Security review for AI-built apps

I check your app's security before someone else does

Did you build a web app, mobile app or API with Claude Code, Codex, Cursor or Lovable? I personally review the code, the configuration and how it handles data, and you get a report with the issues I found, ranked by priority. Then I fix them, unless you would rather your team does it.

Before building products I spent five years in application security, doing penetration tests and security code reviews. Today I apply that experience to apps written with AI agents.

10+ years in software development · 5 years in application security · Fastweb Hall of Fame 2020

Luca Di Domenico, security review for AI-built apps

An AI agent writes code that works, but it doesn't always check who can use it

AI agents quickly produce features that pass the tests. Security issues, though, don't show up when you use the app: they're found by whoever goes looking for them, sometimes before you do.

Other users' data

Changing a number in the URL or in a request is enough to read someone else's orders, documents or profile.

Keys in the wrong place

An API key in the frontend code or in the repository lets anyone use your services on your bill.

A database with no rules

A table without access rules, such as Supabase RLS, can be read and changed from outside.

What I check

Where AI-built apps go wrong most often

Authentication and permissions

Login, sessions, roles and the checks that stop one user from seeing or changing another user's data.

API keys and secrets

Where keys, tokens and passwords are stored, who can read them and what happens if they leak.

Database rules

Table permissions, access rules such as Supabase RLS, functions and high-privilege keys.

Input and injection

How the app handles data coming from users and external services: SQL injection, XSS, file uploads.

Dependencies

Libraries with known vulnerabilities, unused packages or packages from untrusted sources.

Hosting configuration

Environment variables, security headers, CORS, test environments reachable from outside.

Personal data

What personal data you collect, where it ends up and whether anyone can see it without the right to.

AI features and agents

Prompt injection, data sent to the model and the permissions of agents that take actions in the app.

How it works

From application to re-check

01

Application

You tell me what the app does, what stack it runs on and what worries you.

02

Quote

I assess the size and scope of the review and send you a quote.

03

Review

You give me access to the code and the configuration. I analyze them personally.

04

Report

You get the issues I found ranked by priority, why they matter and how to fix them.

05

Fixes

I fix the issues I found. If you or your team would rather fix them, tell me and the report is all you need.

06

Re-check

After the fixes, mine or yours, I check the app once more. The re-check is included.

If your team does the fixes, the report can also be handed to an AI agent.

What the review covers

Included
  • Manual code review
  • Review of the database, hosting and third-party service configuration
  • Report with issues ranked by priority
  • Guidance on how to fix each issue
  • Fixing the issues, unless you prefer to do it yourself
  • One re-check after the fixes
Not included
  • Penetration testing of the live app
  • Certifications or compliance audits
  • A guarantee that no vulnerabilities remain
Who does the review

I do the review myself, not an automated tool

I spent five years in application security at specialized consulting firms: penetration tests on web apps, mobile apps and APIs, security code reviews and secure coding training for development teams.

I've been building software for more than ten years, and today I build products with AI agents every day. I know where they go wrong because I use them.

Fastweb Hall of Fame 2020

In 2020 Fastweb added me to its Hall of Fame after a responsible security disclosure.

See the reference on Fastweb

The security review is for you if...

It may not be for you if

The app doesn't exist yet and you want it built for you: in that case have a look at software development. If you want to learn to build with AI and check your app's security yourself, there's my secure vibe coding course (in Italian).

Which path is right for you?

Software development

You want it built for you

You trust me with the design, the development and the technical responsibility of the product.

Explore software development
You are here · Security review

You already have an app to check

I review the code and the configuration, fix the issues and check the app again.

Request a security review
Course (in Italian)

You want to learn to do it yourself

In my secure vibe coding course you learn to build with AI and to check your app's security.

Explore the course

Frequently asked questions

Answers to the most common questions before requesting a security review.

No. The service is designed for apps built with agents such as Claude Code, Codex, Cursor or Lovable, but I also review apps built without AI.

It depends on the size of the app, the stack and the scope of the review. After your application I send you a quote.

It depends on the size of the app. You'll find the timing in the quote.

Access to the repository and a description of how the app is set up: database, hosting and third-party services. We agree on the details before starting.

Yes, I fix them. If you or your team would rather do it, tell me: the report explains how to fix each issue.

Yes. One re-check after the fixes is included, whether I make them or you do.

Yes. If you ask, I sign it before I see the code.

No. I review the code and the configuration from the inside, with access to the repository. I don't simulate an attack on the live app.

The review finds the most important issues and puts them in order, but no review can guarantee that no vulnerability is left.

Want to know how secure your app is?

Tell me what it does, how it was built and what worries you. I'll reply personally, with a proposal or with the reason it isn't the right service.