| name | ultraqa |
| description | Autonomous QA cycling until all tests pass |
| argument-hint | <goal or acceptance criteria> |
| level | 3 |
UltraQA runs autonomous QA cycles until the goal is met. It repeatedly tests, diagnoses failures, and fixes until the solution is verified working.
<Use_When>
- User says "ultraqa", "run QA cycles", "test until passing"
- Need to verify all edge cases
- Want automated regression testing
- Building critical path functionality
</Use_When>
<Do_Not_Use_When>
- Single quick test is sufficient
- Tests are already comprehensive
- Exploratory manual testing needed
</Do_Not_Use_When>
<QA_Cycle>
Each cycle:
- Run tests
- If fail: diagnose the failure
- If fail: fix the root cause
- Repeat until pass (max 5 cycles)
If the same failure persists 3 times:
- Flag as fundamental issue requiring human input
- Report what needs to change and why
</QA_Cycle>
1. **Define Goal**: What does "done" mean? List acceptance criteria
- Present the criteria to the user
- Ask: "I'll run QA cycles until all these pass. Correct?"
- **CONFIRMATION REQUIRED before proceeding**
- Run Tests: Execute test suite
- Diagnose: If failures, identify root causes
- Fix: Implement fixes for root causes
- Repeat: Run tests again
- Verify: Confirm all criteria met
User: "ultraqa: verify the payment flow handles all edge cases"
Why good: Clear goal, testable criteria
User: "ultraqa: make sure the app works"
Why bad: No specific criteria to verify
<Final_Output>
## QA Report
### Goal
[Original acceptance criteria]
### Test Results
- Tests run: N
- Passed: N
- Failed: N
### Issues Found
| Issue | Fix Applied | Status |
|-------|-------------|--------|
| ... | ... | Fixed |
### Conclusion
[Goal met / Goal not met with blockers]
</Final_Output>