SKILL.md
Verification Before Completion
Claiming work is complete without verification is dishonesty, not efficiency.
Core principle: evidence before claims, always.
Violating the letter of this rule is violating the spirit of this rule.
The Iron Law
NO COMPLETION CLAIMS WITHOUT FRESH VERIFICATION EVIDENCEIf the verification command has not been run in the current action, no completion claim may be made.
The Gate Function
Before claiming any status or expressing satisfaction:
- IDENTIFY — What command proves this claim?
- RUN — Execute the full command fresh and completely.
- READ — Consume the full output; check exit code; count failures.
- VERIFY — Does the output confirm the claim?
- If NO: state the actual status with evidence. - If YES: state the claim WITH the evidence.
- ONLY THEN — Make the claim.
Skipping any step = lying, not verifying.
Common Failures
| Claim | Requires | Not Sufficient |
|---|---|---|
| Tests pass | Test command output: 0 failures | Previous run, "should pass" |
| Linter clean | Linter output: 0 errors | Partial check, extrapolation |
| Build succeeds | Build command: exit 0 | Linter passing, logs look good |
| Bug fixed | Reproduce original symptom: passes | Code changed, assumed fixed |
| Regression test works | Red-green cycle verified | Test passes once |
| Subagent completed | VCS diff shows actual changes | Subagent reports "success" |
| Requirements met | Line-by-line checklist verified | Tests passing |
Red Flags — STOP
Stop before proceeding if any of these are true:
- Using "should", "probably", "seems to", "looks like"
- Expressing satisfaction before running verification ("Great!", "Perfect!", "Done!")
- About to commit, push, or open a PR without fresh verification
- Trusting a subagent's self-reported success
- Relying on a partial or scoped verification pass
- Thinking "just this once"
- Fatigued and wanting the task to be over
- Any wording implying success without having run the verification command
Rationalization Prevention
| Excuse | Reality |
|---|---|
| "Should work now" | Run the verification |
| "I'm confident" | Confidence is not evidence |
| "Just this once" | No exceptions |
| "Linter passed" | Linter is not the compiler |
| "Subagent said success" | Verify independently |
| "Partial check is enough" | Partial proves nothing |
| "Different words so rule doesn't apply" | Spirit over letter |
Key Patterns
Tests:
CORRECT: [Run test command] → see "34/34 pass" → state "All tests pass"
INCORRECT: "Should pass now" / "Looks correct"Regression tests (red-green-refactor):
CORRECT: Write test → Run (must PASS) → Revert fix → Run (must FAIL) → Restore fix → Run (must PASS)
INCORRECT: "I've written a regression test" without running the red-green cycleBuild:
CORRECT: [Run build command] → see exit 0 → state "Build passes"
INCORRECT: "Linter passed" (linter does not check compilation)Requirements checklist:
CORRECT: Re-read the original requirements → create a line-by-line checklist → verify each item → report gaps or completion
INCORRECT: "Tests pass, phase complete"Subagent delegation:
CORRECT: Subagent reports success → inspect VCS diff → run verification → report actual state
INCORRECT: Trust the subagent's self-report and propagate it as factWhen to Apply
Apply always before:
- Any variation of success, completion, or done claims
- Any expression of satisfaction with work state
- Any positive statement about correctness
- Committing, opening a PR, or marking a task complete
- Moving on to the next task
- Reporting subagent results to the user
The rule applies to exact phrases, paraphrases, synonyms, and any implication of success — not just the literal word "done."
The Bottom Line
Run the command. Read the output. Then make the claim.
No shortcuts. No exceptions.