ci(permit): rubric denies on P0/P1 only; add severity discipline - #71
Draft
peycheff-com wants to merge 1 commit into
Draft
ci(permit): rubric denies on P0/P1 only; add severity discipline#71peycheff-com wants to merge 1 commit into
peycheff-com wants to merge 1 commit into
Conversation
Pairs with helm-ai-kernel#701, which stops the reducer counting P2 as blocking. The reviewer rubric has to move in the same authority generation: until it does, a reviewer told to DENY on any P2 still denies through REVIEW_DENIED and the reducer change accomplishes nothing. Also states severity discipline explicitly. Observed across 36 permits: one lane raised 39 blocking findings to the other's 7, and several were requests for work outside a phase-1 library PR's stated scope. A finding that asks for work the change does not claim to do is at most P2. Not merge-ready: gate policy is promoted through the authority-generation path, not an ordinary PR.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Draft on purpose. This is the reviewer half of a gate-policy change; the reducer half is helm-ai-kernel#701. Both are generation-3 candidates and neither should be merged under the generation they amend — see HELM-392.
Why
scripts/autonomous_release_permit.pytells both lanes "ALLOW only when you found no P0, P1, or P2 issue". Paired with a conjunctive quorum, that makes any substantial change unmergeable by iteration: each round of fixes closes the reported findings and the next review returns a fresh crop.Measured across 36 evaluated permits and their review artifacts:
anthropic/claude-fable-5openai/gpt-5.6-sol19 of 24 denials were single-handed from one lane. Three kernel PRs carry
permit round 9in their history; a tenth round on #671/#672/#673 closed every reported finding and drew three new P1s with zero repeats.What changes
Verdict rule. ALLOW when no P0 and no P1 exists. P2 and P3 are reported and still return ALLOW. DENY stays for P0/P1 or insufficient evidence on a safety-critical change.
Severity discipline, stated explicitly. Assign the severity the defect carries — do not inflate a P2 to force attention, do not deflate a real P1 because the change is large. And: a finding that asks for work outside the change's stated scope is at most P2. Scope is set by the change, not by what a complete implementation would eventually need; a phase-1 library change is not deficient for lacking an integration it does not claim. Several denials in the measured set were exactly that.
Sequencing
Without the reducer half, a reviewer told to DENY on P2 still denies through
REVIEW_DENIEDand this change accomplishes nothing. Without this half, the reducer stops counting P2 as blocking while the reviewer keeps voting DENY on it. They move in the same authority generation or not at all.python3 -m pytest tests/test_autonomous_release_permit.py— 27 passed.