Quality Gates Before Completion

Iron Rule

NEVER claim completion without evidence.

This skill activates before ANY "done", "complete", "finished", or "ready" statement.

When This Skill Activates

• About to say "done" or "complete"
• Task appears finished
• User asks "is this done?"
• Before creating a PR
• Before merging code

Mandatory Quality Checks

1. Tests Pass

# Run relevant tests
pytest tests/
npm test
cargo test

Show the output. Don't just say "tests pass."

2. No Regressions

• Run the FULL test suite, not just new tests
• Existing functionality still works
• No new warnings or errors

3. Type Checks (if applicable)

# Python
mypy src/
pyright src/

# TypeScript
tsc --noEmit

4. Linting Clean

# Python
ruff check .

# JavaScript/TypeScript
eslint .

# Rust
cargo clippy

5. Manual Verification

Actually TEST the feature:

• Run the code
• Check the output
• Verify expected behavior

Evidence Format

Before claiming "done", provide:

## Verification Evidence

### Tests
Command: pytest tests/test_feature.py -v
Result: 5 passed in 0.23s

### Type Check
Command: mypy src/
Result: Success: no issues found

### Manual Test
Action: Called API endpoint with test data
Result: Returned expected response {status: "ok"}

Quick Checklist

[ ] Tests written and passing
[ ] Existing tests still pass
[ ] Type checks clean
[ ] Linting clean
[ ] Manually verified behavior
[ ] Showed evidence in response

Integration

This skill combines with:

• verification-before-completion (from Superpowers)
• test-driven-development (ensures tests exist)
• Project-specific test commands

Common Excuses (Don't Accept)

• "It should work" → Run it and prove it
• "Tests aren't set up" → Set them up first
• "Minor change, doesn't need tests" → It always needs verification
• "I'll add tests later" → No, add them now

Anti-Patterns

• Claiming done without running code
• Running only new tests, not regression suite
• Ignoring linting warnings
• Saying "I tested it" without showing output