reviewing-an-artifact

When to use

Review a standardized authored artifact against its own published standard rather than as application code — e.g. a SKILL.md / agent-skill definition. Detects the artifact and applies its rubric. The collapsed entrypoint for artifact-shaped review.

How this works

Rank the relevant lenses below by relevance to what is being reviewed, pick the breadth from the depth mode (default review), then for each selected lens load its bundle and apply it:

• Read reference/lenses/<lens>/body.md — the lens's checklist and examples. Open reference/lenses/<lens>/tool-rules.md or sources.md only if deeper tooling/provenance is called for.
• After the lenses run, merge their findings with the procedure in reference/synthesis.md — one deduplicated, ranked report with a single verdict.

Depth modes

Routing first ranks every lens whose scope the change touches by relevance — it is no longer a hard cap. A depth mode then sets the breadth (how far down the ranked list to run, plus room for judgment calls above that floor) and the severity floor. Pick the mode from the request; default to review.