Python Engineering Standards

This skill codifies 2026 Python best practices — modern tooling (uv, ruff), strict typing, and idiomatic patterns that produce maintainable, performant code.

1. Project Setup & Structure

Tooling: uv (Not pip/pipenv/poetry)

uv is the standard package manager for Python in 2026. It replaces pip, pipenv, and poetry with a single, fast Rust binary.

uv init <project-name>       # Initialize a new project
uv add <package>              # Add a dependency
uv add --dev <package>        # Add a dev dependency
uv run <command>              # Run a command in the project's venv
uv sync                      # Install all deps from uv.lock
uv python install 3.13        # Install a specific Python version

• ALWAYS commit both pyproject.toml AND uv.lock to version control.
• NEVER use requirements.txt for new projects — it is a legacy pattern.
• NEVER use pip install directly — use uv add to keep the lock file in sync.

Directory Layout: src/ Layout

my-project/
├── .python-version          # Pin Python version (e.g., 3.13)
├── pyproject.toml           # Single source of truth for config
├── uv.lock                  # Deterministic lock file
├── src/
│   └── my_package/
│       ├── __init__.py
│       ├── main.py
│       └── models.py
└── tests/
    ├── __init__.py
    ├── conftest.py          # Shared fixtures
    └── test_main.py

• Use `src/` layout for packages and production apps — it prevents tests from accidentally importing the local source instead of the installed package.
• Group by domain, not file type — prefer users/, billing/, auth/ over models/, utils/, services/.
• Use `kebab-case` for directory names, snake_case for Python modules.

pyproject.toml Configuration (PEP 621)

See references/config-templates.md for the standard pyproject.toml template.

2. Type Hints (Mandatory)

Type hints are not optional — they are foundational design tools.

Rules

• Use built-in generics (Python 3.9+): list[str], dict[str, int], tuple[int, ...]
• Use union syntax (Python 3.10+): str | None instead of Optional[str]
• Avoid `Any` — use object, Unknown, or narrow with type guards
• Use `TypedDict` for dict-shaped data with known keys
• Use `Protocol` for structural subtyping instead of ABCs
• Run static analysis: mypy --strict or pyright

Examples

# ✅ Good
def get_user(user_id: int) -> User | None:
    ...

class Config(TypedDict):
    host: str
    port: int
    debug: bool

# ❌ Bad
def get_user(user_id):  # Missing type hints
    ...

def process(data: Any):  # Untyped escape hatch
    ...

3. Code Quality: ruff

Ruff is the all-in-one linter and formatter for Python. It replaces Flake8, Black, isort, pyupgrade, and more.

ruff check --fix .    # Lint with auto-fix
ruff format .         # Format (replaces Black)

Configuration in pyproject.toml

See references/config-templates.md for the standard Ruff configuration.

• NEVER use # noqa to suppress linter errors unless the suppression is explicitly justified by a comment explaining _why_.
• NEVER use # type: ignore — fix the type error or use a proper type guard.

4. Error Handling

Patterns

• Use specific exception types — never catch bare except: or except Exception:
• Create custom exceptions inheriting from a base project exception
• Use `from` for exception chaining: raise ValueError("msg") from original_error
• Never swallow exceptions silently — at minimum, log them

# ✅ Good
class AppError(Exception):
    """Base exception for the application."""

class NotFoundError(AppError):
    """Raised when a resource is not found."""

try:
    user = get_user(user_id)
except DatabaseError as e:
    raise NotFoundError(f"User {user_id} not found") from e

# ❌ Bad
try:
    user = get_user(user_id)
except:           # Bare except — catches SystemExit, KeyboardInterrupt
    pass          # Silently swallowed

5. Testing with pytest

Structure

• AAA pattern: Arrange, Act, Assert — one concern per test
• Use fixtures for setup/teardown, not classes with setUp()/tearDown()
• Use `conftest.py` for shared fixtures across test modules
• Use `@pytest.mark.parametrize` for data-driven tests

@pytest.fixture
def client(tmp_path: Path) -> TestClient:
    """Create a test client with a temporary database."""
    app = create_app(db_path=tmp_path / "test.db")
    return TestClient(app)

@pytest.mark.parametrize("status_code,expected", [
    (200, True),
    (404, False),
    (500, False),
])
def test_is_success(status_code: int, expected: bool) -> None:
    assert is_success(status_code) == expected

Running Tests

uv run pytest                           # Run all tests
uv run pytest tests/test_main.py        # Run specific file
uv run pytest -k "test_user"            # Run by pattern
uv run pytest --cov=src --cov-report=term-missing  # Coverage

6. Advanced Patterns & Security

For advanced guidance on asyncio concurrency, Pydantic v2 data validation, and Python security best practices, see the reference file:

[Read: references/advanced-patterns.md](references/advanced-patterns.md)

9. Anti-Patterns (Never Do These)

10. Web Frameworks Quick Reference