Audit Deep

Runs every audit skill in parallel against one repo, reconciles findings by severity, and proposes a single prioritized remediation plan. Replaces the "run six audits manually and try to remember what each said" pattern.

Auto-invocation triggers

• User asks "is this project healthy", "audit this repo", "tech debt review"
• Weekly per-active-repo via launchd (combine with diagnostic-skills schedule)
• Pre-release (before any version bump on a production-bound repo)
• After significant architecture change or new team member onboarding

Workflow

Phase 1 — Parallel audit dispatch (always)

Invoke in parallel via Agent tool or sequential Skill calls:

• test-health — suite proportionality, coverage, runtime
• config-drift-detect — gate compatibility
• hook-effectiveness — hooks fire/exit/latency stats
• performance-audit or performance-test — runtime perf
• security-audit — secrets, deps, OWASP
• mcp-audit — MCP server usage
• plugin-audit — plugin enabled-vs-used
• socket-audit — supply chain (npm only)
• forge-audit — if Forge ecosystem repo

Each returns a structured verdict + findings.

Phase 2 — Reconcile by severity

Aggregate all findings into one ranked list:

• CRITICAL — blocks merge / release / production safety
• HIGH — degrades workflow significantly
• MEDIUM — measurable but not blocking
• INFO — track but no action

Cross-reference: a HIGH from config-drift that explains a HIGH from test-health is reported as one root cause, not two findings.

Phase 3 — Remediation plan

For each CRITICAL + HIGH:

• Recommend the specific composite skill to fix it (fix-the-suite, secrets-rotate, gate-relax, etc.)
• Estimate the effort
• Sort by impact-per-effort

Phase 4 — Memory + handoff

Write the report to ~/.claude/projects/<slug>/memory/audit_deep_<repo>_<date>.md so trends are visible across audits. Update MEMORY.md index.

Reconciliation

Single report:

AUDIT DEEP — <repo> — <date>

Overall health:  <SCORE/100>

CRITICAL (N):
  [test-health]    1467 tests vs target 40-150 (37x ceiling)
                   Root cause: [config-drift] 99% functions gate
                   Fix: /fix-the-suite (estimated 2-4h)

HIGH (N):
  [hook-effectiveness] turn-counter spam every 10 turns
                       Fix: applied 2026-05-08 (commit 04ec576)

  [security-audit] 2 high-severity vulns in transitive deps
                   Fix: /dependency-update-batch (estimated 30min)

MEDIUM (N): ...
INFO (N):   ...

REMEDIATION PLAN (ranked by impact-per-effort):
  1. /fix-the-suite (resolves 1 CRITICAL + 2 MEDIUM)
  2. /dependency-update-batch (resolves 1 HIGH)
  3. /secrets-rotate ANTHROPIC_API_KEY (it's been 90 days)

Outputs / Evidence

• Per-audit raw verdicts
• Reconciled severity-ranked findings
• Effort-sorted remediation plan
• Memory file written for trend tracking

Failure / Stop Conditions

• Any audit skill errors → mark as PARTIAL, continue with the rest
• All audits return CLEAN → write a "no findings" memory; a clean baseline is itself valuable evidence
• If user invokes during active development → defer non-blocking audits to next scheduled run; only run the ones gating immediate work