::{GatewayEngine, GatewayObserveRequest, GatewayObserveResult, GatewaySubmitResult};
use base64ct::{Base64UrlUnpadded, Encoding as _};
use serde::{Deserialize, Serialize};
@@ -12,11 +12,25 @@ use std::time::Duration;
use tokio::io::{AsyncReadExt as _, AsyncWriteExt as _};
use tokio::net::UnixStream;
use tokio::time::{Instant, sleep_until, timeout_at};
+use zeroize::{Zeroize as _, Zeroizing};
+
+mod content;
+#[cfg(test)]
+#[path = "app/content_tests.rs"]
+mod content_tests;
+pub use content::{
+ APP_CONTENT_HISTORY_SCHEMA, APP_CONTENT_SCHEMA, APP_DISCLOSURE_ACK_SCHEMA,
+ APP_DISCLOSURE_SCHEMA, AppContent, AppContentHistory, AppContentResult, AppDisclosure,
+ AppDisclosureAcknowledgement, AppDisclosureTarget,
+};
+pub use content::{AppDisclosureReception, receive_content_disclosure};
/// Schema of a proof/action submission frame.
-pub const APP_REQUEST_SCHEMA: &str = "auths.gateway-submit/1";
+pub const APP_REQUEST_SCHEMA: &str = "auths.gateway-submit/2";
/// Schema of an observation request frame.
-pub const APP_OBSERVE_SCHEMA: &str = "auths.gateway-observe/2";
+pub const APP_OBSERVE_SCHEMA: &str = "auths.gateway-observe/4";
+/// Independently approved recovery acquisition and evidence refinement.
+pub const APP_RECOVERY_SCHEMA: &str = "auths.gateway-recover/1";
/// Largest frame either side accepts.
pub const MAX_FRAME_BYTES: usize = 8 * 1024 * 1024;
@@ -73,10 +87,26 @@ impl SessionClock {
/// Returns [`read_frame`]'s codes, or `gateway.ipc.read-timeout` when a
/// deadline passes first.
pub async fn read_frame(&self, stream: &mut UnixStream) -> Result, &'static str> {
+ self.read_frame_limited(stream, MAX_FRAME_BYTES).await
+ }
+
+ /// Reads a frame under a tighter protocol limit, clamped to the global
+ /// frame bound. An oversized advertised length is refused before allocation.
+ ///
+ /// # Errors
+ /// Returns the bounded frame/read/deadline failures of [`Self::read_frame`].
+ pub async fn read_frame_limited(
+ &self,
+ stream: &mut UnixStream,
+ maximum: usize,
+ ) -> Result, &'static str> {
let until = self.deadline.min(Instant::now() + self.limits.frame_read);
- timeout_at(until, read_frame(stream))
- .await
- .unwrap_or(Err("gateway.ipc.read-timeout"))
+ timeout_at(
+ until,
+ read_frame_limited(stream, maximum.min(MAX_FRAME_BYTES)),
+ )
+ .await
+ .unwrap_or(Err("gateway.ipc.read-timeout"))
}
/// Writes one frame within the response-write limit and the session
@@ -133,6 +163,58 @@ pub struct AppSubmission {
pub proof_b64: String,
/// Canonical action bytes, unpadded base64url.
pub action_b64: String,
+ /// Native canonical exact-effect approval, unpadded base64url.
+ pub approval_b64: String,
+}
+
+/// Bounded decoded transport input. This is data only, not verified authority.
+pub struct DecodedSubmission {
+ /// Canonical proof bytes for native verification.
+ pub proof: Zeroizing>,
+ /// Canonical action bytes for native verification.
+ pub action: Zeroizing>,
+ /// Parsed exact-effect approval; native entry still validates it.
+ pub approval: ExactEffectApproval,
+}
+
+impl AppSubmission {
+ /// Decodes one bounded proof/action/approval frame without authorizing it.
+ ///
+ /// # Errors
+ /// Refuses wrong schema, oversized fields and invalid native encoding.
+ pub fn decode(&self) -> Result {
+ let invalid = "gateway.submit.invalid-encoding";
+ if self.schema != APP_REQUEST_SCHEMA
+ || self.approval_b64.len() > (MAX_EFFECT_APPROVAL_BYTES * 4).div_ceil(3)
+ {
+ return Err("gateway.submit.invalid-frame");
+ }
+ if self.proof_b64.len() > (4 * 1024 * 1024 * 4_usize).div_ceil(3)
+ || self.action_b64.len() > (64 * 1024 * 4_usize).div_ceil(3)
+ {
+ return Err(invalid);
+ }
+ let proof =
+ Zeroizing::new(Base64UrlUnpadded::decode_vec(&self.proof_b64).map_err(|_| invalid)?);
+ let action =
+ Zeroizing::new(Base64UrlUnpadded::decode_vec(&self.action_b64).map_err(|_| invalid)?);
+ let approval =
+ Zeroizing::new(Base64UrlUnpadded::decode_vec(&self.approval_b64).map_err(|_| invalid)?);
+ let approval = ExactEffectApproval::from_private_bytes(&approval).map_err(|_| invalid)?;
+ Ok(DecodedSubmission {
+ proof,
+ action,
+ approval,
+ })
+ }
+}
+
+impl Drop for AppSubmission {
+ fn drop(&mut self) {
+ self.proof_b64.zeroize();
+ self.action_b64.zeroize();
+ self.approval_b64.zeroize();
+ }
}
/// An observation request frame.
@@ -145,16 +227,76 @@ pub struct AppObservation {
pub request: GatewayObserveRequest,
}
-/// Every application frame is exactly one of the two closed schemas.
-#[derive(Deserialize)]
-#[serde(untagged)]
-enum AppFrame {
- Submit(AppSubmission),
- Observe(AppObservation),
+/// One separately authorized provider recovery read. The nested submission
+/// uses the same bounded canonical decoder as mutation and operator recovery.
+#[derive(Deserialize, Serialize)]
+#[serde(deny_unknown_fields)]
+pub struct AppRecovery {
+ /// Must equal `APP_RECOVERY_SCHEMA`.
+ pub schema: String,
+ /// Original retained mutation, never a provider path.
+ pub source_operation: String,
+ /// Native read authority and exact approval.
+ pub submission: AppSubmission,
+}
+
+impl AppRecovery {
+ /// Decodes transport data only; native compilation still establishes authority.
+ ///
+ /// # Errors
+ /// Refuses obsolete schema, invalid operation and invalid native envelope.
+ pub fn decode(&self) -> Result {
+ if self.schema != APP_RECOVERY_SCHEMA
+ || crate::LogicalOperationId::parse(&self.source_operation).is_err()
+ {
+ return Err("gateway.submit.invalid-frame");
+ }
+ self.submission.decode()
+ }
+}
+
+/// Redacted native facts or a native refusal. No response body or signed
+/// provider value is released through this channel.
+#[derive(Serialize)]
+#[serde(tag = "outcome", rename_all = "kebab-case")]
+pub enum AppRecoveryResult {
+ /// Distinct retained mutation/acquisition facts, not a claim of success.
+ RecoveryReport {
+ /// Native redacted source/acquisition projection.
+ report: crate::GatewayRecoveryReport,
+ },
+ /// Native refusal/indeterminate outcome preserves its stable code.
+ Refused {
+ /// Original native stage and refusal code.
+ result: GatewaySubmitResult,
+ },
+}
+
+// The enum and operator-visible catalog share this one declaration, so an
+// application variant cannot be added without entering the interface inventory.
+macro_rules! application_frames {
+ ($($variant:ident($frame:ty) => $schema:expr),* $(,)?) => {
+ /// Every application frame selects one closed native operation.
+ #[derive(Deserialize)]
+ #[serde(untagged)]
+ enum AppFrame { $($variant($frame)),* }
+ /// Complete application control-frame schemas, from the native dispatcher.
+ pub const APPLICATION_REQUEST_SCHEMAS: &[&str] = &[$($schema),*];
+ };
+}
+application_frames! {
+ Submit(AppSubmission) => APP_REQUEST_SCHEMA,
+ Observe(AppObservation) => APP_OBSERVE_SCHEMA,
+ Recover(AppRecovery) => APP_RECOVERY_SCHEMA,
+ Content(AppContent) => APP_CONTENT_SCHEMA,
+ Disclose(AppDisclosure) => APP_DISCLOSURE_SCHEMA,
+ ContentHistory(AppContentHistory) => APP_CONTENT_HISTORY_SCHEMA,
+ DisclosureAck(AppDisclosureAcknowledgement) => APP_DISCLOSURE_ACK_SCHEMA,
}
/// Which closed frame `bytes` parse as, schema included: `submit`,
-/// `observe`, or `None`.
+/// `observe`, `recover`, `content`, `disclose`, `content-history`,
+/// `disclosure-ack`, or `None`.
#[cfg(feature = "fuzzing")]
pub(crate) fn frame_kind(bytes: &[u8]) -> Option<&'static str> {
match serde_json::from_slice::(bytes).ok()? {
@@ -164,6 +306,11 @@ pub(crate) fn frame_kind(bytes: &[u8]) -> Option<&'static str> {
AppFrame::Observe(observation) => {
(observation.schema == APP_OBSERVE_SCHEMA).then_some("observe")
}
+ AppFrame::Recover(recovery) => recovery.decode().ok().map(|_| "recover"),
+ AppFrame::Content(request) => request.decode().ok().map(|_| "content"),
+ AppFrame::Disclose(request) => request.decode().ok().map(|_| "disclose"),
+ AppFrame::ContentHistory(request) => request.validate().ok().map(|()| "content-history"),
+ AppFrame::DisclosureAck(request) => request.decode().ok().map(|_| "disclosure-ack"),
}
}
@@ -171,9 +318,60 @@ pub(crate) fn frame_kind(bytes: &[u8]) -> Option<&'static str> {
/// harness that follows the same dispatch against a counting provider.
pub trait GatewayApplication {
/// Verifies and attempts one exact action.
- fn submit(&self, proof: &[u8], action: &[u8]) -> impl Future