skillize-any-libs

A meta-skill that generates SKILL.md for any library by analyzing its source code and real-world usage in popular dependent projects.

Problem

LLMs may not be trained on all libraries (Python packages, Rust crates, npm packages, etc.), causing them to not know how to use them properly. This meta-skill helps generate comprehensive SKILL.md files that teach LLMs how to use any library.

Solution

The skill guides you through 8 steps:

1. Parse - Extract library name from user prompt
2. Detect - Find the package registry and detect language
3. Discover - Search GitHub for dependent projects
4. Ask - Let user choose reference projects
5. Clone - Clone library and selected repos
6. Analyze - Extract usage patterns, examples, APIs
7. Generate - Write SKILL.md
8. Cleanup - Remove temp files

Supported Languages

Language	Registry	Search Pattern	Tested
Rust	crates.io	depends:{lib}	✅ Yes
Python	PyPI	import {lib}	❌ Not yet
JavaScript	npm	require("{lib}")	❌ Not yet
Go	pkg.go.dev	"{lib}" in go.mod	❌ Not yet
C/C++	GitHub only	#include	❌ Not yet

Usage

Install the Skill

Copy this skill to your AI editor's skills directory:

Claude Code / OpenCode:

cp -r . ~/.claude/skills/skillize-any-libs
# or
cp -r . ~/.config/opencode/skills/skillize-any-libs

Use the Skill

When user says "skillize the lib X", load this skill and follow the workflow in SKILL.md.

Example:

• User: "skillize the lib tokio"
• You: Follow the 8-step workflow, generate SKILL.md

Output

Generated SKILL.md is saved to:

~/.claude/skills/<lib_name>-usage/SKILL.md

Requirements

• git - cloning repositories
• gh CLI (optional) - GitHub search
• curl / jq (optional) - API calls
• Network access to package registries

Philosophy

No hardcoded scripts. The meta-skill guides the LLM through the process dynamically, making it adaptable to any language or package manager.