---
description: Run the auditor agent against recent work or a specific file/task
argument-hint: "[file-path or task description]"
allowed-tools:
  - Read
  - Glob
  - Grep
  - Agent
  - Bash(date:*)
---

Invoke the auditor for a quality review of recent work.

## Procedure

### Step 1 — Identify scope

If an argument was provided (file path or task description): use that as the audit scope.
If no argument: audit the last completed task (from today's daily note or task board).

### Step 2 — Select tier

Based on scope:
- T1: Single file or quick check (daily)
- T2: Completed feature or content piece (standard)
- T3: Multi-day work or weekly review (deep)
- T4: Full system audit (monthly — avoid unless needed)

### Step 3 — Invoke auditor

Spawn the `auditor` agent via Task tool:
```
Agent(auditor): Audit [scope] at tier T[N]. Read CLAUDE.md and knowledge-base.md first.
Report: PASS / FAIL / WARN / INCIDENT with specific findings.
```

### Step 4 — Surface findings

Print the auditor's verdict.
- PASS: note any WARNs for awareness
- FAIL: surface required actions — user decides next step
- INCIDENT: flag for immediate review

### Step 5 — Update task board (if needed)

If FAIL items require follow-up work: add to Task Board → Today.