Setting Up a Python Library with uv

Overview

This skill scaffolds a Python library project using uv in the current working directory.

Workflow

Step 1: Gather Project Details

Ask the user for:

1. •Python version - default to 3.12 if not specified
2. •Project description (optional) - for pyproject.toml and README

Step 2: Initialise the Project

Run from the current directory (user has already created and cd'd into it):

uv init --lib . --python <version>

Step 3: Add Dev Dependencies

uv add --dev ipython jupyterlab ruff pytest pytest-cov

Step 4: Create Project Structure

Create directories:

mkdir -p tests
mkdir -p nbs
mkdir -p scratch/scripts
mkdir -p scratch/nbs
touch tests/__init__.py

• •tests/ - test files
• •nbs/ - notebooks (tracked in git)
• •scratch/scripts/ - throwaway scripts (ignored)
• •scratch/nbs/ - throwaway notebooks (ignored)

Add scratch/ to .gitignore:

echo "scratch/" >> .gitignore

Create a minimal test file tests/test_placeholder.py:

def test_placeholder():
    """Placeholder test to verify pytest works."""
    assert True

Step 5: Create README.md

Create a minimal README using the directory name as the project name:

# <directory-name>

<description if provided>

Step 6: Create CLAUDE.md

Create a CLAUDE.md with project context:

# <project-name>

Python library using uv for dependency management.

## Commands

Run `just` to see available commands (test, lint, format, shell, jupyter).

## Project Structure

- `src/<package>/` - library source code
- `tests/` - test files
- `nbs/` - notebooks (tracked)
- `scratch/` - throwaway scripts and notebooks (gitignored)

## Dependencies

- Add runtime deps: `uv add <package>`
- Add dev deps: `uv add --dev <package>`
- Run anything: `uv run <command>`

Step 7: Create justfile

Create a justfile with common development commands:

# List available commands
default:
    @just --list

# Run tests
test *args:
    uv run pytest {{args}}

# Run tests with coverage
test-cov:
    uv run pytest --cov

# Format code
fmt:
    uv run ruff format .

# Lint code
lint:
    uv run ruff check .

# Lint and fix
lint-fix:
    uv run ruff check . --fix

# Format and lint
check: fmt lint

# Start IPython shell with autoreload
shell:
    uv run ipython -i -c "get_ipython().run_line_magic('load_ext', 'autoreload'); get_ipython().run_line_magic('autoreload', '2')"

# Start Jupyter Lab
jupyter:
    uv run jupyter lab

Step 8: Git Repository Setup

First, discover available GitHub organisations:

gh org list

Then ask the user using AskUserQuestion with these options:

1. •No remote - Keep it local only
2. •Public repository (personal) - Create public repo in personal account
3. •Private repository (personal) - Create private repo in personal account
4. •Organisation repository - Show list of orgs from gh org list and let user pick, then ask public/private

If user selects organisation, present the orgs discovered and ask which one, then ask public or private.

Creating the repository:

For personal repos:

gh repo create <repo-name> --public|--private --source . --push

For organisation repos:

gh repo create <org-name>/<repo-name> --public|--private --source . --push

If no remote wanted:

git init
git add .
git commit -m "Initial commit: scaffold Python library with uv"

Checklist

• • Confirm Python version (default 3.12)
• • Get optional project description
• • Run uv init --lib .
• • Add dev dependencies (ipython, jupyterlab, ruff, pytest, pytest-cov)
• • Create directories (tests/, nbs/, scratch/scripts/, scratch/nbs/)
• • Add scratch/ to .gitignore
• • Create placeholder test
• • Create README.md
• • Create CLAUDE.md
• • Create justfile with dev commands
• • Fetch available GitHub orgs with gh org list
• • Ask about git remote setup
• • Initialise git and optionally create remote

Notes

• •Always use uv init --lib . for library projects (creates proper src layout)
• •The project name is derived from the current directory name
• •Verify uv and gh are installed before starting
• •Use gh auth status to check GitHub authentication if needed