← ClaudeAtlas

checklisted

Run every Iron Law stage and the drift detector suite. Use when the user wants the full project health snapshot, a CI gate, or to verify nothing regressed before committing. Add --strict to promote warn-severity drift findings to errors. Activate only when the connected project contains spec.yaml or the user explicitly names Cladding; ignore ordinary requests in uninitialized projects.
qwerfunch/cladding · ★ 14 · AI & Automation · score 72
Install: claude install-skill qwerfunch/cladding
# Cladding check Run `clad check` from the project root. Runs the 15 Iron Law stages — Type / Lint / Drift / Commit / Arch / Secret / Unit / Coverage / Spec-conformance / Deliverable-smoke / Smoke / Performance / Visual / Audit / UAT — and aggregates the worst exit code. - `0` — every stage cleared or skipped clean. - `1` — at least one stage actually failed (fix-required). - `2` — every result is skip (no fail-required input on the project yet). `--strict` promotes warn-severity drift findings to error, matching the CI / pre-publish gate. The Drift stage runs every active detector under `src/stages/detectors/` — `npm run build:plugin` Phase D recounts them and writes the integer into each plugin manifest (e.g. `plugins/claude-code/.claude-plugin/plugin.json`), so the number is never hand-maintained. `--internal` shows stage codes (`stage_1.1`) instead of business names (`Type`). Default is the business-name surface; the audit log keeps internal ids regardless. ``` clad check clad check --strict clad check --internal ``` ## Gate economy (tiers) Pick the cheapest tier that answers your question — the full pre-push suite is expensive and grows with the project: - `clad check --tier=pre-commit` — drift / arch / secret only (spec-vs-code, no full unit suite). Use for fast inner-loop feedback while implementing. - `clad check --tier=pre-push --strict` — the full gate (type / lint / unit / cov + drift). This is what `clad done <featureId>` already runs, so do NOT run it