---
name: audit
description: Targeted read-only audit of a code surface - a11y, debt, doc-drift, perf - or re-verify a prior audit report
argument-hint: <a11y|debt|doc-drift|perf|verify> <target, or a link to an external tracker/doc>
disable-model-invocation: true
allowed-tools:
- Bash(git log *)
- Bash(kit a11y-check *)
---
## Dispatch
The first word of the arguments is the kind; everything after it is the kind's arguments.
| Kind | Procedure |
| ----------- | ------------------------------------------------ |
| `a11y` | [reference/a11y.md](reference/a11y.md) |
| `debt` | [reference/debt.md](reference/debt.md) |
| `doc-drift` | [reference/doc-drift.md](reference/doc-drift.md) |
| `perf` | [reference/perf.md](reference/perf.md) |
| `verify` | [reference/verify.md](reference/verify.md) |
Security audits use the built-in `/security-review`, and simplification the built-in `/simplify <path>`, which reviews a path beyond the diff.
1. Missing or unknown kind: ask via AskUserQuestion with the kinds above as options.
2. Read the kind's reference file.
3. For `a11y`, `debt`, `doc-drift`, and `perf`: resolve external context as `investigate` step 0 describes, then dispatch the auditor agent (subagent_type: kit:auditor, foreground) with the reference's steps and the resolved arguments; it writes the report, you relay its summary. For `verify`, follow the reference yourself: it dispatches the auditor with its own finding list.
4. In the reference, the `ARGUMENTS` token (written with a leading dollar sign) means everything after the kind word in the Arguments line below, and a backticked command prefixed with an exclamation mark means run that command via Bash and use its output.
Arguments: `$ARGUMENTS`
