How-to: run the audits¶
The framework has 5 layered audits + a pre-commit gate.
Quick reference¶
make audit # run all 5 audits in sequence
make audit-quick # structural + semantic only (fast)
make test # pytest with coverage
make all # lint + typecheck + test + audit (full local gate)
What each audit catches¶
| Audit | What it validates |
|---|---|
scripts/audit.py |
YAML/JSON parse errors; SKILL.md frontmatter; curated registry cross-refs; registry-full.yaml consistency; agent/orchestrator/routing references |
scripts/audit_semantic.py |
preferred_skills exist; governs_areas exist; sub_areas declared in CAPABILITY-GRAPH; routing intent classes match handlers |
scripts/audit_paranoid.py |
README drift; spec cross-refs; JSON schema well-formedness; curated tier fields; stray scripts |
scripts/audit_deep.py |
Inventory partials; template TBD markers; template cross-refs; README skill count drift |
scripts/verify_all.py |
ARTIFACTS-INVENTORY summary counts; markdown link resolution; cross-refs in YAML examples |
Exit codes (pre-commit)¶
| Code | Meaning |
|---|---|
| 0 | All checks passed |
| 1 | Structural or semantic issue (commit aborted) |
| 2 | _meta drift detected — run python scripts/sync_registry.py and re-stage |
Bypass (not recommended)¶
CI will still catch you — see .github/workflows/audit.yml.