What a human review
actually changes.
Ten failures we find again and again in applications built with AI coding tools. For each one: the code as the tool wrote it, the code as it should be, and how to check your own project in a couple of minutes.
Every example is illustrative and written for this page. We never publish client code.
6 critical
4 high
Lovable · Bolt · Cursor · Replit · Claude Code · v0
Critical — fix these first
Critical findings
Each of these can be exploited by someone with no special access and no tooling beyond a browser.
High — fix these next
High-severity findings
Not immediately exploitable, but they are what turns a working prototype into an application nobody can safely operate.
Not sure which of these you have?
That is what the free health check is for. Send us the repository and a senior engineer will tell you which of these are in your project, where, and what to fix first.