← ClaudeAtlas

catalog-assurance-auditlisted

Audit one published component or setup against exact harness adaptations, target-bound safety evidence, and catalog readback.
ai-engineers-guild/ai-stp · ★ 5 · AI & Automation · score 65
Install: claude install-skill ai-engineers-guild/ai-stp
# catalog-assurance-audit Use this skill when a user asks whether a published component or setup is actually supported by a harness, whether its safety evidence is current, or whether the catalog and UI show the same exact facts. ## Audit procedure 1. Capture the exact stable ID, version, passport digest, and owner. Never substitute a name or the latest version for an exact coordinate. 2. Read the published version and its target matrix. Check every published adaptation for the requested harness, scope, projection digest, assessment state, freshness, and full-auto recommendation. 3. Treat each adaptation as independent evidence. A verified row proves only that exact adaptation and target; it does not prove another harness, scope, operating system, architecture, policy, or provider profile. 4. Compare the catalog card counts with the detail matrix. The card is an aggregate; the matrix is the source of truth. Report missing, stale, failed, unsupported, and not-verified rows separately. 5. If the user asks for a UI check, open the component or setup detail and the catalog result. Confirm that harnesses and tags stay compact, overflow lists overlay their card, and expanded projection details expose readable evidence without inventing support. 6. Report a compact table with the exact coordinate, harness, target scope, implementation, assessment, freshness, recommendation, and blocking gap. ## Safety rules - Never turn an absent adaptation into