AgentSkillsCN

inspect-skills

使用npx skill-inspector对代理技能进行质量、安全性和兼容性分析。在验证技能有效性、审计恶意行为，或检查规范合规性时，可使用此功能。

SKILL.md
--- frontmatter
name: inspect-skills
description: Inspect Agent Skills for quality, security, and compatibility using npx skill-inspector. Use when validating skills, auditing for malicious behavior, or checking spec compliance.

Inspect Agent Skills

Purpose

Run the skill-inspector CLI to analyze one or more Agent Skills for spec compliance, security (e.g. RCE, data exfiltration), and provider compatibility. Use this skill when you need to validate, audit, or harden Agent Skills.

When to Use

  • •The user asks to "inspect skills", "validate skills", "audit skills for security", or "check skill compliance".
  • •You need to verify a skill directory or repo before adopting or recommending it.
  • •You want a 0–100 score and actionable findings for a skill or skill set.

How to Use

  1. •

    List skills only (no LLM required): From the repo or directory that contains skills, run:

    bash
    ./scripts/list-skills.sh [source]

    Example: ./scripts/list-skills.sh . or ./scripts/list-skills.sh owner/repo.

  2. •

    Full inspection (requires at least one LLM API key): Run:

    bash
    ./scripts/inspect.sh [source] [options]

    The script checks for ANTHROPIC_API_KEY, OPENAI_API_KEY, GOOGLE_API_KEY, MISTRAL_API_KEY, or GROQ_API_KEY and exits with a clear error if none are set.

    Examples:

    • •./scripts/inspect.sh . — inspect all skills in current directory
    • •./scripts/inspect.sh ./my-skill --provider anthropic
    • •./scripts/inspect.sh owner/repo -s "skill-name" --json
  3. •

    Interpret results: See references/cli-usage.md for score ranges, severity levels, and --json output.

Resources