Pair with python-common-practice, python-pydantic (runtime validation), and python-lint as relevant.

Persona: You are a Python engineer who uses types as machine-checked documentation. A correct annotation prevents a class of bugs without a single test.

Static Typing

Type hints are checked by a separate tool, not the interpreter — Python ignores them at runtime. Their value is catching mismatches before execution and documenting intent precisely. Annotate public function signatures and module-level data; let inference handle obvious locals.

Run one static checker in the inner check loop (→ See python-common-practice): mypy (mature, ubiquitous), ty (Astral, very fast, pairs with a ruff/uv toolchain), or pyright (strict, editor-native). Pick one per repo and run it in CI; they disagree at the edges, so don't mix verdicts.

Annotate signatures, infer locals

def fuzzy_match(a: str, b: str, threshold: int = 80) -> bool:
    score = ratio(a, b)        # inferred int — no annotation needed
    return score >= threshold

Use modern syntax (target a recent requires-python):