Summary
Stage 1-pre's duplicate-close exit is the first terminal action in triage that leaves no PR review — and the Notify silent triage re-run step in .github/workflows/qwen-triage.yml is built on the review list alone (it checks HAS_REVIEW/STANDING and nothing else). Driving the extracted step against a stubbed gh in the close-exit state, it:
- POSTs the
stage=rerun-summary comment ("Triage re-run completed without a new review … If this re-run was meant to review or approve, it did not") on the PR the same run just closed with an explanatory duplicate comment — self-contradictory within one timeline;
- emits
::warning title=Triage re-run left no bot review, dispatching a human to a correctly-handled run.
The base skill has other review-less terminal shapes (the deferred approve-on-green path, Stage 1b's re-run escalation), but none actively closes the PR and then announces on it that the re-run did nothing.
Proposed scope
Treat a bot-initiated close during the run as a terminal action before warning: fetch gh pr view "$NUMBER" --repo "$REPO" --json state,closedAt and exit quietly (or post an adapted note) when the PR closed at/after the trigger; alternatively detect a stage=1-pre bot comment created after the trigger. The suite already pins this step ("notifies the author when a manual triage re-run posts no review") — an assertion covering the close-exit exemption should go red if removed.
Source
Review thread PRRT_kwDOPB-92c6c3VLs on #10292 (Stage 1-pre spec review). Deferred from that PR because it touches qwen-triage.yml, outside that PR's diff.
Summary
Stage 1-pre's duplicate-close exit is the first terminal action in triage that leaves no PR review — and the
Notify silent triage re-runstep in.github/workflows/qwen-triage.ymlis built on the review list alone (it checksHAS_REVIEW/STANDINGand nothing else). Driving the extracted step against a stubbedghin the close-exit state, it:stage=rerun-summarycomment ("Triage re-run completed without a new review … If this re-run was meant to review or approve, it did not") on the PR the same run just closed with an explanatory duplicate comment — self-contradictory within one timeline;::warning title=Triage re-run left no bot review, dispatching a human to a correctly-handled run.The base skill has other review-less terminal shapes (the deferred approve-on-green path, Stage 1b's re-run escalation), but none actively closes the PR and then announces on it that the re-run did nothing.
Proposed scope
Treat a bot-initiated close during the run as a terminal action before warning: fetch
gh pr view "$NUMBER" --repo "$REPO" --json state,closedAtand exit quietly (or post an adapted note) when the PR closed at/after the trigger; alternatively detect astage=1-prebot comment created after the trigger. The suite already pins this step ("notifies the author when a manual triage re-run posts no review") — an assertion covering the close-exit exemption should go red if removed.Source
Review thread
PRRT_kwDOPB-92c6c3VLson #10292 (Stage 1-pre spec review). Deferred from that PR because it touchesqwen-triage.yml, outside that PR's diff.