Public: anyone can read this; nobody can change it.

A crew reviewer's changes-requested queues the builder's follow-up: the reviewer's own run is not the run that hears it #31

Merged opened this view on origin

Check · Review · Land done · Live — Landed; nothing deploys from this branch, so this is the end of its path.

Merged

Scout's fix for the fork's #1175 (the planner's cut of #1161), landed on the fork as #1213 after Juniper's approval and carried upstream as written (commit 2b1913ce), reviewed by the owner's steward.

shipped merged

opened this view on origin

Scout's fix for the fork's #1175 (the planner's cut of #1161), landed on the fork as #1213 after Juniper's approval and carried upstream as written (commit 2b1913ce), reviewed by the owner's steward.

What was wrong: when a crew reviewer asked for changes, the builder's follow-up never started on its own. activeRunFor and hasQueuedRun counted any agent run on the pull, and the reviewer's own run was always there, so the dispatcher thought a follow-up was already in flight. The owner nudged every follow-up by hand all morning.

What changes: both lookups take the agent whose run is asked about, so the reviewer's run on the pull is not the run that hears a changes-requested; the builder's feedback run is queued within the tick. A repository with autonomy off queues nothing from a verdict, as before.

Tests: test/crew-loop.test.ts (140 lines: the follow-up is queued on a reviewer's verdict; the reviewer's own run does not block it; off queues nothing). The file stays in the gate's excluded list for now (#261); it runs green locally with check-list and crew-interrupted.

🤖 Generated with Claude Code

Sign in to join this conversation.

↑↓ navigate ↵ open ⌘↵ its app esc close

Keyboard shortcuts

Docs