Verification Before Completion

The Iron Law

NO COMPLETION CLAIMS WITHOUT FRESH VERIFICATION EVIDENCE

Never say "done", "fixed", or "working" without running verification and showing the output.

The Gate Function

BEFORE making any completion claim:

1. IDENTIFY - What command proves this claim?
2. RUN - Execute the FULL command (fresh, complete)
3. READ - Full output, check exit code, count results
4. VERIFY - Does output actually confirm the claim?
   • If NO: State actual status with evidence
   • If YES: State claim WITH evidence
5. ONLY THEN - Make the claim

Common Patterns

Tests

✅ Run: npm test
   Output: "47 passing, 0 failing"
   Claim: "All tests pass"

❌ "Tests should pass now"
❌ "Looks correct"
❌ "I fixed it"

Bug Fixes

✅ REQUIRES:
   1. Test original symptom passes
   2. All other tests still pass

   Run: npm test
   Output: "48 passing (including new regression test)"
   Claim: "Bug fixed, regression test added"

❌ "Code changed, should be fixed"
❌ "Applied the fix"

Regression Tests (TDD Red-Green)

✅ REQUIRES:
   1. Write test → Run (PASS with fix)
   2. Revert fix → Run (MUST FAIL)
   3. Restore fix → Run (PASS)

   Claim: "Regression test verified - fails without fix, passes with it"

❌ "I've written a regression test" (without red-green verification)

Build/Compile

✅ Run: npm run build
   Output: "Build completed successfully"
   Claim: "Build passes"

❌ "Should compile now"
❌ "Fixed the type error"

Linting

✅ Run: npm run lint
   Output: "0 errors, 0 warnings"
   Claim: "Linting passes"

❌ "Lint errors fixed"

Red Flags - STOP

You're about to violate this skill if you're:

• Using "should", "probably", "seems to"
• Expressing satisfaction before verification ("Great!", "Perfect!", "Done!")
• About to commit/push/PR without running tests
• Trusting your own success report without fresh run
• Relying on partial verification ("the main test passes")
• About to say "fixed" without showing test output

Verification Commands by Context

JavaScript/TypeScript

npm test                    # All tests
npm test -- --grep "name"   # Specific test
npm run build               # Build
npm run lint                # Lint
npm run typecheck           # Types

Python

pytest                      # All tests
pytest path/to/test.py -v   # Specific file
pytest -k "test_name"       # Specific test
mypy .                      # Type check
ruff check .                # Lint

Go

go test ./...               # All tests
go test ./path/... -v       # Specific package
go build ./...              # Build
golangci-lint run           # Lint

Rust

cargo test                  # All tests
cargo test test_name        # Specific test
cargo build                 # Build
cargo clippy                # Lint

Claim Templates

Successful Completion

Verified: [command run]
Output: [relevant output or summary]
Result: [specific claim with evidence]

Partial Completion

Verified: [command run]
Output: [what actually happened]
Status: [X of Y complete, specific blockers]

Failed Verification

Verified: [command run]
Output: [actual error/failure]
Issue: [what's actually wrong]
Next: [what needs to happen]

Constraints

• Never claim success without fresh verification - Always run the command
• Never use hedging language - No "should", "probably", "I think"
• Never trust previous runs - Run verification fresh each time
• Never skip showing output - Include the evidence
• Never verify partially - Run the full suite, not just one test
• Always include the command - Show what was run
• Always include the output - Show what happened