review: verdicts record a disposition per lifted claim - #231
Merged
Conversation
The change the 2026-08-18 falsified-by efficacy review recommended (agent_failure_modes.md item 6 Outcome): across 22 ship gates the ledger could not distinguish a healthy adversarial claim pass from a skipped one — both wrote a bare 'review CLEAN'. Now, when the ReviewSurface lifts any 'claims to falsify', the verdict carries one disposition line per claim (basis-cited / idle / FINDING unverified-claim), written by the reviewer at verdict time; a bare CLEAN over a non-empty claims surface is malformed evidence. An empty surface requires nothing, so the 74-95% of ships that lift no claims gain no busywork. Surfaces: faculty AGENTS.md step 2a + verdict mapping; _review.py human-emit epilogue (printed only when claims lifted); AUTONOMY.md autonomous-ship-gate review leg; ship_library reference.md gate evidence line. No trigger-vocabulary change (measured as neither empty nor saturated). 2 new pinning tests; suite 351 passed. Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01WH4NizvBK2jki2Uh5TMABh
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The change the 2026-08-18 falsified-by efficacy review recommended (
agent_failure_modes.mditem 6 Outcome, merged in #230): across 22 ship gates the ledger could not distinguish a healthy adversarial claim pass from a skipped one — both wrote a bare "review CLEAN". Executes the PyAutoMind promptreview_claim_dispositions.md; the Mind-side record ships in the paired PyAutoMind PR on the same branch name.The rule: when the ReviewSurface lifts any
claims to falsify, the reviewing agent's verdict carries one disposition line per claim —— written by the reviewer at verdict time, never by the author. A bare CLEAN over a non-empty claims surface is malformed evidence, not CLEAN. An empty surface requires nothing, so the 74–95% of ships that lift no claims gain no busywork. No trigger-vocabulary change (measured as neither empty nor saturated).
Surfaces:
agents/faculties/review/AGENTS.md— step 2a disposition format + step 3 malformed-evidence ruleagents/faculties/review/_review.py— human-emit epilogue demands dispositions, guarded to print only when claims were liftedAUTONOMY.md— autonomous-ship-gate review leg carries the requirementskills/ship_library/reference.md—--autovalidation-checklist gate linedocs/agent_failure_modes.md— item 6 Outcome updated from "filed" to "implemented"tests/test_review_claims.py— 2 new pinning tests; suite 351 passedDogfood: the faculty was run on this branch itself — surface produced, zero claims lifted from the commit message, no disposition demand printed on the empty surface (the guard's no-busywork path), reviewer verdict CLEAN with no dispositions owed.
🤖 Generated with Claude Code
https://claude.ai/code/session_01WH4NizvBK2jki2Uh5TMABh
Generated by Claude Code