AgentSkillsCN

audit

验证智能体与技能生态系统的完整性。适用于检查配置合规性或生态系统健康状况时使用。

SKILL.md
--- frontmatter
name: audit
description: Validate agent and skill ecosystem integrity. Use when checking configuration compliance or ecosystem health.
argument-hint: "[--scope agents|skills|all] [--fix]"
category: workflow

/audit

Usage

bash
/audit                     # Audit all agents and skills
/audit --scope agents      # Agents only
/audit --scope skills      # Skills only
/audit --fix               # Auto-fix issues

Description

Validate Claude Code configuration ecosystem. Checks agents and skills for YAML compliance, template adherence, and integrity.

Behavior

  1. Validate: Run validation scripts
  2. Report: Present findings with severity
  3. Fix (with --fix): Apply safe automatic fixes

Validation Checks

Agents:

  • YAML frontmatter parsing
  • Required fields: name, description, tools, model, category, color
  • Template compliance (AGENT_TEMPLATE.md)
  • SYSTEM BOUNDARY statement present
  • No Task tool access

Skills:

  • SKILL.md frontmatter syntax (name, description required)
  • Skill directory structure (SKILL.md present)
  • Description quality and completeness
  • Bundled resource references valid (scripts/, references/, assets/)

Expected Output

text
User: /audit

AUDIT REPORT - Configuration Ecosystem
──────────────────────────────────────
Scope: all
Agents: 12 total | 12 compliant | 0 issues
Skills: 20 total | 20 compliant | 0 issues
Overall Compliance: 100%

✅ All validation checks passed

With Issues

text
User: /audit --scope agents

AUDIT REPORT - Agent Validation
──────────────────────────────
Agents: 12 total | 10 compliant | 2 issues

Issues Detected:

HIGH:
  - debugger.md: Missing SYSTEM BOUNDARY statement

MEDIUM:
  - researcher.md: Description exceeds recommended length

Run `/audit --fix` to auto-apply safe fixes

Fix Mode

text
User: /audit --fix

[Validation runs...]

Issues Fixed:
  ✅ debugger.md: Added SYSTEM BOUNDARY statement
  ✅ researcher.md: Truncated description

Post-fix Compliance: 100%

Notes

  • Runs validation scripts directly
  • Safe fixes applied with --fix
  • Pre-commit hooks should run this automatically
  • Typical execution: 30-60 seconds