Test

Actions

Commit rule: After any write action completes successfully, always run the Cleanup step before returning to the user. Do not wait for the user to request a commit.

Run

Input: Marker/tag, path, topic, "all", "ci", or empty (fast/default)

Resolve input to a runner command via config commands:

Report: Pass/fail counts, failure patterns, slow tests, next steps.

Create

Input: Goal, module path, feature, or ticket

Steps:

1. Analyze goal, identify modules/classes under test
2. Research existing tests: config commands.collect filtered by topic
3. Determine test location: config path_mapping rules (source pattern → test path)
4. Identify available fixtures: config fixtures list with descriptions
5. Design: happy path, edge cases, errors, parametrized/table-driven
6. Apply markers/tags: config markers list
7. Write tests following config test_pattern and existing project style
8. Verify: config commands.single with new test path
9. Check coverage: config commands.coverage

Research Gate (SHARED/research-framework.md): Read the code under test and at least one existing test in the same area before writing.

Fix

Input: Test path, "last", marker, or empty

Steps:

1. Reproduce with verbose/long-trace output
2. Analyze: trace failure, determine if test or code is wrong
3. Categorize and fix:

   Category	Fix
   Test bug	Update test
   Code bug	Fix source
   Environment	Fix setup/fixtures
   Flaky	Fix race condition, timing, shared state

4. Post-edit verification (SHARED/verify-framework.md)
5. Verify: run original test + related tests

See references/perf.md for slow test optimization.

Coverage

Input: Module path, topic, "gaps", "report"

Steps:

1. Run coverage: config commands.coverage (uses coverage_package for scope)
2. Identify gaps: uncovered functions (High), error paths (High), edge cases (Medium)
3. Design tests following existing patterns and config test_pattern
4. Write to location per config path_mapping
5. Verify coverage improved

Perf (Slow Tests)

Input: Path, marker, threshold, "all", "report"

Steps:

1. Find slow tests using duration reporting
2. Categorize: setup overhead, I/O, database, external calls, data generation
3. Profile and analyze root cause
4. Apply optimizations:

   Issue	Fix
   Repeated setup	Widen fixture/setup scope
   Real external dep	Mock or in-memory substitute
   Large test data	Reduce to minimum needed
   External process	Mock subprocess/network

5. Update markers/tags to reflect actual speed
6. Verify timing improved and coverage maintained

See references/perf.md for detailed strategies.

Cleanup

Uses SHARED/commit-framework.md with:

• file_scope: git status --porcelain filtered to config test_dir
• prefix: test
• verify_cmd: config commands.fast

See references/cleanup.md for details.

Output Format

## Test {Action}: {scope}

### Summary
- **Scope**: {marker/path}
- **Status**: PASSED/FAILED

### Results
| Status | Count |
|--------|-------|
| Passed | X |
| Failed | Y |

### {Action-specific details}

### Next Steps
- {recommendation}