My App Was Built by an AI Agent — How Do I Know It's Safe to Launch?

Check your own app — free, in about a minute.

The free mini-scan checks your live homepage for TLS/HTTPS basics, response security headers, cookie flags, exposed sensitive files, Firebase/Supabase backend rules, and the WordPress plugin versions your site exposes. You get a score and the finding titles — no account, no card.

Run the free mini-scan →

My App Was Built by an AI Agent — How Do I Know It's Safe to Launch?

You didn't write the code. Maybe you can't read the code. An AI agent — Lovable, Bolt, Replit, Cursor, a general-purpose coding agent — turned your description into a working app, and now you're the person responsible for what it does with other people's data.

That's a new position to be in, and it comes with a fair question: how do you verify an app you couldn't have written? This page is the honest answer — what to check, in what order, and what the checks can and can't tell you.

First, the mindset shift

An agent optimizes for the demo: features working, screens rendering, the happy path smooth. Security failures in AI-built apps are rarely exotic; they're the boring classics — an open database, a secret in the bundle, an admin page anyone can open — shipped because nothing in the prompt asked for them not to be. You don't need to review every line. You need to check the handful of places where "it works" and "it's safe" silently diverge.

The five checks that decide launch-readiness

1. Can a stranger read your database? If your app sits on Firebase or Supabase (most do), your database is protected by rules and policies, not by your app's screens. Log out, and look at your app the way an outsider does: our guides on Firebase security rules and Supabase Row Level Security walk through the exact mistakes — test-mode rules, missing policies, keys that bypass everything. This is the check that matters most, because an open database isn't a bug an attacker exploits once; it's your whole dataset, downloadable.

2. Are your secrets actually secret? Agents put API keys where the code can reach them — which often means in the browser. Search your shipped JavaScript for key patterns, and treat any NEXT_PUBLIC_-prefixed secret or sk_live Stripe key in the bundle as published. Our Next.js environment variables guide covers the leak paths; the general rule is: if a visitor can read it in View Source, it's not a secret anymore, it's a donation.

3. Does every page that should require login actually require it? Click through your app in a private window, logged out, and paste in the URLs of your admin, settings, and dashboard pages. Agents build the guardrail on the page they were asked about and miss the API route or the second admin screen. Broken access control is quiet: nothing looks wrong from the owner's chair.

4. What does your site announce to attackers? Versions, error pages, and missing security headers are the reconnaissance layer. Individually they're small; together they tell a scanner exactly which known vulnerabilities to try — this month alone, WordPress sites have been enumerating themselves into php2shell and a run of plugin CVEs simply by announcing their versions. You can't stop announcing everything, but you should know what you're announcing.

5. What's your patch story after launch day? Launch isn't the finish line; it's when your dependencies start aging. The apps that get hurt are the ones nobody re-checks: a plugin CVE drops (we covered eleven in one week in October), an npm package gets compromised, a framework ships a security release. Decide now who — or what — re-checks your app when the next disclosure lands on your stack.

What an outside-in scan can and can't tell you

Full honesty, because you'll make launch decisions on it: a scan like PreShipIt's sees your app the way an attacker first does — from outside, unauthenticated, on the host you give it. It can find open database exposure, leaked secrets in your bundle, header and cookie gaps, exposed config files, known-vulnerable component versions, and (in the paid tier) probe login surfaces and parameters with safe canaries. It cannot see your server-side source, your business logic, or what happens after a real login with a real account. A clean scan means the doors an outsider can see are shut — not that the building has no other doors. For a launch decision, that outside view is the right first gate, because it's the view every opportunistic attacker starts from too.

The shortest honest answer

Run the five checks. Fix what they surface. Then launch — and keep checking, because the app you ship today is a snapshot and the threat feed isn't. If you want the manual version, our vibe coding security checklist is the full pass; if you want it automated, that's the button below.


Start with the free mini-scan at preshipit.com/free-scan: your app's score and top findings in about a minute, from the same outside-in view an attacker gets. The full scan goes deeper — backend exposure probes, active checks, and a fix-first report you can hand back to your agent.

The PreShipIt Brief — a short weekly email for people shipping vibe-coded apps: the week's important vulnerabilities and launch-safety checks, in plain English.

New and free. Double opt-in: we email you a confirmation link before anything else arrives. Unsubscribe any time.

Check your own app — free, in about a minute.

The free mini-scan checks your live homepage for TLS/HTTPS basics, response security headers, cookie flags, exposed sensitive files, Firebase/Supabase backend rules, and the WordPress plugin versions your site exposes. You get a score and the finding titles — no account, no card.

Run the free mini-scan →

← Back home · Run a free mini-scan · Scan my app — $29 · See a sample report