Skip to content

fix: check agent-work delivery readiness at launch and handoff - #28

Open
ahmad-ajmal wants to merge 1 commit into
ahmad/fix-triage-requeuefrom
ahmad/issue-12-delivery-readiness
Open

ahmad-ajmal wants to merge 1 commit into
ahmad/fix-triage-requeuefrom
ahmad/issue-12-delivery-readiness

Conversation

@ahmad-ajmal

@ahmad-ajmal ahmad-ajmal commented Oct 6, 2026 •

Copy link
Copy Markdown
Collaborator

Apps that queue agent work can remain healthy after a one-shot test while later tasks stay unclaimed. This change reports delivery readiness separately from server health and requires a live delivery check at creator/modify handoff.

Changes

  • Share bounded, read-only delivery inspection across fresh/idempotent serve, list, and bridge. Report absent/stale/running local bridges, observed waiting work, and unknown queue reads without implicitly starting agent runs.
  • Record the endpoint selected by a background bridge and flag bridges still watching dev, or legacy records with an unknown endpoint. Verify app identity before sending credentials during live queue inspection.
  • Explain the independent bridge lifecycle on app stop, and avoid equating process liveness with successful delivery or absence of a local bridge with absence of an external subscriber.
  • Require creator/modify to verify live delivery and leave a detached bridge for authorized handoffs. Document static detection limits and add CLI lifecycle regressions.

Validation

  • pnpm -r build, pnpm -r typecheck, and pnpm -r test passed; conformance: 86/86. The full CLI suite was rerun successfully after the final wording correction. Three existing symlink checks were skipped because symlinks were unavailable on Windows.
  • Fresh React/Node blueprint E2E on Windows with a real SQLite database: scaffold → install → dev → validate (no unchecked steps) → promote → live serve.
  • Real Pi completed five tasks, including work queued after the detached launch command exited, work submitted after app stop/serve, and a retry after an actual harness failure.
  • Browser checks covered validation, status/filter/empty states, waiting/no-listener, progress/input-required, completion/failure/cancellation, retry, offline recovery, and 390px layout. CLI CRUD, duplicate refusal, wrong dev target, stale bridge records, idempotent start, and unknown queue state also passed.
  • git diff --check passed. Test processes were stopped. Fresh browser E2E coverage is limited to the React/Node blueprint and Pi headless route.

Dependencies and scope

Registered in GitHub stack #26 above #27, #25 and #21, targeting main. The PR's direct base is ahmad/fix-triage-requeue, keeping its diff focused on this change.

Bridge opt-in remains unchanged. OS reboot startup, crash supervision, and external subscriber discovery are outside this change.

Closes #12.

@ahmad-ajmal ahmad-ajmal self-assigned this Oct 6, 2026
@ahmad-ajmal ahmad-ajmal linked an issue Oct 6, 2026 that may be closed by this pull request
8 tasks
@ahmad-ajmal
ahmad-ajmal added this pull request to stack #26 October 6, 2026 16:45
@ahmad-ajmal
ahmad-ajmal requested a review from zfoong October 6, 2026 16:48
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Check agent-work delivery readiness at launch and handoff

1 participant