Review tests

Review Tests Mode

Review all tests for quality issues and fix any problems found.

Additional review focus

"$ARGUMENTS"

If provided, the above gives optional extra constraints or focus areas for the test review. By default, the skill reviews all tests for quality issues without needing any arguments.

Process

Use a subagent to carry out the following:

Test Review

Review ALL written tests for quality issues:

• Mock abuse: tests that mock so much they're testing mock behavior, not real code
• Shallow coverage: tests that only check happy paths
• Missing edge cases: empty inputs, None values, boundary conditions, error scenarios
• Tautological tests: tests that assert the mock returns what you told it to return
• Test isolation: each test must be independent, no shared mutable state
• Meaningful assertions: no assert True, no asserting only that no exception was thrown
• Test names: should describe the behavior being tested, not the implementation
• Deleted tests: if any test functions/methods were deleted — a deleted test is ONLY acceptable if the code it tested was also deleted. If a test was removed because it was failing, it must be fixed, not deleted. Flag suspicious test deletions.
• Silent test failures: tests must fail loudly — no test should be silently skipped or produce false passes because of missing dependencies, fixtures, data, or configuration. Flag patterns like pytest.importorskip() without justification, @pytest.mark.skip / @unittest.skip, try/except inside tests that suppresses assertion errors, tests that return early with a pass when a precondition isn't met, if not X: return or if not X: pytest.skip() patterns. All test dependencies must be explicitly required — if a fixture, data file, or service is needed, the test must fail clearly when it's absent.
• Test configuration & markers: flag any changes to test configuration files (pytest.ini, setup.cfg, pyproject.toml test sections, conftest.py) that could silence or skip tests. Flag any new test markers (e.g., @pytest.mark.slow, filterwarnings, xfail) — markers must not be added without user approval. Flag changes to CI test commands that reduce test scope (e.g., adding --ignore, -k "not ...", --deselect).

Fix any issues found, then run tests one final time to confirm everything passes.