Test Runner
Expert skill for test execution and failure diagnosis.
Quick Start
Test Checklist:
- [ ] Environment configured
- [ ] Tests run or failures captured
- [ ] Failed tests diagnosed
- [ ] Coverage report generated
When to Use
- Running test suites
- Debugging failing tests
- Setting up CI/CD pipelines
- Analyzing test coverage
- Identifying flaky tests
Framework Commands
JavaScript
jest # Run all
jest path/to/test.ts # Specific file
jest --coverage # With coverage
jest --watch # Watch mode
Python
pytest # Run all
pytest tests/test.py # Specific file
pytest --cov=src # With coverage
pytest -v # Verbose
Rust
cargo test # Run all
cargo test test_name # Specific test
cargo test -- --nocapture # Show output
Go
go test ./... # Run all
go test -v ./... # Verbose
go test -cover ./... # With coverage
Bash
bats tests/ # Run all
bats tests/file.bats # Specific file
Diagnosing Failures
Step 1: Identify Error Type
| Pattern |
Likely Cause |
Action |
AssertionError: expected X got Y |
Logic error |
Check implementation |
TimeoutError |
Async issue |
Check async/await |
ModuleNotFoundError |
Missing dep |
Install dependency |
panic: runtime error |
Go panic |
Check nil/bounds |
Step 2: Isolate Test
pytest test.py::test_func -v # Python
jest --testNamePattern="test" # Jest
cargo test test_name # Rust
Step 3: Check Environment
pip list | grep pytest # Python
npm list jest # Node
./scripts/setup-test-db.sh # Test data
Coverage Analysis
Generate Reports
# Jest
jest --coverage --coverageReporters=text-summary
# Python
pytest --cov=src --cov-report=html
# Rust
cargo tarpaulin --out Html
# Go
go test -coverprofile=coverage.out ./...
go tool cover -html=coverage.out
Coverage Targets
| Type |
Minimum |
Target |
| New project |
70% |
85% |
| Legacy |
50% |
70% |
| Critical path |
90% |
95% |
Flaky Test Detection
# Run multiple times
for i in {1..10}; do pytest || echo "Failed run $i"; done
# Identify patterns:
# - Time-based (timeout issues)
# - State-dependent (order matters)
# - Async race conditions
Best Practices
DO:
- Run tests locally before commit
- Write edge case tests
- Mock external dependencies
- Keep tests fast (<100ms)
- Clean up test state
- Use descriptive names
DON'T:
- Skip tests without TODO
- Hardcode timeouts
- Test implementation details
- Share mutable state
- Ignore flaky tests
- Use sleep without reason
Quality Criteria
- Tests run or failures diagnosed
- Coverage meets targets
- No flaky tests
- Test environment documented
- CI/CD configured
- Clear error messages
Rationalizations
| Rationalization |
Reality |
| "Tests pass on my machine, CI must be flaky" |
Environment differences are real; CI is the source of truth for reproducibility. |
| "I'll fix the failing test later" |
Deferred test failures accumulate and mask real regressions. |
| "Coverage doesn't matter, I write good tests" |
Coverage metrics catch blind spots that human judgment misses. |
Red Flags
- Tests skipped without a tracked follow-up
- Coverage drops below minimum thresholds on new code
- Flaky tests re-run instead of root-caused
References
agents-docs/SKILLS.md - Skill framework guide
