agentleFS
wdm0006wdm0006/python-skills/skills/python/project-setup/SKILL.md

setting-up-python-libraries

Skill111 starsChanged 18 days agoMITView on GitHub

Sets up professional Python library projects with modern tooling (pyproject.toml, uv, ruff, pytest, pre-commit, GitHub Actions). Use when creating new Python libraries, modernizing existing projects to pyproject.toml, configuring linting/testing/CI, or setting up Makefiles and pre-commit hooks.

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 2 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Create a new library with this structure:

    line 10
  2. Very specific:

    Use src/ layout to prevent accidental imports of development code.

    line 24
  3. Very specific:

    Put --fix only in format and the pre-commit hook.

    line 80
  4. Distinctive:

    Create a new project structure:

    line 130

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Python Library Project Setup

Quick Start

Create a new library with this structure:

my-library/
├── src/my_library/
│   ├── __init__.py
│   └── py.typed
├── tests/
├── pyproject.toml
├── Makefile
├── .pre-commit-config.yaml
└── .github/workflows/ci.yml

Use src/ layout to prevent accidental imports of development code.

Core Configuration

For complete templates, see:

  • PYPROJECT.md - Full pyproject.toml with all tool configs
  • CI.md - GitHub Actions and pre-commit setup
  • MAKEFILE.md - Makefile automation patterns

Minimal pyproject.toml

[build-system]
requires = ["setuptools>=61.0", "wheel"]
build-backend = "setuptools.build_meta"

[project]
name = "my-library"
version = "0.1.0"
description = "What it does"
readme = "README.md"
requires-python = ">=3.10"
license = "MIT"  # SPDX expression
dependencies = []

[project.optional-dependencies]
dev = ["pytest>=7.0", "ruff>=0.1", "mypy>=1.0"]

[tool.setuptools.packages.find]
where = ["src"]

Essential Commands

# Setup
uv sync --extra dev
pre-commit install

# Daily workflow
ruff check src tests        # Lint
ruff format src tests       # Format
pytest                      # Test
mypy src                    # Type check

Keep local checks identical to CI

The most common CI failure is not a bug — it's a check that passes locally but fails in CI (or vice versa) because the two run different commands. Two rules prevent an entire class of "green locally, red in CI" (and chronically-red base branch) problems:

1. make lint must be read-only — never --fix. A lint target that runs ruff check --fix mutates your files and almost always exits 0, so pre-existing violations silently sit on the branch while CI's read-only ruff check goes red. Put --fix only in format and the pre-commit hook. make lint should run the exact commands CI runs.

2. CI (and make lint) must check formatting too. ruff check and ruff format are different tools: the linter passing says nothing about formatting. Gate on both, or formatting drift ships / turns a branch red unexpectedly:

ruff check src tests          # lint rules
ruff format --check src tests # formatting — REQUIRED, not implied by ruff check

Lint the same paths in the Makefile and CI (add examples/, docs/, etc. if they contain Python) — a narrower local scope lets violations accumulate in dirs CI checks. Configure ruff under [tool.ruff.lint] (not the deprecated top-level select/ignore), and keep requires-python and [tool.ruff] target-version in sync so ruff doesn't apply upgrade rules for a version you don't support.

For coverage, prefer running pytest --cov with a terminal report (--cov-report=term-missing) in the CI log over uploading to a third-party service — no external account, token, or network dependency in the gate.

Key Decisions

ChoiceRecommendationWhy
Layoutsrc/Catches packaging bugs early
Build backendsetuptoolsMature, broad compatibility
LinterruffFast, replaces flake8+isort+black
Python range>=3.10Don't pin exact versions
DependenciesMinimalMove optional deps to extras

Checklist

Project Setup:
- [ ] src/ layout with py.typed marker
- [ ] pyproject.toml (not setup.py)
- [ ] Makefile with dev/test/lint/format (lint read-only, no --fix)
- [ ] `make lint` runs the exact `ruff check` + `ruff format --check` CI runs
- [ ] Build-gating tools pinned (linter, formatter, toolchain, test runner) so upstream releases don't flip green/red on unrelated PRs
- [ ] .pre-commit-config.yaml
- [ ] .github/workflows/ci.yml
- [ ] README.md, LICENSE, CHANGELOG.md
- [ ] .gitignore

Helper Script

Create a new project structure:

python scripts/create_project.py my-library --author "Name" --license MIT

--license takes an SPDX identifier (default MIT); a LICENSE file is generated for MIT and BSD-3-Clause.

Learn More

This skill is based on the Guide to Developing High-Quality Python Libraries by Will McGinnis. See these posts for deeper coverage:

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.