| name | tiangolo-library-skills |
| description | Install and manage AI agent skills from Python/JS libraries so agents always use up-to-date patterns |
| triggers | ["install library skills for my project","set up agent skills from dependencies","add library skills for Claude Code","update agent skills for my libraries","use library-skills to configure my AI agent","install FastAPI skills for my agent","set up .agents directory with library skills","run library-skills to scan dependencies"] |
Library Skills
Skill by ara.so — Daily 2026 Skills collection.
Library Skills lets AI coding agents use libraries as intended — always up to date. Libraries (e.g. FastAPI, Streamlit) embed official AI skills in each release, and library-skills discovers those installed packages and wires their skills into your project's .agents directory as symbolic links. When you upgrade a library, its skills update automatically.
Installation & Quick Start
No global install required. Run directly with uvx (Python) or npx (JavaScript/TypeScript):
uvx library-skills
npx library-skills
For Claude Code (which doesn't yet support the standard .agents directory):
uvx library-skills --claude
What It Does
- Scans your project's installed dependencies (e.g. from the active virtualenv or
node_modules)
- Finds libraries that publish their own skills at agentskills.io
- Prompts you to select which skills to install
- Creates symbolic links under
.agents/ (and optionally .claude/skills/) pointing into the installed package
Because they are symlinks, upgrading the library (e.g. pip install -U fastapi) automatically updates the skill content — no need to re-run library-skills.
CLI Reference
uvx library-skills [OPTIONS]
| Option | Description |
|---|
--claude | Also install skills in .claude/skills/ for Claude Code compatibility |
--help | Show help message and exit |
Directory Structure After Installation
my-project/
├── .agents/
│ └── fastapi -> /path/to/venv/lib/python3.x/site-packages/fastapi/skills/
├── .claude/
│ └── skills/
│ └── fastapi -> /path/to/venv/lib/python3.x/site-packages/fastapi/skills/
├── src/
│ └── main.py
└── pyproject.toml
Python Project Workflow
1. Create and activate a virtualenv with your dependencies
uv venv
source .venv/bin/activate
uv add fastapi streamlit
2. Run library-skills
uvx library-skills
3. For Claude Code users
uvx library-skills --claude
JavaScript/TypeScript Project Workflow
1. Install dependencies
npm install
pnpm install
2. Run library-skills
npx library-skills
npx library-skills --claude
Real Code Example: FastAPI with Up-to-Date Skills
After installing FastAPI skills, your agent will use current patterns, not deprecated ones.
from fastapi import FastAPI
from pydantic import BaseModel
app = FastAPI()
class Item(BaseModel):
name: str
price: float
@app.get("/items/{item_id}")
async def read_item(item_id: int, q: str | None = None):
return {"item_id": item_id, "q": q}
@app.post("/items/")
async def create_item(item: Item):
return item
With skills installed, the agent reads the embedded SKILL.md from the FastAPI package and applies the documented patterns — including any new features added in the latest release.
How Library Authors Publish Skills
If you are a library author who wants to embed skills in your package:
- Create a
skills/SKILL.md (or similar) inside your package directory
- Publish the package to PyPI or npm
- Register at agentskills.io so
library-skills can discover it
my-library/
├── my_library/
│ ├── __init__.py
│ └── skills/
│ └── SKILL.md ← embedded skill, shipped with every release
├── pyproject.toml
└── README.md
Common Patterns
Re-running after upgrading libraries
Skills update automatically via symlinks. But if you add new libraries that have skills, re-run:
uvx library-skills
Committing .agents/ to version control
Symlinks can be committed so the whole team benefits:
git add .agents/
git commit -m "Add library skills for fastapi"
Teammates need the same virtualenv path for symlinks to resolve, so consider using relative symlinks or documenting the setup.
Checking which skills are installed
ls -la .agents/
Troubleshooting
uvx library-skills finds no skills
- Make sure you have a virtualenv activated (or the packages are installed in the current environment)
- The library must ship its own skills — not all libraries do yet; check agentskills.io
Symlinks are broken after moving the project
Symlinks are absolute by default. Re-run uvx library-skills from the new location to recreate them.
Claude Code doesn't load skills from .agents/
Use the --claude flag to also install into .claude/skills/:
uvx library-skills --claude
Skills are stale after pip install -U fastapi
Skills update automatically via symlinks — no action needed. If they don't, verify the symlink target points into the virtualenv:
readlink .agents/fastapi
Key Concepts
| Concept | Explanation |
|---|
| Library Skill | A SKILL.md (or similar file) shipped inside a library package |
.agents/ directory | Standard location where agent skills are discovered by AI coding tools |
| Symlink | library-skills uses symlinks so skills stay in sync with installed library version |
| agentskills.io | Registry of libraries that publish agent skills |
--claude | Flag to also populate .claude/skills/ for Claude Code compatibility |
Links
