You spent 10 hours building.
The AI said yes to everything.
Turns out it had no idea what your actual plan was.
This is not a hypothetical. This is Tuesday.
The vibe coder isn't the problem. The problem is that nobody built a check for the AI that's supposed to be running the show. The model doesn't know what it doesn't know. It just keeps saying yes — confidently, fluently, completely wrong.
Who watches the watcher?
That's the question ComplianceGate was built to answer.
It's an open-source compliance layer that lives in your repo and runs automated checks against your codebase — not against what the AI claimed the plan was, but against what the plan actually says. The rules. The requirements. The decisions someone made before the session started.
One command:
npx compliancegate install
Then it tells you — in plain terms — where your build drifted from your intent.
If you're using PM Butler, this is the layer that holds it honest. The Gatekeeper flags the risk. ComplianceGate checks whether anyone actually did something about it.
The AI is not your project manager. It's a very confident collaborator with no memory, no accountability, and no skin in the game.
Build accordingly.