ffca-audit git:20261021.79ecd2aA

Immutable. This exact content never changes and is served at /api/v1/blob/c974aecbfd07b901.

---
name: ffca-audit
description: >
  Audit an FFCA monorepo for architecture compliance: mechanical layer, naming, and cycle checks
  plus a qualitative source review, producing a per-package verdict table. Use in an FFCA monorepo,
  a features/ folder whose packages are named {feature}_domain, {feature}_data, and
  {feature}_presentation, when auditing architecture compliance, reviewing the full dependency
  graph, assessing FFCA adoption in an existing project, or before opening a PR that adds or moves
  packages. Triggers on "audit the architecture", "is the architecture still clean", "check the repo
  follows FFCA", "architecture health check", "find layer violations", "any dependency cycles",
  "review the dependency graph", and "are we ready to open this PR". Reports only and never
  auto-fixes. If the repo instead uses packages/ with _repository and _api_client packages, defer to
  vgv-ai-flutter-plugin's layered-architecture skill.
allowed-tools: Task Read Glob Grep Bash
effort: high
---

# FFCA Audit

Whole-repo architecture health check. The hook keeps individual pubspec edits compliant; this skill answers whether the repo as a whole is healthy.

The audit itself is performed by the **`ffca-layer-auditor` agent**, so the work runs in its own context and can be reused by build, review, and pre-PR flows. This skill is the entry point that dispatches it.

## What to do

1. Confirm the repo is FFCA-shaped (a `features/` folder with `{feature}_{layer}` packages). If not, say so and stop.
2. Launch the `ffca-layer-auditor` agent (subagent type `ffca-layer-auditor`) against the repo. The agent runs `dart run scripts/validate_layers.dart --all` for the mechanical pass (naming, layer rules, cycles, feature ownership, imports and barrels), then adds the source-level checks (barrel scope, DTO leakage, Command/Query necessity, module entry, misplaced packages, high fan-in) and reads `${CLAUDE_PLUGIN_ROOT}/references/ffca/` for the detail behind each.
3. Relay the agent's output: the per-package verdict table and the severity-ordered findings list. Each finding states the rule and the fix. The agent is read-only and does not auto-fix.

If the agent is unavailable for any reason, fall back to running `dart run ${CLAUDE_PLUGIN_ROOT}/scripts/validate_layers.dart --all` yourself and reporting its violations, then note that the qualitative source-level checks were skipped.