potiuk opened a new pull request, #368:
URL: https://github.com/apache/airflow-steward/pull/368

   ## Summary
   
   - `stale-sweeps.md` Sweep 4 (stale ready-for-review label) was wrongly 
counting maintainer comments that predate the `ready for maintainer review` 
label-add as evidence of author silence. Tighten the "Common trigger" so it 
only fires when a maintainer comment postdates the label-add timestamp.
   - The fix mirrors the "after the label-add timestamp" pattern that 
`classify-and-act.md` row F4 already uses for regression detection — convention 
now consistent across the skill.
   
   ## Why
   
   When `ready for maintainer review` is added, the queue moves to the 
maintainers. Pre-label maintainer comments are part of the conversation that 
*got* the PR to ready, not proof of post-promotion silence. The naive check 
meant a PR promoted on day N could be closed on day N+2 by Sweep 4b — exactly 
backwards from what the label means.
   
   ## Live incident
   
   apache/airflow#66642 — contributor was promoted to `ready for maintainer 
review` on 2026-05-24 and closed on 2026-05-26 with *"no author response for 15 
days since the last maintainer comment"*. The cited maintainer comment was from 
2026-05-11, well before the promotion. Contributor pushed back politely; PR 
reopened 2026-05-28; this PR is the skill-side fix.
   
   ## Test plan
   
   - [ ] `skill-and-tool-validate` exits 0 (verified locally).
   - [ ] On the next morning sweep, confirm no freshly-promoted PR (< 7 days 
since label-add) appears as a Sweep 4 candidate.
   - [ ] On a real stale ready-labeled PR (label > 7d, maintainer comment AFTER 
label, author silent ≥ 7d), confirm Sweep 4b still fires correctly.


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to