No scan without consent

Pre-launch security checklist for AI-built web apps

Scanners read your source code. But incidents happen in what you deployed. The five items below are the most common real-world failures in apps built with Lovable, Bolt, or v0 — each with how to check it yourself and how we verify it from the outside.

A scan of ~1,400 live vibe-coded apps found 400+ exposed secrets and 175 PII exposures (Escape.tech, 2025). 45% of AI-generated code carries an OWASP Top 10 flaw (Veracode, 2025).

1. Are API keys shipped in your browser bundle?

In your browser devtools → Network/Sources, search for strings like sk-, AKIA, Bearer. A key in the frontend is visible to everyone. Move it to the server (edge function/backend). — Our audit fingerprints key candidates in the actually-served bundle.

2. Is your database open to the world?

If you use Supabase or similar, check that Row Level Security (RLS) is on and the anon key can't read whole tables. Many vibe-coded apps ship without RLS. — We check sensitive paths and the public surface from the outside, read-only.

3. Any APIs or paths open without auth?

Open paths like /api/*, /admin, /.env, /.git while logged out. A 200 is a problem. — We probe these common paths from the outside and mark what we couldn't read as “undetermined.”

4. Are security headers and TLS set?

Confirm HTTPS is enforced, the certificate is valid, and headers like CSP, HSTS, X-Content-Type-Options are present. Without them the app still runs, but the attack surface grows. — Our receipt lists header/TLS status as items.

5. What about what you can't check yourself?

Even after all of the above, sweeping the whole deployed reality alone is hard. We audit it from the outside, consent-based and read-only, and report verdicts as pass/fail/undetermined — honestly writing down even what we couldn't see. It's free.

Get my web app audited — free →

Related: Request an audit · llms.txt · Operator star365 · Trading fleet metaquant · Sitemap