Python Project Best Practice

An opinionated baseline for Python projects that both people and agents work in. Two entry points: scaffold a new project (Workflow A), or score and modernize an existing one (Workflow B). The same conventions drive both.

The bias throughout: every operation has one canonical, discoverable, non-interactive command. That is what makes a project workable by an agent, and it happens to be what makes it pleasant for humans.

When to use

• "Start a new Python project / package / CLI tool"
• "Set up pyproject.toml", "which should I use, uv or poetry / pip-tools?"
• "How do I make this installable so someone can just run it?"
• "Modernize this repo" — setup.py, requirements.txt, poetry, conda, flat layout
• Choosing between ruff/black, mypy/ty, extras/dependency-groups, just/taskipy
• Adding a CLI, notebooks, an HTTP surface, or a Rust extension to a Python repo