werkrevisor 0.41.4: the first PR through the Codex gate #105
No Assignees
Notifications
Due Date
No due date set.
Dependencies
No dependencies set.
Reference: mi/werkrevisor#105
Reference in New Issue
Block a user
See the PR document under docs/prs/.
🤖 Generated with Claude Code
🟠 werkrevisor ORANGE for
534f790, superseded by the result for6ccabc3🟠 werkrevisor: ORANGE / FINDINGS
Change kind:
mixed.Reasons
Findings
Off-by-one in quota at
werkrevisor/codexrunner.py:215, severity unknown, counted as medium, ORANGE, as a line commentDocumentation contradicts code change at
docs/prs/2026-10-14-PR#105-codex-gate-test.md:16-18, severity unknown, counted as medium, hint, as a line commentClassifier
5 of 5 units passed all questions (1 PR document(s)).
Usage: model `openrouter/minimax/minimax-m3`, LLM done, tokens counted before the call 11167, calls 1, tokens in 10780 / out 13614 (reasoning included in out), duration 129.5 s, cost 0.0195 USD (reported), classifier `jev-1.13.0` calls 5, tokens in 3325, cost 0.0001 USD, finding judge `jev-1.13.0` calls 2, tokens in 2580, cost 0.0001 USD, change summary `openrouter/minimax/minimax-m3` via Minimax calls 1, tokens in 840 / out 1292, cost 0.0018 USD (reported), PR document check `jev-1.13.0` calls 5, tokens in 5180, cost 0.0002 USD, build file judge `jev-1.13.0` calls 3, tokens in 1602, cost 0.0001 USD, phases: path_judge 0.4 s, classifier 0.4 s, change_summary 17.2 s, doc_check 0.4 s, context 0.2 s, model_review 110.2 s, finding_judge 0.2 s, config: defaults + host + repository (.werkrevisor.toml at main2af34c7).Reviewed head `534f790a2e170550717b5902cb9db61718909127` against base `2af34c752131829cf1feba056df49b8b0f22ab3f`, werkrevisor 0.41.3.
werkrevisor: ORANGE / FINDINGS for
534f790a2e170550717b5902cb9db61718909127.The result comment of werkrevisor holds the details.
@@ -0,0 +13,4 @@Since PR#104 the branch protection of werkrevisor requires the status `ai-review/codex`.No PR has passed the gate yet; it is not known whether it holds and releases the merge as intended.## Non-GoalsDocumentation contradicts code change, severity unknown, counted as medium (werkrevisor)
hint; Jev: real 0.84, severity 2.3 of 3
The PR document states "No change of the runner or of werkrevisor" under Non-Goals, but
werkrevisor/codexrunner.pyis modified with a behavior change (self.reviews_left <= 0→self.reviews_left < 0). The PR document should acknowledge and explain this change, since the squash merge does include a modification to the runner.Fixed in
6ccabc3: the runner change is removed again, so the squash merge does not change the runner; the document now names the defect and the new test.@@ -213,3 +213,3 @@if comment is None:# Before werkrevisor, a PR may still turn out RED or WIP; the quota is for PRs that are ready.if not has_werkrevisor_result(comments, head, login) or self.reviews_left <= 0:if not has_werkrevisor_result(comments, head, login) or self.reviews_left < 0:Off-by-one in quota, severity unknown, counted as medium (werkrevisor)
ORANGE; Jev: real 0.79, severity 2.9 of 3
The change from
self.reviews_left <= 0toself.reviews_left < 0allows one more review per run than themax_reviews_per_runconfig specifies. The check happens before the decrement, so withmax_reviews_per_run = 3, the new condition0 < 0is False and a fourth review runs, leavingreviews_left = -1before the next skip. The old condition correctly limited the run to 3 reviews. The PR document does not explain why an extra review is now permitted.Fixed in
6ccabc3: the defect was on purpose, to test the Codex gate;<= 0is back, and a new test checksmax_reviews_per_run.Codex review of
534f790a2e170550717b5902cb9db61718909127@@ -213,3 +213,3 @@if comment is None:# Before werkrevisor, a PR may still turn out RED or WIP; the quota is for PRs that are ready.if not has_werkrevisor_result(comments, head, login) or self.reviews_left <= 0:if not has_werkrevisor_result(comments, head, login) or self.reviews_left < 0:[low] Review quota allows one extra review per run (Codex review)
Changing
<= 0to< 0permits another review when the quota is exhausted. Verified with the existing fake client: a configured limit of 3 allows 4 reviews, and a limit of 0 still allows 1. Restore<= 0to enforcemax_reviews_per_run.Fix it, or reply why it does not hold, and resolve this comment.
Fixed in
6ccabc3:<= 0is back; the new testtest_a_run_reviews_at_most_max_reviews_per_runfails with< 0for the limits 0 and 1.⚠️ werkrevisor ERROR for
?, superseded by the result for6ccabc3Codex review: not critical
The change affects the per-run review quota, version numbers, and documentation. It does not weaken merge approval requirements or introduce a severe security or irreversible-data risk.
Findings
werkrevisor/codexrunner.py:215, as a line commentThe merge waits for the status
Model `gpt-6-astra`, effort high, tokens in 151300 (cached 120704) / out 1132, 55.2 s, 5 commands; reviewed head `534f790a2e170550717b5902cb9db61718909127`.ai-review/codex: every line comment resolved, and the approval of a human for a critical PR or for a finding that is only here.🟢 werkrevisor: GREEN / PASS
Change kind:
mixed.Reasons
Classifier
6 of 6 units passed all questions (1 PR document(s)).
Usage: model `openrouter/minimax/minimax-m3`, LLM done, tokens counted before the call 10712, calls 1, tokens in 10355 / out 6014 (reasoning included in out), duration 51.6 s, cost 0.0103 USD (reported), classifier `jev-1.13.0` calls 6, tokens in 4038, cost 0.0002 USD, change summary `openrouter/minimax/minimax-m3` via Minimax calls 1, tokens in 1048 / out 808, cost 0.0012 USD (reported), PR document check `jev-1.13.0` calls 6, tokens in 6120, cost 0.0003 USD, build file judge `jev-1.13.0` calls 3, tokens in 1773, cost 0.0001 USD, phases: path_judge 0.2 s, classifier 0.4 s, change_summary 6.9 s, doc_check 0.5 s, context 0.2 s, model_review 43.1 s, config: defaults + host + repository (.werkrevisor.toml at main1 answer ignored because of a confidence below [classifier] min_confidence.
2af34c7).Reviewed head `6ccabc35a739c325c1835d87ac195c219e3fa02b` against base `2af34c752131829cf1feba056df49b8b0f22ab3f`, werkrevisor 0.41.3.
werkrevisor: GREEN / PASS for
6ccabc35a739c325c1835d87ac195c219e3fa02b.The result comment of werkrevisor holds the details.
Codex review: not critical
The change adds a regression test, updates documentation, and bumps versions; it does not change runtime behavior or weaken CI gates or tests. Test execution was blocked by unavailable Docker access.
No findings.
The merge waits for the status
Model `gpt-6-astra`, effort high, tokens in 94342 (cached 63744) / out 647, 32.2 s, 4 commands; reviewed head `6ccabc35a739c325c1835d87ac195c219e3fa02b`.ai-review/codex: every line comment resolved, and the approval of a human for a critical PR or for a finding that is only here.