Audit Skills

Audit Skills

Audit Claude Code skills for quality, compliance, delegation pattern, and maintainability.

Initialization

Before auditing, initialize the environment:

Get the current UTC date, capture the project root path, ensure the temp directory exists, and clean up stale audit files. The skill-development skill provides authoritative validation guidance (auto-loaded when this command runs).

Audit Types

Type A: Standard Skill Audit

For regular skills (git-setup, markdown-linting): YAML frontmatter, naming, file structure, progressive disclosure, documentation quality.

Type B: Meta-Skill Audit

For Claude Code meta-skills (skill-development, docs-management): Zero duplication, delegation pattern compliance, metadata-only content.

The audit auto-detects type based on the skill's name and description.

Arguments

Step 1: Discover Skill Sources

Detect all skill sources in local repo and globally installed plugins.

For local discovery, check marketplace repos, single plugin repos, and .claude/​skills/. For each plugin, scan skills/ directories. Track plugin names for deduplication.

For global discovery, check ~/​.claude/​plugins/ (Unix) or %USERPROFILE%\.claude\plugins\ (Windows). Scan skills/ directories within each plugin. Skip globals with local dev versions.

Step 2: Parse Arguments

Parse flags and skill names. Read audit logs for each source.

Step 3: Present Audit Plan

Display mode, sources discovered, type distribution (A vs B), and audit queue.

Step 4: Execute Audits

For each skill, detect type then spawn the skill-auditor subagent with source, path, type (A or B), and last audit date. Run in parallel batches of 3-5.

Subagents write findings to .claude/​temp/. The main conversation thread collects results and updates audit logs using its Write/​Edit tools.

Step 4.5: Validate Findings

Unless --skip-validation flag is present:

1. Spawn the audit-finding-validator agent with:
   • project_root: The captured project root path
   • audit_type: "skill"
   • audit_files: List of .claude/​temp/​audit-*-skill-*.json file paths
2. Wait for validation to complete
3. Read updated JSON files with validation results
4. Filter out FALSE_POSITIVE findings completely before aggregation
5. Note: Filtered findings are logged to .claude/​temp/​audit-filtered-findings.json

If --skip-validation flag is present:

• Skip validation phase entirely (current speed preserved)
• Present all findings without filtering
• Note in summary: "Validation: Skipped"

Step 5: Final Summary

Report total audited by source and type, results, and details table. Note that global skill fixes must be applied manually.

Include validation statistics (if validation was performed):

• Validation performed: Yes/​No
• Findings validated: X
• False positives filtered: Y
• Verified findings: Z
• Unverified findings: W

Scoring

• PASS (85+): Ready for use
• PASS WITH WARNINGS (70-84): Functional with improvement opportunities
• FAIL (<70): Issues that should be addressed

Audit Log Location

All audit results are written to .claude/​audit/​skills.md.

Use /​audit-log skills to view current audit status.

Example Usage

Example 1: Audit All Skills

User: /​audit-skills

Claude: Discovering skill sources...

## Audit Plan
**Mode**: SMART
- Plugin: claude-ecosystem (41 skills)
- Plugin: code-quality (9 skills)
- Type A: 43, Type B: 7
- Deduplicated: claude-ecosystem (global skipped)

**Will audit**: 12 items

[Spawns skill-auditor subagents]

## Audit Complete
| Source | Name | Type | Result | Score |
| --- | --- | --- | --- | --- |
| plugin | docs-management | B | PASS | 100/​100 |
| plugin | scrape-docs | A | PASS | 95/​100 |

Example 2: Audit Specific Skill

User: /​audit-skills plugin:docs-management
Claude: PASS (Score: 100/​100) - Excellent delegation pattern compliance