You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Route scheduled-run failure issues to an agent fix lane #7535
When the twice-daily whole-tree pr-require-checks run fails on main, post-merge-demotion-alert.yml files a needs-human issue per (signal, SHA). The same issue is the demotion record that holds the unattended-merge rung. Nothing routes the underlying defect to an agent that could fix it, so a reproducible failure waits for a person.
Proposal to evaluate
Keep the deterministic part deterministic: the workflow files one issue per failure class with the failing job, the reproduce command, and a fixed label.
Leave admission to the existing intake and triage lanes: whether a known, reproducible failure class becomes autonomous-eligible is the work-class gate's call, not the alert workflow's.
Decide whether the fix item is separate from the demotion record, so clearing one does not silently clear the other.
Constraints
This changes what agents may do unattended; per AGENTS.md it needs the owner's approval before it lands.
The demotion record's merge-hold semantics must not change as a side effect.
Context
Came out of the check-placement interview (decision Q66), which kept this out of that change.
Question: A failed scheduled run on main files a needs-human issue that is also the demotion record. Nothing sends the defect to an agent. Options: (a) a separate fix issue sent to intake; (b) reuse the demotion issue; (c) keep failures human-only.
Owner's answer: (a), one fix issue per failing check, sent to intake, with the demotion record left needs-human. admission goes through intake triage and your ratification, not the work-class gate the issue names, because #7725 stopped reading work-class labels. A red main stops the drain, so this is throughput work, and your ratification remains the human step. Charge it per failing check, because #7829 now blames only the merge that turned a check red. Basis: #7535 body constraints; pr-pipeline.yaml:301-303 ("reads no work-class: label").
The owner accepted this recommendation as item D12 of a 21-decision batch in an attended backlog-shaping session on 2026-10-18. The batch question: "Accept all 21 batch 7 decisions, with my change to D13 (#6008 goes to the lane, but you merge its PR because it edits AGENTS.md)? I post a 'Decision recorded.' comment on each issue." The owner answered: "Accept, with my D13 change (Recommended)".
Posted by AI from the owner's answer in that session.
Activity
github-actions commentedon Oct 15, 2026
Thanks for the report. Automated triage handed this issue to a maintainer.
Automated triage. A maintainer reviews it.
kyle-sexton commentedon Oct 18, 2026
Decision recorded.
Question: A failed scheduled run on main files a
needs-humanissue that is also the demotion record. Nothing sends the defect to an agent. Options: (a) a separate fix issue sent to intake; (b) reuse the demotion issue; (c) keep failures human-only.Owner's answer: (a), one fix issue per failing check, sent to intake, with the demotion record left
needs-human. admission goes through intake triage and your ratification, not the work-class gate the issue names, because #7725 stopped reading work-class labels. A red main stops the drain, so this is throughput work, and your ratification remains the human step. Charge it per failing check, because #7829 now blames only the merge that turned a check red. Basis: #7535 body constraints; pr-pipeline.yaml:301-303 ("reads nowork-class:label").The owner accepted this recommendation as item D12 of a 21-decision batch in an attended backlog-shaping session on 2026-10-18. The batch question: "Accept all 21 batch 7 decisions, with my change to D13 (#6008 goes to the lane, but you merge its PR because it edits AGENTS.md)? I post a 'Decision recorded.' comment on each issue." The owner answered: "Accept, with my D13 change (Recommended)".
Posted by AI from the owner's answer in that session.
test(pr-pipeline): cover the agent-ready removal on a fix issue reopen
fix(pr-pipeline): report red per-merge full runs to the red-main alert (