Skip to content

Cursor scanner

The code is yoursnow the review is too

An editor that writes large blocks of your app changes who reviews what. Nobody is reading every generated file the way you read your own. This scans the two things that are actually yours: the URL the code ended up on, and the repository it came from.

  • Site scan + repo scan
  • No agent
  • 85 checks in every scan

Why this shape of app

Two halves,
one review

A generated file is not unsafe because a model wrote it. It is unsafe for the same reason a copied file is: nobody re-read it on the way to production. The scan reads both places a copy can land — what the browser gets, and what the repository holds.

  • What the bundle gives away

    The client build is the last place a secret can hide and the first place a stranger looks. The scan reads the JavaScript your app ships for keys that were meant to stay server-side — and it knows the difference between a publishable key that belongs there and a service-role key that does not.

  • Sinks the generator did not mean

    Inner HTML on a value from the URL. A query assembled by string concatenation. A redirect that trusts its parameter. These are the shapes a language model reproduces because they appear in a hundred tutorials, and each one is a named check rather than a hunch.

  • The dependencies arrived uninvited

    A generated app installs what it needs and never removes it. The repository scan lists every package, ranks the vulnerable ones against known-exploited listings rather than raw CVE counts, and tells you whether the dependency is direct or three levels deep — which decides whether you can even fix it.

  • Nothing recorded who did what

    The check most teams discover late: whether anything in the stack is logging. An app assembled in a week rarely has an audit trail, and the incident that follows is reconstructed from the attacker’s goodwill. There is a check for that, and a repository-side look at your infrastructure definitions too.

What runs on your URL

The checks that
bite hardest here

A shortlist, not a special mode: paste the URL and the whole scan runs. These are the checks that most often have something to say about code you did not type yourself.

  • Content-Security-Policy
  • Cookie security flags
  • CORS policy
  • Exposed sensitive files
  • Secrets in JavaScript
  • Source maps
  • Audit logging & monitoring exposure
  • Open redirect through URL parameters
  • SQL injection in query parameters
  • DOM-based XSS sinks fed by URL input
  • JSON Web Tokens published to anonymous visitors
  • Tech stack & CVE

Every one of these runs on every scan, whatever built your app. What each check looks at is on the full list.

Generated code is not riskier than copied code. It is the same code, unreviewed.

Stated plainly

What this page
is not claiming

  • There is no Cursor check in the scanner, and there could not be — Cursor is an editor, it does not host your app or touch your deployment. What runs is the same 85 checks every site gets, plus the repository scan if you install the GitHub App.

  • The repository scan reads your code; the site scan does not. A secret that never made it into the build, a dependency behind a private registry, a policy file that is not deployed — each is visible from one side and invisible from the other. Running both is the point.

  • It is not a penetration test and not a code review. It never logs in, never submits anything of yours, never calls a state-changing endpoint, and refuses private and metadata addresses before a request leaves. It will not tell you the architecture is wrong.

  • A clean scan is a photograph. It is 85 true statements about one URL on one day — and the next merge is the one that would have needed it.

How the scan runs

Three steps,
one URL

No agent to install, no DNS change, no access to your account. The scan reads the site the way a stranger does.

  1. Step 1

    Paste the URL you shipped

    Wherever Cursor left it — your own host, a platform subdomain, a staging box you forgot was public.

  2. Step 2

    Install the GitHub App for the repo

    A short-lived installation token, a clone, and pinned open-source scanners whose downloads are checked against upstream checksums before the image builds. Secrets, dependency vulnerabilities, license policy and an SBOM out the other end.

  3. Step 3

    Take the evidence and the prompt

    Findings carry what was measured — the file, the version, the header — with a prompt written for the same agent that wrote the code, so the fix goes back the way the problem came.

Read it before
your customers do

The URL gets the 85-check scan; the repository gets secrets, dependency vulnerabilities, license policy and an SBOM. Free account, no card.

Scan your site