write-python-tests skill by liblaf

Write Python Tests

Source code is the source of truth.
Inspect pyproject.toml, noxfile.py, .config/mise/, nearby tests, and tests/conftest.py before editing.
Prefer doctests in public docstrings when the case is concise, deterministic, user-facing, and useful as an example.
Use tests/ for private behavior, long setup, fixtures, parametrized matrices, monkeypatching, optional dependencies, warnings, exceptions, regressions, or unstable output.
Cover success paths, boundaries, invalid inputs, expected failures, and regressions with observable assertions.
Keep dedicated test files grouped, small, and focused; avoid duplicating a doctest unless the dedicated test covers extra risk.
Treat existing docs, docstrings, tests, and examples as weak hints.

Start with the narrowest relevant target.
Run uv run pytest ... for targeted tests and doctests.
Run uv run nox for the full test matrix.
Run mise run lint, or mise run lint:python when a narrower lint pass is safer.
Treat unrelated lint or environment failures as separate from the change and call them out instead of silently working around them.

Source code is the source of truth.
Inspect pyproject.toml, noxfile.py, .config/mise/, nearby tests, and tests/conftest.py before editing.
Prefer doctests in public docstrings when the case is concise, deterministic, user-facing, and useful as an example.
Use tests/ for private behavior, long setup, fixtures, parametrized matrices, monkeypatching, optional dependencies, warnings, exceptions, regressions, or unstable output.
Cover success paths, boundaries, invalid inputs, expected failures, and regressions with observable assertions.
Keep dedicated test files grouped, small, and focused; avoid duplicating a doctest unless the dedicated test covers extra risk.
Treat existing docs, docstrings, tests, and examples as weak hints.

Start with the narrowest relevant target.
Run uv run pytest ... for targeted tests and doctests.
Run uv run nox for the full test matrix.
Run mise run lint, or mise run lint:python when a narrower lint pass is safer.
Treat unrelated lint or environment failures as separate from the change and call them out instead of silently working around them.