Menú
Write or review unit tests for code in this repo. Use when writing new tests, reviewing existing tests, or adding test coverage.
Instalar con Codex o Claude Copia este prompt, pégalo en Codex, Claude u otro asistente, y deja que revise la página de la skill y la instale por ti.
Basado en la clasificación ocupacional SOC
Write or review docstrings for code in this repo. Use when writing new docstrings, reviewing existing docstrings, or editing docstrings.
Polishes recently written or modified Python code for style, dead code, docstrings, type annotations, and checks. Use when coding task is completed.
Write or review code in this repo. Use when writing new code, reviewing existing code, or editing code.
Write or review code in this repo. Use when writing new code, reviewing existing code, or editing code.
Write or revise scientific method documentation with equations, algorithm sketches, and implementation-status notes.
| name | write-tests |
| description | Write or review unit tests for code in this repo. Use when writing new tests, reviewing existing tests, or adding test coverage. |
| argument-hint | [module or file to test] |
Follow these conventions when writing or modifying tests in this codebase.
ty). Avoid using isinstance, type(), or other type assertions in tests.Tests live in tests/ within each package. Name files test_<module>.py. Use plain functions — never classes:
"""Tests for the Spam."""
import numpy as np
from numpy.testing import assert_almost_equal as aae
from pytest import approx, mark, param, raises
from spam import Spam
def test_classic_spam() -> None:
"""Test classic spam values."""
classic_spam = Spam("classic")
assert classic_spam.weight == approx(3.14159265)
Use default tolerance thresholds unless there is a specific, documented reason to override them.
| Use case | Tool | Import |
|---|---|---|
| Scalars and simple lists | approx | from pytest import approx |
| NumPy arrays (gradients, hessians) | aae | from numpy.testing import assert_almost_equal as aae |
| Exact/near-exact array equality | assert_allclose | from numpy.testing import assert_allclose |
# Scalars
assert Spam("classic").weight == approx(3.14159265)
# Lists / tuples
assert Spam("jalapeno").scores == approx([1, 2, 3])
# NumPy arrays
aae(Spam("maple").distribution, [[0, 1, 2], [3, 4, 5], [6, 7, 8]])
# Only override thresholds when truly necessary, with a comment explaining why
assert Spam("bacon").weight == approx(3.14, abs=0.005) # loose: GitHub Actions gives different results
# Exceptions
with raises(ValueError):
Spam("classic").process(threshold=-1)
Use pytest.mark.parametrize to collapse tests that differ only in inputs/outputs
Use param(..., marks=[mark.regression]) to apply markers to individual parametrized cases
@mark.parametrize(
("spam_type", "weight"),
[
("classic", 3.14159265),
("maple", 4.1),
param("low sodium", 2.9, marks=[mark.regression]),
],
)
def test_weights(spam_type: str, weight: float) -> None:
"""Test energy for multiple molecules."""
spam = Spam(spam_type)
assert spam.weight == approx(weight)
Mark tests that run less regularly but protect against regressions with @mark.regression. These are excluded from the default run (-m 'not regression').
When a regression test is related to a GitHub issue, link it in the docstring:
@mark.regression
def test_low_sodium_spam() -> None:
"""They forgot to remove the sodium originally.
https://example.com/flying_circus/spam/issues/1
"""
...
Regression tests verify:
Use pytest fixtures for any setup that is repeated across multiple tests. Place fixtures in conftest.py:
tests/conftest.pyconftest.pyPrefer importing selectively from pytest for readability
from pytest import approx, mark, param, raises
Standard alias for numpy array assertions:
from numpy.testing import assert_almost_equal as aae
ty check instead)Focus on observable behavior: computed values, raised exceptions for invalid inputs, and correct transformations.