Skip to content
Merged
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
79 changes: 59 additions & 20 deletions docs/migrations/scanner-cutover-ponsbot-panicbot.adoc
Original file line number Diff line number Diff line change
Expand Up @@ -2,10 +2,10 @@
= Scanner cutover: pons-asinorum and panic-attack → ponsbot and panicbot
:toc: left
:toclevels: 2
:revdate: 2026-11-13
:revdate: 2026-11-14

Status: **DESIGN, with the owner's rulings of 2026-11-12 and 2026-11-13
recorded (§8). Phase 0 (the adapters) is the only phase started.** Nothing
Status: **DESIGN, with the owner's rulings of 2026-11-12, 2026-11-13 and
2026-11-14 recorded (§8). Phase 0 (the adapters) has landed; phase 1 is next.** Nothing
has been cut over: no
workflow, enrolment manifest, grid row or direct lane has changed. Each later
phase lands as its own pull request.
Expand Down Expand Up @@ -132,8 +132,9 @@ stay as external engines and repositories, and only the direct CI lanes retire.
a row's gating from the enrolment's per-reporter `gating` boolean (standards
`scripts/lib/verdict-grid.jq`), not from the envelope: enrolments set
`"gating": true` for ponsbot and leave panicbot without it.
* Hypatia's estate rescan and bridge gate are decided separately (§8 D4); they
are not PR lanes.
* Hypatia's estate rescan and bridge gate are not PR lanes. Under §8.3 D4
(superseded 2026-11-14) they move behind panicbot's optional `bridge` mode
in phase 5.

== 4. ponsbot (new adapter)

Expand Down Expand Up @@ -192,7 +193,7 @@ A thin crate in `bots/ponsbot/`, the same shape as panicbot:
scope filter alone handles them.
* Fix or retire `panicbot-sweep.yml`'s install line (see §2.2).

== 6. Phases (each is its own PR; only phase 0 is started)
== 6. Phases (each is its own PR; phase 0 has landed)

[cols="1,3,2,2"]
|===
Expand Down Expand Up @@ -224,9 +225,9 @@ A thin crate in `bots/ponsbot/`, the same shape as panicbot:
| No `panic-attack` invocation remains outside `bots/panicbot/`

| 5 (Hypatia)
| Per the D4 ruling (2026-11-13): record estate-rescan and the bridge gate as sanctioned direct uses, reading panicbot's pin
| hypatia
| Ruling recorded; the verisim-data shape is unchanged either way
| Per the superseding D4 ruling (2026-11-14): add panicbot's optional `bridge` mode (gitbot-fleet), then move estate-rescan and the bridge gate behind it, reading panicbot's pin
| gitbot-fleet, hypatia
| Bridge mode tested; the verisim-data shape is unchanged; standard scanning still works without the bridge mode
|===

== 7. Risks and rollback
Expand Down Expand Up @@ -271,7 +272,7 @@ G2. **Decided 2026-11-12.** panic-attack does not gate: panicbot is advisory.

=== 8.2 Decided (owner rulings, 2026-11-13)

D4. **Decided 2026-11-13.** Hypatia's estate rescan (verisim-data feed) and
D4. **Decided 2026-11-13; superseded 2026-11-14 (see §8.3).** Hypatia's estate rescan (verisim-data feed) and
the bridge gate (`panic-attack bridge triage`) are **sanctioned direct
uses**, consistent with D0: bridge triage is the CVE-mitigation
lifecycle, not lane scanning. Both stay outside panicbot, and both read
Expand All @@ -291,13 +292,51 @@ D7. **Decided 2026-11-13.** Converge on the Hypatia lane's pins: panic-attack
`410a8f82`, pons `bca1ea28`, the values phase 0 already records in
`bots/panicbot/canon/pin.toml` `[build]` and `bots/ponsbot/canon/pin.toml`.
`fleet-report-back-reusable.yml` (`b689a74d`) and deep-recon (`b5530834`,
`51b8920f`) move to them in phase 1. Later bumps change those files only.

=== 8.3 Open

D2. Shadow-period length for phase 1 (number of PRs or days) and the
acceptable difference between the direct and adapter rows.

D3. Does `hooks/pre-push-gate.sh`'s default (`panic-attack hypatia`) become
`panicbot ponsbot hypatia`, or does pons stay out of the local pre-push
gate because of its build cost?
`51b8920f`) move to them in phase 1. (The fleet reusable already moved to
`410a8f82` in #674.) Later bumps change those files only.

=== 8.3 Decided (owner rulings, 2026-11-14)

D2. **Decided 2026-11-14.** The phase 1 shadow period runs **14 days**. A

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2 Badge Align phase 1 with the 14-day shadow period

This ruling makes elapsed time the phase 1 boundary, but the phase table still says to shadow for a fixed number of PRs and evaluate an agreed sample. When PR volume is unusually high or low, following that table can advance phase 2 before 14 days or prolong the shadow past the decided period. Update the phase 1 change and exit criterion to require the full 14 days.

Useful? React with 👍 / 👎.

difference in evidence text between the direct and adapter rows is
acceptable; a difference in verdict (status or gating outcome) is not, and
blocks phase 2.

D3. **Decided 2026-11-14.** `hooks/pre-push-gate.sh`'s default becomes
`panicbot ponsbot hypatia` (the full set), accepting pons' build cost
(phase 4).

D4 (superseded). **Superseded 2026-11-14** (D4 of 2026-11-13 stood for one
day). The owner's reasoning: panicbot is the harness that lets panic-attack
Comment on lines +309 to +310

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2 Badge Mark the replacement D4 as superseding

This paragraph contains the operative 2026-11-14 decision, yet its heading labels that decision itself “superseded”; the old 2026-11-13 D4 is already correctly marked superseded in §8.2. A phase 5 implementer can therefore read the bridge plan as withdrawn. Label this entry as revised/superseding and reserve “superseded” for the prior D4.

Useful? React with 👍 / 👎.

work as a bot inside the Hypatia system. It must still run as a plain
scanner without any bridge capability, but when it runs as a bot it should
be self-contained as far as it can be. Consequence: panicbot gains an
*optional* `bridge` mode (a thin subprocess call to `panic-attack bridge
triage`, same pin, same envelope discipline as `scan`). Hypatia's estate
rescan and bridge gate move behind panicbot in phase 5 instead of staying
Comment on lines +313 to +316

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2 Badge Define an adapter path for the estate rescan

The optional mode described here only invokes panic-attack bridge triage, so it cannot replace the estate rescan, which §2.2 identifies as an assail run feeding verisim-data. The existing panicbot scan path emits translated fleet findings rather than the raw report consumed by that pipeline, while phase 5 requires the verisim-data shape to remain unchanged. Specify a separate raw estate-rescan mode or leave that use direct instead of moving both workflows behind the bridge mode.

Useful? React with 👍 / 👎.

direct uses. The standard scan path must not depend on the bridge mode.
*Interpretation, to confirm in review:* the owner's answer did not name the
bridge mode; this reading is mine.

R14. **Decided 2026-11-13 (relayed).** finishingbot is retired or re-pointed
at the deed (D312). The design of that re-point is *not yet written*: it
needs the deed's current shape and a decision whether finishingbot's rows
are replaced by a deed-conformance row. Tracked as a phase 6 PR once phase
2 settles the grid shape; nothing is cut over for it here.

=== 8.4 Authority and ownership (relayed owner rulings, 2026-11-13)

* Pin authority: `bots/*/canon/pin.toml` in gitbot-fleet. Standards reusable
pins are owned by `propagate-sha-bump.sh` with the closure audit, and
Dependabot handles third-party actions only. The pin change in standards is
Comment on lines +329 to +331

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2 Badge Assign scanner-pin updates to a capable mechanism

For phase 1 and subsequent re-pins, this assigns the scanner pins copied into standards workflows to propagate-sha-bump.sh, but that script only handles reusable_workflow_sha_bump_needs_propagation findings and dispatches updates to consumers whose uses: reference contains an old reusable-workflow SHA; it never edits standards' internal panic-ref, pons-ref, or scanner revision values. Following this authority statement therefore leaves those copies stale relative to bots/*/canon/pin.toml; name or add the mechanism that actually synchronizes the scanner pins.

Useful? React with 👍 / 👎.

Tsunami-breaker's (standards#1509).
* The cutover thread owns the cutover edits to `hypatia-scan-reusable.yml` and
`deep-recon-reusable.yml`. The thread that opened phase 0 (#670) is archived,
so the cutover thread is the one that wrote this design.
* One auto-merge actuator (the estate sweep with `merge_ready` gates); the
Inbox Steward auto-merge job is dropped and its monitoring kept. CodeRabbit
and Codex are advisory.

=== 8.5 Open

None for phases 0 to 4. The bridge mode of 8.3 D4 is open to review.
Comment on lines +340 to +342

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2 Badge Keep the unresolved phase 2 and 3 choices open

The claim that phases 0–4 have no open questions contradicts two unresolved choices earlier in this document: §5 still says to decide how counts_excluded is preserved before the direct row is removed, and §4 says phase 3 must choose between native adapter SARIF and a conversion step. Treating these as closed can let phases 2 or 3 proceed without preserving their existing output contracts, so list them here or settle them explicitly.

Useful? React with 👍 / 👎.

Loading