Repository navigation
ci: only reviews run under the review contract count against the budget (ADR-1129) - #965
Conversation
…et (ADR-1129) A pull request reviewed by the workflow that preceded the cap and the verdict line had its budget used by reviews that posted no verdict: the review was skipped on every new head and the merge was held for ever (ai-architect-mcp-spec #182, #184, #181; ai-architect-mcp-codebase #484, #481, #478, #476 on 2026-11-10). A run now counts only if it has the Claude review budget job. Co-Authored-By: Claude Fable 5.2 <[email protected]> Claude-Session: https://claude.ai/code/session_018kyZYsWnKSjpX4oNV19PCF Signed-off-by: Claude <[email protected]>
Codex Review Summary: ✅ CompletedThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
|
Claude finished @cdeust's task in 1m 41s —— View job Review of PR #965 (review 1 of 3)
No Blockers. I posted three Notes that need no action before merge. Verdict: approve. I could not run pytest, ruff or |
Claude review 1 of 3: PR #965I checked the diff with five reviews: code quality, security, tests, docs and performance. None of them found a Blocker.
Notes, no action before merge
Verdict: approve |
|
CI autopilot : #965 en retard de 4 commits sur Le rebase sur Generated by Claude Code |
Conflict: wiki/manifest.json, the ADR-1128 and ADR-1129 entries are both kept. The ADR-1130 entry is added: its decision file came with #975 without its manifest line, which made check_project_wiki.py fail on main. Co-Authored-By: Claude Fable 5.2 <[email protected]> Claude-Session: https://claude.ai/code/session_018kyZYsWnKSjpX4oNV19PCF Signed-off-by: Claude <[email protected]>
|
Claude finished @cdeust's task in 1m 52s —— View job Review 2 of 3 of PR #965
The merge kept the branch's change intact. In I did not run pytest or ruff in this job. The review checked the code by reading it. Verdict: approve |
Claude review 2 of 3: PR #965There were no Blockers in review 1. Since b528fc6 the only new commit is the merge
Notes, no action before merge
View job run |
Summary
The review budget counts only reviews run under the review contract (ADR-1129, amends ADR-1119).
A pull request opened before a repository received the contract (the cap, then the
Verdict:line the merge waits for) was reviewed by the earlier workflow, sometimes more than four times, with no verdict. Its budget was then already used: the review was skipped on every new head, no verdict could exist, and the resolver held the merge for ever.claude-rereviewgrants a fourth review, never a fifth, so it did not help.Measured on 2026-11-10: ai-architect-mcp-spec #182, #184, #181 and ai-architect-mcp-codebase #484, #481, #478, #476. #182 had four earlier reviews; relabelled and reopened, its run 42769093930 still ended "3 of 3 reviews used".
ran_the_actionnow selects thePR Reviewstep only in a run that has theClaude review budgetjob, which the earlier workflow did not have. The script still reads run metadata only.Measured
Applied to the real runs of ai-architect-mcp-spec #182 with
jq: the four earlier reviews went from counted to not counted.tests_py/scripts/test_ci_review_budget.py: 22 passed.ruff check,ruff format --check,scripts/check_craftsmanship.py: pass.Generated by Claude Code