MetalHexx/rad-run-tests

Run Tests

Execute the project's test suite and report structured results. This is an instructions-only skill with no template — it guides the agent through discovering and running tests, then reporting results.

When to Use This Skill

• After implementing code changes that need validation
• When a task's acceptance criteria include "all tests pass"
• When the Coder Agent needs to verify test results before completing a task
• When checking if prior test failures have been resolved after a corrective task

Workflow

Step 1: Discover the Test Runner

Find the project's test configuration:

1. Check package.json for scripts.test (Node.js projects)
2. Check for jest.config.*, vitest.config.*, playwright.config.*
3. Check for pytest.ini, setup.cfg, pyproject.toml (Python projects)
4. Check for Cargo.toml (Rust projects)
5. Check for go.mod (Go projects)
6. Read orchestration.yml for any custom test commands

Step 2: Run All Tests

Execute the discovered test command. Common patterns:

• Node.js: npm test or npx jest or npx vitest run
• Python: pytest or python -m pytest
• Rust: cargo test
• Go: go test ./...

Step 3: Parse Results

Capture and structure the output:

• Total tests run
• Tests passing
• Tests failing (with failure details)
• Tests skipped
• Coverage percentage (if available)

Step 4: Run Targeted Tests (if applicable)

If the task handoff specifies particular test files:

1. Run only those specific test files first
2. Then run the full suite to check for regressions

Step 5: Report Structured Results

Format results for inclusion in task output:

Tests: {passing}/{total} passing
Failures:
- {test name}: {error message}
Coverage: {X}% (if available)
Build: {pass/fail}

Key Rules

• Actually run the tests: Never assume results — execute the command and parse real output
• Report failures with details: Include the test name, file, and error message for every failure
• Run the full suite: Even if only specific files were changed — check for regressions
• Note flaky tests: If a test passes on retry, note it as potentially flaky
• Build must also pass: Run the build command alongside tests — both must succeed

Troubleshooting

| Issue | Resolution | |-------|------------| | No test runner found | Check project root for config files; ask human if unclear | | Tests timeout | Increase timeout or note as an issue | | Missing dependencies | Run install command (npm install, pip install, etc.) first | | Environment issues | Check for required env vars, database connections, or services |