Agent Skills: Audit Hooks Command

Audit Hooks Command

Audit Claude Code hooks for quality, compliance, 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 any stale audit files if the user confirms. The hook-management skill provides authoritative validation guidance (auto-loaded when this command runs).

What Gets Audited

• hooks.json configuration structure and validity
• Hook script existence and structure
• Matcher configuration appropriateness
• Environment variable naming conventions
• Decision control usage
• Test existence and coverage

Command Arguments

| Argument | Description | | --- | --- | | (none) | Smart mode: audit only modified, never-audited, or stale (>90 days) hooks | | --force | Audit ALL hooks regardless of status | | --skip-validation | Skip finding validation (faster, but may include false positives) | | --plugin-only | Only audit local plugin hooks | | --project-only | Only audit project hooks (.claude/hooks/) | | --global-only | Only audit globally installed plugin hooks | | hook-name | Audit specific hook(s) by name | | plugin:name | Explicitly target plugin hook |

Step 1: Discover Hook Sources

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

For local discovery, check marketplace repos, single plugin repos, and .claude/hooks/. Track plugin names for deduplication.

For global discovery, check ~/.claude/plugins/ (Unix) or %USERPROFILE%\.claude\plugins\ (Windows). Skip globals that have local dev versions.

Step 2: Parse Arguments

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

Step 3: Present Audit Plan

Display mode, sources discovered, deduplication status, and audit queue.

Step 4: Execute Audits

For each hook, spawn the hook-auditor subagent with source, path, 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: "hook"
   • audit_files: List of .claude/temp/audit-*-hook-*.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, results, and details table. Note that global hook 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

Important Notes

Deduplication

Local dev repo plugins take precedence over globally installed versions. Global hooks are read-only - report findings but recommend manual fixes.

Cross-Platform Paths

| Platform | Global Plugins | | --- | --- | | Unix | ~/.claude/plugins/ | | Windows | %USERPROFILE%\.claude\plugins\ |

Audit Log Location

All audit results are written to .claude/audit/hooks.md.

Use /audit-log hooks to view current audit status.

Example Usage

Example 1: Audit All Hooks

User: /audit-hooks

Claude: Discovering hook sources...

## Audit Plan
**Mode**: SMART
- Plugin: claude-ecosystem (4 hooks)
- Local: .claude/hooks/ (2 hooks)
- Deduplicated: claude-ecosystem (global skipped)

**Will audit**: 3 hooks

[Spawns hook-auditor subagents]

## Audit Complete
| Source | Hook | Result | Score |
| --- | --- | --- | --- |
| plugin | prevent-backup-files | PASS | 100/100 |

Example 2: Audit Specific Hook

User: /audit-hooks plugin:prevent-backup-files
Claude: PASS (Score: 100/100)