python-init

name	python-init
description	> Use when this capability is needed.

Python Project Initializer

Scaffold a new Python project using modern conventions. The output is a ready-to-use repository with a pyproject.toml-based build, src layout, dev tooling, and pre-configured instructions for coding agents.

What Gets Created

<project-name>/
  src/
    <package_name>/
      __init__.py
      py.typed
  tests/
    __init__.py
    test_placeholder.py
  .github/
    workflows/
      ci.yaml
  pyproject.toml
  README.md
  LICENSE
  .gitignore
  .python-version
  CLAUDE.md
  AGENTS.md

Step-by-Step

1. Gather Information

You need at minimum:

• Project name (kebab-case, used for the repo directory and package distribution name)
• Package name (snake_case, used for the importable Python package — default: project name with hyphens replaced by underscores)
• One-line description
• Python version (default: 3.11; prefer 3.11+ for modern typing syntax)

Optional:

• License (default: MIT)
• Initial dependencies
• Whether it's a CLI tool (adds a [project.scripts] entry)
• Author name and email

2. Create the Project

Read the templates in templates/ and references in references/ for content. Adapt templates based on user answers from step 1.

Key files and their purposes:

pyproject.toml

Refer to references/pyproject-guide.md for the full annotated example. Key points:

• Use [build-system] with hatchling (simple, standards-compliant, no setup.py needed)
• Pin Python version with requires-python
• Define [project.optional-dependencies] for dev/test/docs extras
• Configure all tools (ruff, pytest, mypy) in pyproject.toml — no separate config files

CLAUDE.md

This file gives Claude Code project-specific context. Start with the template in templates/CLAUDE.md.template, which includes:

• Build and test commands
• Project structure orientation
• Code style preferences
• Common workflows

The user should update this as the project evolves.

AGENTS.md

This file gives broader agent instructions (not Claude-specific). Start with the template in templates/AGENTS.md.template, which includes:

• Architecture overview placeholder
• Contribution conventions
• Testing expectations
• What not to do

Github workflows

By default, we want two GitHub workflows at package initialization: ci.yaml and python-publish.yaml. Use the templates in templates/ as a starting point, and modify as needed.

3. Initialize Git and Install

cd <project-name>
git init
git add -A
git commit -m "Initial project scaffold"

# Create a virtual environment and install in editable mode
python -m venv .venv
source .venv/bin/activate
pip install -e ".[dev]"

4. Create Notebook Entry

If the lab notebook is set up at ~/codebox/notebook/, create a project entry:

mkdir -p ~/codebox/notebook/projects/<project-name>/sessions

Write an initial STATUS.md and add the project to PROJECTS.md.

5. Verify

Run the following to confirm everything works:

python -c "import <package_name>; print(<package_name>.__version__)"
pytest
ruff check src/

Conventions

• Always use src layout. It prevents accidental imports from the working directory.
• Always include py.typed. Signals PEP 561 compliance for downstream type checkers.
• No setup.py, setup.cfg, or requirements.txt. Everything lives in pyproject.toml.
• Ruff for linting and formatting. Replaces black, isort, flake8, and pyflakes.
• Pytest for testing. Configured in pyproject.toml, not pytest.ini or conftest.py (conftest.py is fine for fixtures, just not config).

---

Source: briney/codebox — distributed by TomeVault.