From project-init
Audit, score, and improve CLAUDE.md files in the current project. Generates a quality
How this skill is triggered — by the user, by Claude, or both
Slash command
/project-init:claude-md-auditThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Audit, score, and improve CLAUDE.md files in the current project. Generates a quality
Audit, score, and improve CLAUDE.md files in the current project. Generates a quality report before making any changes, then applies targeted improvements after confirmation.
Find all CLAUDE.md files under the target directory (default: current working directory,
or the path given in $ARGUMENTS):
find . -name "CLAUDE.md" -o -name ".claude.local.md" 2>/dev/null | sort
File types and their roles:
| Type | Location | Purpose |
|---|---|---|
| Project root | ./CLAUDE.md | Primary project context — shared with team via git |
| Local overrides | ./.claude.local.md | Personal settings — gitignored, not shared |
| Global defaults | ~/.claude/CLAUDE.md | User-wide defaults across all projects |
| Package-specific | ./packages/*/CLAUDE.md | Module-level context in monorepos |
| Subdirectory | Any nested path | Feature or domain-specific context |
Note: Claude auto-discovers CLAUDE.md files in parent directories, so monorepo setups work without duplication.
For each file found, score it against the criteria in
${CLAUDE_PLUGIN_ROOT}/references/quality-criteria.md.
Quick scoring summary (full rubric in reference):
| Criterion | Weight | Question |
|---|---|---|
| Commands/workflows | High | Are build, test, run, and lint commands present? |
| Architecture clarity | High | Can Claude understand the directory structure and key files? |
| Non-obvious patterns | Medium | Are gotchas, quirks, and project-specific rules documented? |
| Conciseness | Medium | Is it free of verbose explanations and obvious filler? |
| Currency | High | Does the content reflect the current codebase state? |
| Actionability | High | Are instructions copy-paste ready — not vague? |
Grade scale:
Always output the report before making any changes.
## CLAUDE.md Quality Report
### Summary
- Files found: X
- Average score: X/100
- Files needing update: X
### File-by-File Assessment
#### 1. ./CLAUDE.md (Project Root)
**Score: XX/100 (Grade: X)**
| Criterion | Score | Notes |
|---|---|---|
| Commands/workflows | X/20 | ... |
| Architecture clarity | X/20 | ... |
| Non-obvious patterns | X/15 | ... |
| Conciseness | X/15 | ... |
| Currency | X/15 | ... |
| Actionability | X/15 | ... |
**Issues:**
- [specific problems]
**Recommended additions:**
- [what should be added]
After printing the report, ask: "Shall I apply the recommended improvements?"
After the user confirms, apply improvements using these guidelines:
Only add genuinely useful information:
Never add:
Show a diff for each proposed change before writing:
### Update: ./CLAUDE.md
**Why:** Build command was missing — sessions couldn't run the project.
```diff
+ ## Quick Start
+
+ ```bash
+ npm install
+ npm run dev # dev server on port 3000
+ ```
For CLAUDE.md templates appropriate to the detected project type, load the matching
reference from `${CLAUDE_PLUGIN_ROOT}/../references/project-types/`.
---
## Phase 5: Apply updates
Write approved changes using the Edit tool. Preserve existing content structure —
only add or update the specific sections identified in Phase 4.
---
## Structural notes to share with the user
After completing the audit, mention these practices if they aren't already in use:
1. **`<important>` tags** — wrap critical rules that must not be ignored:
```markdown
<important>Never edit files outside src/ without explicit confirmation.</important>
200-line limit — CLAUDE.md content after ~200 lines may be truncated. Split
domain-specific sections into .claude/rules/ files and import with @path.
# key shortcut — during any Claude session, pressing # prompts Claude to
incorporate session learnings into CLAUDE.md automatically.
.claude.local.md — personal preferences not shared with the team (add to
.gitignore); put user-wide preferences in ~/.claude/CLAUDE.md.
npx claudepluginhub bcmcpher/my-skills --plugin project-initGuides creation, editing, and verification of skills for AI coding agents using test-driven development with subagent scenarios. Use when authoring or debugging skills.