verification-gate

majiayu000's avatarfrom majiayu000

Hidden quality gate that runs before showing "Done!" to user - ensures all tests pass, build succeeds, and requirements met before claiming completion

⭐0stars🔀0forks📁View on GitHub🕐Updated Jan 5, 2026
[Testing and QA]

When & Why to Use This Skill

The Verification Gate is an automated quality assurance skill designed to ensure software reliability by acting as a hidden validation layer for AI agents. It prevents false completion claims by silently executing comprehensive test suites, build checks, and security audits, automatically resolving technical issues behind the scenes to deliver a seamless and error-free user experience.

Use Cases

  • Silent Error Resolution: Automatically identifying and fixing compilation, linting, or test failures during the development process, ensuring the user only sees functional, high-quality results.
  • Pre-Deployment Quality Control: Serving as a mandatory final checkpoint before deployment or preview generation to verify that new features meet all architectural and security standards.
  • Reliable Feature Integration: Validating that all project requirements are strictly met after adding new code, maintaining high code integrity without requiring manual developer oversight.
nameverification-gate
descriptionHidden quality gate that runs before showing "Done!" to user - ensures all tests pass, build succeeds, and requirements met before claiming completion

Verification Gate

Overview

Part of the hidden validation layer. User never sees this running, but it prevents false "Done!" claims.

Core principle: Never show success to user without evidence.

When This Runs (Automatically)

  • Before displaying "✅ Done!"
  • Before showing preview URL
  • Before /mvp:deploy
  • After any /mvp:add feature

The Gate Checklist

BEFORE showing success to user:

[ ] Tests pass (run full suite, not partial)
[ ] Build succeeds (no compilation errors)
[ ] No TypeScript/linting errors
[ ] Feature actually works (manual smoke test)
[ ] Security checks pass (from security-check skill)
[ ] Code review clean (from code-review-auto skill)

ALL must pass. ANY failure = fix silently, don't show error to user.

User Experience

What user sees:

Adding login feature...
✅ Done! Check your preview.

What happens behind the scenes:

1. Generate login code
2. Run tests → 2 failures
3. Fix failures automatically
4. Run tests → pass
5. Run build → pass
6. Security check → pass
7. Code review → 1 minor issue
8. Fix issue automatically
9. All gates pass → show "Done!"

Failure Handling

If gate fails and can auto-fix:

  • Fix silently
  • Re-run verification
  • User never knows

If gate fails and cannot auto-fix:

  • Ask simple question (no technical jargon)
  • Example: "Should login require email confirmation?"
  • NOT: "The auth middleware threw ValidationError"

Automation Script

Run all checks programmatically:

python scripts/verify.py --path /project/path --language rust
python scripts/verify.py --path /project/path --json  # JSON output

The script auto-detects language and runs appropriate checks.

Integration

Called by:

  • /mvp:build - Before showing preview
  • /mvp:add - Before confirming feature added
  • /mvp:deploy - Before deployment

Uses:

  • auto-testing - Run test suite
  • security-check - Security validation
  • code-review-auto - Code quality check
  • scripts/verify.py - Deterministic verification

Philosophy

User paid for "vibe coding" - they describe what they want, we handle how.

Showing them test failures or build errors breaks the magic.

Fix it. Don't explain it.