I thought AI-generated code was the easy part. Security was the scary part.

I've been building VibeSafe because I kept seeing the same pattern with AI-built apps:

The app works.
The UI looks good.
The deployment succeeds.

But what's actually happening underneath?

Are secrets exposed?
Is authentication implemented correctly?
Are there insecure endpoints?
Are there runtime errors?
Does the deployed app actually behave the way the builder thinks it does?

That's what we're trying to solve with VibeSafe.

We're currently testing a workflow that can scan code/files/GitHub, test a live website, and run Launch Check against the actual application.

I'm still figuring out what the minimum useful security check for an AI-built app should be.

If you build with Cursor, Claude, ChatGPT, Lovable, Bolt, Replit, etc. — what's the security check you wish existed before you hit “deploy”?