Audit Components

Audit Claude Code components for quality, security, and schema compliance.

Arguments:

• $1 (optional): Component type to audit OR --all for all types
  • Types: agent, skill, command, hook
  • Default: --all if not specified

Full arguments: $ARGUMENTS

Workflow

Single Type Audit

/agent-builder:audit agent
    ↓
Task → agent-builder
    ↓
Audit all agents in project

Full Audit (--all)

/agent-builder:audit --all
    ↓
Parallel Task invocations
    ↓
├─ agent-builder: Audit agents
├─ skill-builder: Audit skills
├─ command-builder: Audit commands
└─ hook-builder: Audit hooks
    ↓
Consolidated report

Audit Categories

Agents

• Schema compliance
• Naming conventions
• Tool permissions
• Model selection
• Content quality

Skills

• Directory structure
• No model field (critical)
• {baseDir} usage
• Auto-invoke triggers
• Resource organization

Commands

• Model format (version alias, not short)
• Argument documentation
• Security (Bash validation)
• Description clarity

Hooks

• Security (7 categories)
• Matcher specificity
• JSON schema
• Script validation

Examples

Audit All Components

/agent-builder:audit --all

Audit Only Agents

/agent-builder:audit agent

Audit Only Skills

/agent-builder:audit skill

Audit Only Commands

/agent-builder:audit command

Audit Only Hooks

/agent-builder:audit hook

Report Format

## Audit Report

### Summary
| Type | Count | Pass | Warn | Fail |
|------|-------|------|------|------|
| Agents | 5 | 4 | 1 | 0 |
| Skills | 3 | 2 | 0 | 1 |
| Commands | 8 | 7 | 1 | 0 |
| Hooks | 2 | 2 | 0 | 0 |

### Critical Issues
- [List of critical issues requiring immediate attention]

### Warnings
- [List of warnings for improvement]

### Recommendations
1. [Prioritized recommendations]

Execution

When invoked:

1. Parse audit scope from $1
2. Delegate via Task to appropriate builder(s)
3. For --all: parallel Task invocations to all builders
4. Aggregate and return consolidated report