python

Python Skill

You are working on a Python project. Apply these best practices.

Project Setup

• Use pyproject.toml for all project metadata and tool configuration — avoid legacy setup.py / setup.cfg.
• Use uv or poetry for dependency management; commit the lockfile.
• Use Python 3.12+ features: type parameter syntax (type Point = tuple[int, int]), tomllib, match statements.

Type Hints

• Annotate all function signatures and class attributes.
• Use built-in generics directly: list[str], dict[str, int], tuple[int, ...] (no need to import from typing in 3.9+).
• Use X | Y union syntax instead of Union[X, Y] (Python 3.10+).
• Use TypedDict for dict shapes; use dataclasses.dataclass or Pydantic models for structured data.
• Run mypy --strict or pyright in CI.

Code Style

• Follow PEP 8; use ruff for linting and formatting (replaces flake8 + black + isort).
• Use f-strings for interpolation — not % formatting or .format().
• Use pathlib.Path instead of os.path for file system operations.
• Use context managers (with statements) for resources (files, DB connections, locks).

Async

• Use asyncio with async def / await for I/O-bound concurrency.
• Use asyncio.gather() for concurrent coroutines.
• Use anyio or trio for more structured concurrency when needed.
• Add from __future__ import annotations to defer evaluation of annotations in async code (Python 3.9).

Testing

• Use pytest with pytest-asyncio for async tests.
• Use tmp_path fixture for temporary files — never use hardcoded paths in tests.
• Use monkeypatch to override environment variables and module attributes.
• Use pytest.raises(ValueError, match="…") for expected exception messages.

Security

• Never use eval() or exec() with user input.
• Use secrets module for cryptographic random values — not random.
• Validate and sanitize all external inputs; use Pydantic for structured validation.
• Pin dependencies in requirements.txt or the lockfile; run pip-audit or safety in CI.