Project Audit

$SKILL_DIR denotes the physical directory containing the selected SKILL.md; it is path notation, not an ambient environment variable or repository-root lookup. Resolve it before running the audit from an arbitrary consumer cwd.

When NOT to Use

• Security-specific review (use change-verdict --mode security)
• Runtime performance profiling
• Mid-development review (use change-verdict --mode code)

Procedure

1. Run node "$SKILL_DIR/scripts/audit.js" --json to collect deterministic scores
2. Parse the JSON output — overall_score, status, dimensions, checks, findings, next_actions
3. If status = Blocked (P0 findings) — highlight critical gaps, suggest immediate fixes
4. If status = Needs Work (P1 findings) — format improvement roadmap by dimension
5. If status = Healthy — summarize strengths, note any P2 improvements
6. Add qualitative interpretation beyond the scores (e.g., "test ratio is good but concentrated in unit tests")

Script Integration

The audit script runs 12 deterministic checks across 5 dimensions:

Scoring Model

• Each check: 1 (pass) / 0.5 (partial) / 0 (fail) / N/A (skipped)
• Dimension score: applicable_sum / applicable_count * 100
• Overall score: average of dimension scores
• Confidence: applicable_checks / total_checks per dimension

Status Determination

Script Failure Fallback

If the script fails, report the error and suggest running manually:

node "$SKILL_DIR/scripts/audit.js" --json

Output

## Project Audit Report

| Field | Value |
|-------|-------|
| Repo | [name] |
| Score | **[N]/100** |
| Status | [icon] [status] |

### Dimensions
[table of dimension scores]

### Checks
[list of check results with suggestions]

### Next Actions
[prioritized action items]

## Gate: ✅/⛔

References

• references/check-catalog.md — Check definitions, scoring criteria, ecosystem detection (read when investigating a specific check result)
• references/output-template.md — Report format examples and JSON schema (read when customizing output)

Verification

• Script ran successfully
• All 12 checks executed (or marked N/A with reason)
• Qualitative interpretation added beyond raw scores
• Next actions are actionable (include commands where applicable)