# SkillAnything

A meta-skill that generates production-ready AI agent skills for Claude Code, OpenClaw, Codex and other platforms from a single target, such as a CLI tool, a REST API, a Python library or a described workflow.

Given a target, SkillAnything runs a seven-phase pipeline: analyze auto-detects the target type and extracts its capabilities; design maps those capabilities to a skill architecture; implement generates a SKILL.md file with scripts and references; test plan auto-generates evaluation cases and trigger queries; evaluate benchmarks the skill with and without it in place and grades the results; optimize improves the skill's description through a train/test loop; and package produces multi-platform distribution packages.

Target auto-detection covers a CLI tool through `which` and `--help` parsing, a REST API through an OpenAPI or Swagger specification URL, a library through its package name on pip or npm, a workflow described step by step, and a service documented on the web.

It installs by cloning its repository into a platform's skills directory, for example `~/.claude/skills/skill-anything` for Claude Code, `~/.openclaw/skills/skill-anything` for OpenClaw, or `~/.codex/skills/skill-anything` for Codex.

Website: https://github.com/AgentSkillOS/SkillAnything