/audit

name: audit
description: "Project-wide health audit pipeline that fans out to all analysis skills in parallel, evaluates findings, and produces a unified report at .turbo/audit.md. Use when the user asks to \"audit the project\", \"run a full audit\", \"project health check\", \"audit my code\", \"codebase audit\", or \"comprehensive review\"."

Audit

Project-wide health audit. Fans out to all analysis skills, evaluates findings, and writes `.turbo/audit.md` and `.turbo/audit.html`. Analysis-only — does not apply fixes.

Task Tracking

At the start, use `TaskCreate` to create a task for each phase:

1. Scope and partition 2. Threat model 3. Run analysis skills 4. Run `/evaluate-findings` skill 5. Generate markdown report 6. Generate HTML report

Step 1: Scope and Partition

If `$ARGUMENTS` specifies paths, use those directly (skip the question).

Otherwise, use `AskUserQuestion` to confirm scope:

• **All source files** — audit everything
• **Specific paths** — user provides directories or file patterns
• **Critical paths** — heuristically identify high-risk areas (entry points, auth, data handling, payment processing)

Once scope is determined:

1. Glob for source files in the selected scope. Exclude generated and vendored directories (`node_modules/`, `dist/`, `build/`, `vendor/`, `__pycache__/`, `.build/`, `DerivedData/`, `target/`, `.tox/`, and others appropriate to the project). 2. Partition files by top-level source directory. Cap at 10 partitions. If more than 10 top-level directories exist, group related directories or use `AskUserQuestion` to narrow scope. If a single directory contains 50+ files, sub-partition it by its immediate subdirectories.

Step 2: Threat Model

Check whether `.turbo/threat-model.md` exists. When it exists, collect each surface's cited paths from **Entry points and sinks** and **Hot files**, ignoring any `:line` suffix. Treat the model as superseded when a surface has no cited path that still resolves. When it exists and is not superseded, continue to Step 3.

If missing or superseded, use `AskUserQuestion` to ask whether to create one before proceeding. The security review benefits from threat model context, but creating one adds time.

• **Yes** — launch an Agent tool call (`model: "opus"`, no `name`) whose prompt instructs it to invoke the `/create-threat-model` skill via the Skill tool. Wait for it to report before continuing; do not relaunch it if it has not yet reported.
• **No** — continue without a threat model.

Step 3: Launch All Analysis Agents

Before dispatching, read the project's test configuration and CI workflow to identify any test tier that resets a shared external resource between tests, such as a database, a fixed port, or a cache. Such tiers have no cross-process interlock, so agents running them concurrently wipe each other's state and return failures indistinguishable from defects in the code under review. Name any such tier to every agent as off-limits.

Emit all analysis Agent tool calls below in one assistant message. Each Agent call uses `model: "opus"` and no `name`. Wait for every agent to report before continuing. Do not begin the next step on a partial set, and do not relaunch an agent that has not yet reported. Each Agent's prompt instructs the subagent to invoke its assigned skill via the Skill tool, with the partition's file list passed in for partitioned skills, and to treat the shared working tree and its git index as read-only — any empirical check runs in an isolated `git worktree` created under `$TMPDIR` and discarded afterward. HEAD stays where it is: read other refs with `git show <ref>:<path>` rather than `git checkout` or `git switch`. Refer to that worktree by absolute path in every command and join chained steps with `&&`, so a failed step cannot leave the rest running in the shared checkout. Run teardown and verification as their own commands. Give that worktree its own dependency install rather than reaching the shared tree's install by any route: removing a worktree deletes through symlinks, and a redirected suite writes into the shared install. When its own install is not possible, the check is left unrun and reported as such. Afterward the agent verifies that `git worktree list` no longer shows the worktree, that `git status --short` is clean, that HEAD is still on the branch it started on, and that the shared tree's dependency directory still resolves (a destroyed install leaves `git status` clean, since it is gitignored). Damage the agent cannot repair is reported with the exact repair command in place of findings.

Expect (6 partitioned rows × number of partitions, plus 5 project-wide rows) Agent tool calls total. State the count explicitly when emitting the calls.

Partitioned Skills

For each skill below, launch **one Agent per partition** with the partition's file list in the prompt. Pass `(skip peer review)` annotations through to `/review-code` as an opt-out so it runs internal reviews only — `/peer-review` is scheduled as its own row to avoid duplicate peer-review runs.

| Skill | Scope | |---|---| | `/review-code` with `correctness` (skip peer review) | File list | | `/review-code` with `security` (skip peer review) | File list | | `/review-code` with `api-usage` (skip peer review) | File list | | `/review-code` with `consistency` (skip peer review) | File list | | `/review-code` with `simplicity` (skip peer review) | File list | | `/peer-review` | File list |

Project-Wide Skills

| Skill | Notes | |---|---| | `/review-code` with `coverage` (skip peer review) | Project-wide | | `/review-dependencies` | Project-wide | | `/review-tooling` | Project-wide | | `/review-agentic-setup` | Project-wide | | `/find-dead-code` | Has its own partitioning |

Step 4: Run `/evaluate-findings` Skill

Aggregate all findings from all agents. Run the `/evaluate-findings` skill once on the combined set.

Step 5: Generate Markdown Report

Write `.turbo/audit.md` using the templat