| name | test-and-fix |
| description | Run unit tests and automatically fix code failures, regression bugs, or test mismatches. Use when tests are failing, after implementing new features, or to repair "broken" tests. |
Test and Fix Loop
Purpose
An autonomous loop for the agent to identify, analyze, and fix failing unit tests using pytest.
Loop Logic
- Identify: Run
make test to identify failing tests (pytest with pytest-cov; terminal shows coverage summary).
- Analyze: Examine the
pytest output to determine:
- The failing test file and line number.
- The expected vs actual values (assertion errors).
- Tracebacks for runtime errors.
- Fix: Apply the minimum necessary change to either the source code (if it's a bug) or the test code (if the test is outdated).
- Verify: Re-run
make test (or uv run pytest path/to/failing_test.py for speed).
- If passed: Move to the next failing test or finish if all are resolved.
- If failed: Analyze the new failure and repeat the loop.
Termination Criteria
- All tests pass (as reported by
make test).
- Reached max iteration limit (default: 5).
- The error persists after multiple distinct fix attempts, indicating a need for human intervention.
Examples
Scenario: Fixing a logic error
make test fails in src/your_package/tests/test_dummy.py due to an assertion or import error.- Agent inspects the failing test and the implementation under
src/your_package/.
- Agent applies the minimum fix in source or test so behavior matches the intended contract.
make test passes.
Resources
