| name | review |
| description | Run one of the 10 standing governance-forum reviews (taxonomy_blueprint.md Section 8.2 — discovery, scope, design, architecture, api_contract, threat_privacy, test_readiness, release_readiness, incident_postmortem, service_review) over an existing artifact or change. Auto-invocable when a user asks for a focused review pass rather than new production. The actual orchestration lives in .claude/workflows/review.js, invoked here via the Workflow tool. |
| model | sonnet |
| argument-hint | <forum> <subject> |
/review — governance-forum review pipeline (thin wrapper)
This skill is a thin wrapper. All real orchestration logic lives in .claude/workflows/review.js.
Procedure
- Determine which of the 10 forums applies (ask the user if ambiguous — do not guess for a governance action):
discovery, scope, design, architecture, api_contract, threat_privacy, test_readiness, release_readiness, incident_postmortem, service_review.
- If there's an active run (
.fable/current-run exists), use that run_id; otherwise mint a fresh one the same way /run does, since review artifacts still need a home under .fable/<run_id>/artifacts/.
- Invoke the Workflow tool:
Workflow({ name: "review", args: { forum: "<forum>", run_id: "<run_id>", subject: "<what's being reviewed>" } }).
- Report the forum's exit-criteria verdict (per taxonomy_blueprint.md §8.2's "Exit criteria" column) plainly: did this forum's requirements get met, or not, and why.
Constitution citations
Per N9, any reviewer that proposes changes (not just critiques) must state Preserved Invariants vs Changed Behaviors — review.js instructs every dispatched reviewer of this explicitly. Per N7, reviewers are always typed agents from FORUM_AGENTS in review.js, never generic dispatch.
