Clear stale monitor execution state

[alvarosanchez]

Summary

• Clear stale monitor-only executionState when GitHub Sync turns a fired monitor PR wait into a healthy unassigned in_review state.
• Add regression coverage for a triggered Paperclip monitor whose linked PR remains review-ready.
• Repair the real host path where local Paperclip REST status patches succeed but do not clear raw executionState by issuing a focused SDK state-clear patch afterward.

Root Cause

Paperclip issue monitors leave historical monitor state in executionState.monitor after a monitor fires. When GitHub Sync later observed the linked PR as a healthy maintainer/review wait, it cleared the assignee but left that monitor-only execution state behind. Paperclip liveness recovery then saw an in_review issue with executionState but no resolvable currentParticipant, and created invalid_review_participant recovery work.

Validation

• pnpm build
• pnpm typecheck
• pnpm test
• Disposable real-host verification with rebuilt local plugin installed into paperclipai@2026.512.0:
  • Created an in_review Monitor PR child assigned to a test Micronaut Engineer agent.
  • Scheduled and triggered a Paperclip monitor so the issue had fired monitor-only executionState.
  • Linked the issue to real GitHub PR micronaut-projects/micronaut-core#12668, verified CLEAN with green checks.
  • Ran installed plugin sync.runNow through the Paperclip plugin action bridge.
  • Confirmed final issue state: status=in_review, no assignee, executionState=null.
  • Confirmed Paperclip liveness preview returned findings=0 and no invalid_review_participant item.

Model Used

• Model ID/version: Codex GPT-5, as exposed by this Codex session.
• Context window size: not exposed by the local Codex desktop runtime to this agent.
• Capabilities used: shell commands, local file edits, git/GitHub CLI, Paperclip disposable host verification, GitHub API via gh.

[Copilot]

Pull request overview

This PR improves GitHub Sync’s handling of Paperclip monitor-triggered issues by ensuring stale monitor-only executionState is cleared when a fired monitor PR wait becomes a healthy, unassigned in_review state. This prevents Paperclip liveness recovery from generating invalid_review_participant work due to leftover monitor state.

Changes:

• Broaden “healthy maintainer wait” detection to allow clearing monitor-only/stale executionState while keeping PRs in in_review and unassigned.
• Add a regression scenario asserting that a triggered monitor wait ends with executionState = null.
• Add a follow-up SDK patch after successful local REST PATCHes to reliably clear raw executionState on real hosts.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File	Description
src/worker.ts	Expands maintainer-wait execution-state eligibility and adds a post-REST SDK patch to clear executionState.
tests/plugin.spec.ts	Adds a regression scenario for a monitor-triggered executionState and asserts it is cleared after sync.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.