audit

--- description: >- Structured PASS/FAIL audit of any target against any criteria. Takes a subject and a topic as arguments — the subject is what to audit, the topic defines what to check. Reads the target, builds criteria from the topic, and reports a table with specific violations and fix recommendations. /audit, audit this, run an audit, check against, compliance check, audit for, does this pass, review against criteria --- # Audit Run a structured PASS/FAIL audit of a target against a topic. ``` /audit [target] [topic] /audit skills/discuss progressive disclosure /audit app/ops/foo_op.rb ops pattern compliance /audit .context/PLAN-auth.md plan template completeness /audit client/apps/sponsor naming conventions ``` ## Process ### Step 1: Parse Target and Topic From the argument, identify: - **Target** — file path, directory, glob, or conceptual scope to audit - **Topic** — what criteria to evaluate against If ambiguous, ask: ``` What should I audit, and against what criteria? ``` ### Step 2: Build Criteria Based on the topic, construct a checklist of auditable criteria. Sources: 1. **Codebase conventions** — read `.context/codebase/CONVENTIONS.md`, `CLAUDE.md`, or relevant config files 2. **Domain knowledge** — apply known best practices for the topic (e.g., progressive disclosure layers, ops pattern requirements, accessibility standards) 3. **Reference documents** — if a spec or template exists for the topic, read it and extract checkable items 4. **Prior audits** — check for existing `.context/` files that define expected structure Present the criteria before auditing: ``` Auditing: [target] Against: [topic] Criteria: 1. [checkable item] 2. [checkable item] ... Proceed? ``` ### Step 3: Audit For each criterion, read the target and evaluate. Collect evidence: - **PASS** — criterion met, cite the evidence (file:line or specific content) - **FAIL** — criterion not met, cite what's missing or wrong - **N/A** — criterion doesn't apply to this target ### Step 4: Report ``` ## Audit: [target] against [topic] | # | Criterion | Status | Evidence | |---|-----------|--------|----------| | 1 | [item] | PASS | [citation] | | 2 | [item] | FAIL | [what's wrong] | | 3 | [item] | N/A | [why] | **Result:** X/Y passed (Z%) ``` ### Step 5: Recommendations For each FAIL, provide: - What specifically needs to change - Where (file:line if applicable) - Before/after example if the fix is non-obvious Keep recommendations concrete and actionable — no vague "consider improving X."