ticket-review

Run the ticket-review check, which reads the ticket cold and asks whether an agent could start from it alone. It verifies the goal statement, the three context layers, resolvable paths, vault independence, and truthful relations, then posts a one-line pass or a gap comment.

A ticket fails when paths do not resolve, accounts or connectors lack named identities, secrets require vault access, blockers are unfinished, or relations point at nonexistent tickets. Vault dependence alone fails the check, since anything the builder needs must travel onto the ticket.

ready-review grades epics and stories at discovery-exit using five checks, while ticket-review is the engineering lead's last look at dispatch time, after design and build-prep are done. Both exist because they catch different escapes at different stages.

No. The freshness rule requires the context running the check to not have produced the work being graded. If the session wrote the prep notes or acceptance criteria, it must spawn a subagent and hand it only the ticket ids and the five questions.

Use review-and-merge when verifying finished built work sitting in Review. Since the 26 Aug 2026 reassignment, ticket-review only covers the pre-dispatch pickability check; the built-work verification discipline moved entirely to review-and-merge.