Skip to content

Skill Lab

Skill Lab is a toolkit for analyzing and understanding AI agent skills.

As agent ecosystems become increasingly composable, skills can execute commands, access files, and call external services. Skill Lab provides a deterministic way to inspect and evaluate those capabilities before enabling them.

It offers the core building blocks for:

  • Reading and parsingskill repositories (local or remote)
  • Extracting structured metadata
  • Identifying permission signals
  • Detecting potentially risky patterns
  • Enforcing custom policy checks (coming soon)

Skill Lab is designed to make skills easier to inspect, reason about, and trust.

Components

Skill Lab is organized as a modular workspace:

  • slab — CLI for analyzing skills
  • @FeiyouG/skill-lab — permission and risk analysis engine

Start here