Free Download
Built for Lovable, Cursor, Bolt.new, and Replit builders. The security holes AI reliably creates.
In independent testing, 61% of AI-built features were functionally correct. Only 10.5% were secure. (CMU SusVibes, 2025)
What AI doesn't check
These aren't edge cases. They're the default output of every major AI coding tool. The code works. The security doesn't.
Token presence ≠ token validity
AI checks if a token exists. It doesn't check if it's expired, tampered with, or still valid server-side.
No Stripe signature verification
AI builds working Stripe integrations. It almost never adds `stripe.webhooks.constructEvent()`. Anyone can POST a fake payment.
Row-level security off by default
Supabase starts with RLS disabled. AI tools add queries that work, but never enable the policies that scope data per user.
NEXT_PUBLIC_ prefix on secrets
In Next.js, any variable prefixed `NEXT_PUBLIC_` is bundled into the client. AI tools add this prefix to make vars "accessible", and accidentally public.
Six categories. Built for AI-generated code.
API Key Exposure
The most common mistake: keys in bundles, git history, and env files
Auth Bypass Patterns
How AI-generated auth code gets bypassed in predictable ways
Webhook Verification Gaps
Unverified payment events that let anyone fake a successful payment
SQL Injection in AI Code
How AI builds queries that look safe but aren't parameterized
CORS Misconfiguration
Wildcard CORS in production; what it means and how to fix it
Session Handling Failures
Token expiration, session fixation, and logout that doesn't actually log out
Get the checklist
Free. Sent to your inbox immediately.
A few items from the list
Want someone to actually check your app?
The checklist tells you what to look for. A $349 Vibe Check means a senior developer looks for you, independently, with a full written report and launch verdict in 5 business days.
Get a $349 Vibe Check