Use before declaring a long, multi-stage, or artifact-producing task done — prepare a completion CANDIDATE (not a completion). Enumerate touched user-visible surfaces, reconcile thread expectations against the plan, gather real evidence paths, and run the external gate; any uncovered surface, failed check, or review item forces blocked. Use at task wrap-up for long/high-stakes work. Do not use to declare completion yourself — only an external verifier grants `complete`.
Use before declaring a long, multi-stage, or artifact-producing task done — prepare a completion CANDIDATE (not a completion). Enumerate touched user-visible surfaces, reconcile thread expectations against the plan, gather real evidence paths, and run the external gate; any uncovered surface, failed check, or review item forces blocked. Use at task wrap-up for long/high-stakes work. Do not use to declare completion yourself — only an external verifier grants `complete`.
Wrapping up a long / multi-stage task, or any task that produces user-visible artifacts (runs, dashboards, reports, exports).
Procedure
Enumerate the touched files / user-visible surfaces this task changed.
Reconcile: compare the plan/objective against expectations expressed in the thread (not just what's written in the plan). Differences are surfaced, never silently dropped.
Map each touched surface to a check in the protected acceptance_manifest.yaml (read-only). Any user-visible surface with no check → record in review_queue.yaml and mark blocked.
Gather real evidence paths (actual artifacts/configs), not run_state assertions.
Write completion_candidate.yaml (your proposal; status = candidate_complete).
