Cursor apps can be anything
Cursor is a code editor, not a hosting platform. Your app might be Next.js on Vercel, a Vite front end with Supabase, a Python API on a VPS, or something else. So we can't tell from the outside that Cursor wrote it, and we don't need to: Securick checks the live site that's deployed, with the same tests for every stack.
What Cursor-built apps have in common is speed. Code is generated quickly, accepted quickly and deployed quickly. The mistakes below are the ones that slip through when nobody reads the build output.
Mistakes we see in fast-built apps
Secrets behind a "public" prefix
Frameworks decide which environment variables reach the browser by their name. In Next.js only variables starting with NEXT_PUBLIC_ are sent to the browser, and their values are written into the JavaScript at build time (Next.js docs). In Vite the prefix is VITE_ (Vite docs). Give a secret one of these prefixes to fix an error, and it ships to every visitor.
Deploying the whole project folder
When an app is copied to a server as a folder, files like .env, .git and backups can go with it. A public .git folder can let someone download your source code and its history.
Error pages that say too much
A development error page or stack trace in production tells an attacker your framework, file paths and sometimes query details.
CORS opened up to make a bug go away
"Allow all origins" fixes a CORS error in seconds. It also lets any website call your API from a visitor's browser, and with cookies that can mean acting as your users.
Old framework versions
Generated code often pins whatever version the model remembered. If that version of React or Next.js has known flaws, so does your app.
Tell Cursor your rules once
Cursor supports project rules: instructions stored in your repository under .cursor/rules that the AI follows (Cursor docs). A rule such as "never put secret keys in client code or in NEXT_PUBLIC_/VITE_ variables" won't catch everything, but it stops the most common leak before it's written. Then check the deployed result.
What Securick tests on a Cursor-built app
All from our list of tests. Free runs 46, Pro runs 130.
- #10 Secret values in public env variables Free
- #11 .env file public Free
- #12 .git folder public Free
- #13 JavaScript source maps public Free
- #14 Backup files public Free
- #17 Debug or info pages public Free
- #18 Config files public Free
- #31 Error pages hide code details Free
- #29 Not open to every website Free
- #30 Not open to every website with login cookies Free
- #71 React / Next.js with known flaws Pro
- #50 API docs or GraphQL schema open to public Pro
- #23 Content-Security-Policy present Free
- #19 HTTP redirects to HTTPS Free
On a domain you verify, Deep tests check what signed-in users and your API allow:
- #96 Data isolation between users Deep
- #100 Forms and inputs handle unsafe input safely Deep
- #101 Edge/serverless functions require login where needed Deep
- #148 Users can't make themselves admin Deep
- #151 API rate limits Deep
What a finding looks like
Your .git folder can be downloaded
Fix prompt to paste into your builder
Stop serving the .git folder. Deploy only the build output (for example the dist or .next folder) instead of the whole project directory, or add a server rule that blocks every path starting with a dot. Then assume the source code is public: check the git history for secrets and rotate any you find.
Free, Pro and Deep
Free
46 tests from your public URL, 2 scans a day, no sign-up. Every finding with proof.
Pro · $10 one time
30 scans within 30 days: all 130 tests, with a fix prompt for every finding and PDF reports.
Deep · part of Pro
17 more tests (147 in total) that touch your database and signed-in users, only on a domain you verify.
Questions
Does Securick read my code in Cursor?
No. It never sees your editor or repository unless you connect one. The scan reads the deployed site from its public URL, like any visitor's browser would.
My app is a backend API, not a website. Can I scan it?
If it has a public HTTPS address, yes. Tests that look at pages and JavaScript won't find much on an API, and the report marks what didn't apply instead of counting it as a pass.
Which frameworks does it understand?
The tests read what's served, so they work for any framework. Some look for specific things, like NEXT_PUBLIC_ and VITE_ variables or known flaws in React and Next.js versions.
Can I run it on every deploy?
You can re-scan as often as your plan allows: the free scan 2 times a day, Pro up to 30 scans within 30 days. Scanning before each release catches what the last change exposed.