verification-before-completion

name	verification-before-completion
description	Use when about to claim work is complete, fixed, or passing. Requires running verification commands and reading actual output BEFORE making any success claims. Evidence before assertions.

Verification Before Completion

Adapted from superpowers/skills/verification-before-completion. Trimmed.

Core principle

Evidence before claims, always. Claiming "done" without verification = dishonesty, not efficiency.

The Iron Law

NO COMPLETION CLAIMS WITHOUT FRESH VERIFICATION EVIDENCE

If you have not run a verification command in this message, you cannot claim it passes.

Gate function (run mentally before any claim)

BEFORE claiming a status or expressing satisfaction:

1. IDENTIFY: which command would prove this claim?
2. RUN: run the full command (fresh, not partial).
3. READ: read the full output, check exit code, count failures.
4. VERIFY: does the output confirm the claim?
   - No → state the actual status with the evidence.
   - Yes → state the claim WITH the evidence.
5. ONLY THEN: speak the claim.

Skipping any step = lying.

Mapping claim → required evidence

Claim	Required	Not enough
"Tests pass"	Test command output: 0 failures	Old run, "should pass"
"Linter clean"	Linter output: 0 errors	Partial check, extrapolation
"Build OK"	Build command exit 0	Lint passing, "logs look good"
"Bug fixed"	Test that reproduces the bug → passes	Code changed, "assumed fixed"
"Regression test works"	Red-green cycle verified (revert fix → fail; restore → pass)	Test passes once
"Requirement met"	Line-by-line checklist vs spec	"Tests pass, done"

Red flags — STOP

• Using "should", "probably", "seems to", "looks like".
• Expressing satisfaction before verification ("Great!", "Perfect!", "Done!").
• About to commit / push / open a PR before verifying.
• Trusting an agent / subagent "success" report without checking.
• Partial verification.
• Tired and wanting to wrap up.
• ANY wording that implies success when you haven't run verification.

Rationalization → reality

Excuse	Reality
"Should work now"	RUN the verification.
"I'm confident"	Confidence ≠ evidence.
"Just this once"	No exceptions.
"Linter passed"	Linter ≠ compiler ≠ runtime.
"I'm tired"	Exhaustion ≠ excuse.
"Partial check is enough"	Partial proves nothing.

Standard patterns

Tests

✅ "I ran `flutter test test/features/emergency/`. 12/12 pass. Done."
❌ "Tests should pass now."

Build

✅ "I ran `flutter build apk --release`. Exit 0. APK 28MB."
❌ "Lint passed, build is probably OK."

Bug fix with regression test

✅ Write test → run (FAIL) → fix → run (PASS) → revert fix → run (FAIL) → restore → run (PASS) → done.
❌ "I added a test" (red-green-red not actually verified).

Requirements vs spec

✅ Re-read the spec → checklist each requirement → verify each → report any gaps.
❌ "Tests pass, phase complete."

Subagent / parallel work

✅ Subagent reports success → check `git diff` → verify changes are actually right → report actual state.
❌ Trust the subagent report.

When to apply

ALWAYS before:

• Any form of success / completion claim.
• Expressing satisfaction.
• Any positive statement about work state.
• Commit, PR, marking a task done.
• Moving to the next task.
• Finalising a subagent delegation.

This rule applies to:

• The exact phrase.
• Paraphrases and synonyms.
• Implications of success.
• ANY communication suggesting completion / correctness.

Bottom line

There are no shortcuts to verification.

Run the command. Read the output. THEN claim.

Non-negotiable.

Applied to VSmartwatch

Claim	Verification command
"Flutter feature done"	flutter test test/<feature>/ + flutter analyze (zero warnings)
"FastAPI endpoint done"	pytest tests/test_<feature>.py + manual curl with valid auth headers
"Express+Prisma endpoint done"	npm test -- <file>.test.js + npm run lint zero errors
"React feature done"	npm test -- <component> + manual click-through in browser
"Migration applied"	npx prisma migrate status shows no pending OR psql -c "\d <table>" confirms new column
"Cross-repo contract works"	E2E smoke test: producer endpoint returns expected shape + consumer parses it correctly
"Bug fixed"	Regression test that reproduced the bug now passes (RED before fix → GREEN after fix → revert fix locally to confirm RED comes back — don't commit the revert)
"Lint passes"	flutter analyze / black --check . ; isort --check-only . / npm run lint exit 0