Loading...
Loading...
Use this before claiming work is complete, fixed, or tested—before committing or creating a PR—you must run validation commands and confirm the output before claiming success; always back up assertions with evidence
npx skill4agent add jnmetacode/superpowers-zh verification-before-completionNo claiming completion without fresh validation evidenceBefore claiming any status or expressing satisfaction:
1. Identify: What command proves this conclusion?
2. Run: Execute the full command (re-run, full execution)
3. Read: Full output, check exit code, count failures
4. Validate: Does the output support the conclusion?
- If no: State the actual status with evidence
- If yes: State the conclusion with evidence
5. Only then: Draw the conclusion
Skipping any step = lying, not validating| Conclusion | Required | Insufficient |
|---|---|---|
| Tests pass | Test command output: 0 failures | Previous run results, "should pass" |
| Linter has no errors | Linter output: 0 errors | Partial checks, inference |
| Build succeeds | Build command: exit 0 | Linter passes, logs look okay |
| Bug fixed | Original symptom tested: passes | Code changed, assumed fixed |
| Regression test is valid | Red-green cycle verified | Test passed only once |
| Agent completed | VCS diff shows changes | Agent reports "success" |
| Requirements met | Checked against itemized checklist | Tests passed |
| Excuse | Reality |
|---|---|
| "It should work" | Run the validation command |
| "I’m confident" | Confidence ≠ evidence |
| "Just this once" | No exceptions |
| "The linter passed" | Linter ≠ compiler |
| "The agent said it succeeded" | Independent validation |
| "I’m tired" | Fatigue ≠ excuse |
| "Partial checks are enough" | Partial checks prove nothing |
| "This rule doesn’t apply if I phrase it differently" | Spirit matters more than literal wording |
✅ [Ran test command] [Saw: 34/34 pass] "All tests passed"
❌ "Should pass now" / "Looks correct"✅ Write → Run (pass) → Revert fix → Run (must fail) → Restore → Run (pass)
❌ "I wrote a regression test" (no red-green verification)✅ [Ran build] [Saw: exit 0] "Build passed"
❌ "The linter passed" (linter doesn’t check compilation)✅ Re-read plan → Create checklist → Validate item by item → Report gaps or completion
❌ "Tests passed, phase complete"✅ Agent reports success → Check VCS diff → Validate changes → Report actual status
❌ Trust the agent’s report