Local, read-only audit for stale AGENTS.md, CLAUDE.md, and generic SKILL.md instructions.

Catch dangerous AI agent skills before they catch you. Zero-dependency security scanner for Agent Skills, SKILL.md and MCP configs.