claude-setup-audit reads your CLAUDE.md, rules, skills, subagents,
and hooks — then scores them against Anthropic's official guidance and tells you the exact fix
for every issue.
Six dimensions
Each dimension scores 0–10 against principles drawn straight from Anthropic's docs. Findings are never invented — every one traces back to a rubric rule.
Line count and bloat, procedures that belong in skills, and enforcement rules that belong in hooks.
Is each instruction in the right place — CLAUDE.md vs. skill vs. rule vs. hook vs. subagent?
Path-specific rules using paths: frontmatter, with no duplication of CLAUDE.md.
Are deterministic "always/never" requirements enforced by hooks instead of hopeful advisory text?
Procedural content, descriptions specific enough to auto-invoke, and disable-model-invocation on side-effect skills.
Scoped tool lists, explicit model pinning, and correct use of isolation vs. main-thread visibility.
How it works
Run it from your project root. Every recommendation includes the exact before → after — complete file snippets, not vague advice.
Reads CLAUDE.md, .claude/rules/, skills/, agents/, and settings.json.
Grades each of the six dimensions against the rubric sourced from Anthropic.
Writes a full report to claude-setup-audit-report.md in your project root.
Prints your top quick wins — each citing a file, a line, and the exact fix.
> /claude-code-audit:claude-setup-audit
## Audit complete
Overall score: 62/100 (Fair)
| Dimension | Score |
|------------------------|-------|
| 1. CLAUDE.md Health | 6/10 |
| 2. Primitive Placement | 5/10 |
| 3. Rules Hygiene | 8/10 |
| 4. Hooks Usage | 4/10 |
| 5. Skills Quality | 7/10 |
| 6. Subagents Quality | 9/10 |
8 findings total: 2 high, 4 medium, 2 low
Full report written to: ./claude-setup-audit-report.md
Top 3 quick wins:
1. CLAUDE.md lines 45–62: deployment procedure → .claude/skills/deploy/SKILL.md
2. CLAUDE.md line 18: "always run prettier after edits" → PostToolUse hook
3. .claude/rules/api-rules.md: missing paths: frontmatter for src/api/**Get started
It ships as a Claude Code plugin — install it from a marketplace, or point Claude Code straight at the repo. No binary, no dependencies.
# In Claude Code — add this repo as a marketplace
> /plugin marketplace add WayneGoosen/claude-code-audit
# Then install the plugin
> /plugin install claude-code-audit@WayneGoosen> /plugin marketplace add anthropics/claude-plugins-community
> /plugin install claude-code-audit@claude-community# Clone and load the plugin directory directly
git clone https://github.com/WayneGoosen/claude-code-audit.git
claude --plugin-dir ./claude-code-auditWhy a skill, not a linter
A linter can flag "CLAUDE.md is over 200 lines." Only Claude can evaluate "this 30-line
deployment procedure would work better as a skill with disable-model-invocation: true
so engineers invoke it explicitly."
This tool uses Claude's judgment for the architectural questions that can't be mechanically checked — and every recommendation is specific enough to act on immediately.