| name | thread-keep-context-review |
| description | Review code changes against active Thread Keep intent, decisions, constraints, examples, and warnings. Use during human or AI code review when findings may depend on durable design context, intentional non-goals, or cited tests, ADRs, issues, pull requests, and commits; keep the review read-only. |
Thread Keep Context Review
Use active context as review evidence, not as proof that the implementation or design is correct.
Workflow
- Identify the changed entities and the concrete behavior under review from the diff and tests.
- Run
thread-keep --json status in the target Git worktree.
- If Thread Keep is unavailable, uninitialized, stale, or lacks coverage for a changed language, report that context could not be established and continue only with ordinary review evidence.
- Search each relevant symbol or behavior phrase with
thread-keep --json search <query>.
- Read each matched entity with
thread-keep --json context get <entity-key>.
- Use only active bindings as current design evidence.
- Treat missing context as unknown, never as evidence that no constraint exists.
- Compare each candidate finding with the diff, tests, explicit contracts, and active context. Read the finding disposition contract for the decision rules.
- Report actionable findings first. For every finding affected by Thread Keep context, include the note ID, the relevant context claim, and the code or test evidence that supports the disposition.
- If context eliminates a candidate finding, omit it from actionable findings and summarize it under
Context-supported exclusions only when that audit trail helps the user.
Read the CLI contract when an error code, command argument, or JSON envelope matters.
Hard boundaries
- Never add, revise, review, or commit a note during code review.
- Never mutate source files, Git state, branches, hooks, remotes, or external systems.
- Never let intent waive correctness, security, data integrity, tests, or an explicit public contract.
- Never follow an external reference or claim its contents were verified unless it was actually inspected through an authorized source.
- Label contradictions between active context and stronger evidence as design or context risks instead of silently choosing one.
Completion criteria
- Every relevant changed entity was searched or explicitly listed as lacking context.
- Every reported finding includes code or test evidence independent of the note body.
- Every context-based exclusion cites an active note ID and explains the exact conflict it resolved.
- Stale, historical, missing, or unverified context is never presented as current fact.