VTUse cases

Different builders. The same blind spot.

AI changed who can build software and how fast it ships. It didn’t change what attackers look for. Here’s how VibeTect fits the way you work.

01Founders & solo builders

You shipped an app with an AI builder. Now real people are signing up.

Tools like Lovable, Bolt, Replit and v0 made it possible to go from idea to product in days. They’re very good at producing an app that works. They don’t guarantee the database is locked down, or that keys meant for the server stayed on the server.

Risks you’re likely carrying

  • Database tables readable by anyone holding the public key
  • Admin or service keys reachable from the browser
  • Users able to view or edit each other’s records

How VibeTect helps

  • Inspect the exported or synced project on your own computer
  • Findings explained in plain English, ranked by urgency
  • Fix guidance you can hand straight back to your AI builder
supabase/migrations/0001_profiles.sqlTypical first finding
create table public.profiles (  id uuid primary key references auth.users,  full_name text,  phone text,  stripe_customer_id text);
FlaggedThis migration never enables Row Level Security. In a typical Supabase setup, that leaves the table reachable through the auto-generated API with the anon key — which ships in your frontend.
Supabase Row Level Security: what AI app builders get wrong

02Developers using AI assistants

Your assistant writes a growing share of every pull request.

Cursor, Copilot, Claude Code and similar tools make experienced developers dramatically faster. Review attention doesn’t scale the same way: you check that the logic is right, and subtle security mistakes slide through because the code looks idiomatic.

Risks you’re likely carrying

  • Tokens decoded but never verified
  • Ownership checks missing from new endpoints
  • Loosened CORS or disabled checks left behind after debugging

How VibeTect helps

  • A fast local pass before you push, on the code as it is right now
  • Technical detail — CWE references, exact locations — without the noise
  • No repository app to install or permissions to request from your org
middleware/auth.tsTypical first finding
export function requireUser(req: Req, res: Res, next: Next) {  const token = req.cookies.session;  const user = jwt.decode(token) as SessionUser | null;  if (!user) return res.status(401).end();  req.user = user;  next();}
Flaggeddecode() doesn’t verify the signature. Anyone can craft a cookie that claims to be any user.
Broken access control: the bug AI writes most confidently

03Startups & small product teams

The codebase is growing faster than anyone can read it.

When several people pair with AI assistants, the volume of new code outpaces code review. You don’t have a security team, and a quarterly audit can’t keep up with weekly releases. You need a consistent check that anyone on the team can run.

Risks you’re likely carrying

  • Admin features protected only by hiding them in the UI
  • Secrets copied between environments and committed by accident
  • Inconsistent authorization across endpoints written by different people

How VibeTect helps

  • The same inspection on every machine, before every release
  • Severity ranking that makes triage a five-minute conversation
  • Findings specific enough to assign and verify
app/api/admin/users/route.tsTypical first finding
export async function DELETE(req: Request) {  const session = await getSession(req);  if (!session) return new Response(null, { status: 401 });   const { userId } = await req.json();  await db.user.delete({ where: { id: userId } });  return new Response(null, { status: 204 });}
FlaggedThe route checks that someone is signed in — not that they’re an admin. Any user can delete any account.
The vibe coding security checklist

04Agencies & freelancers

You deliver apps that hold your clients’ customers’ data.

AI lets small studios ship more projects, faster. Each one carries your reputation. Clients rarely ask about security until something goes wrong — and many agreements don’t allow you to upload their code to third-party services.

Risks you’re likely carrying

  • Debug settings and verbose errors left on in production configuration
  • Shared credentials embedded in a template you reuse across clients
  • Client data exposed through permissive storage or database rules

How VibeTect helps

  • Check client code on your machine, without sending it anywhere
  • A repeatable pre-handover step for every project
  • Clear findings you can explain to a non-technical client
settings.pyTypical first finding
DEBUG = TrueALLOWED_HOSTS = ["*"] SECRET_KEY = "django-insecure-"CORS_ALLOW_ALL_ORIGINS = True
FlaggedDebug pages expose stack traces and settings, the secret key is hard-coded, and every origin is allowed. Development defaults shipped to production.
How AI-built apps leak API keys

When VibeTect isn’t enough on its own

If you need a formal penetration test, a compliance attestation or a security sign-off for a regulated system, VibeTect isn’t a substitute. It is a sensible step before one: clearing the common, avoidable issues first means expert time goes on the problems only experts can find.

Whatever you build, check it before you ship it.

Run a local security inspection on the project you’re about to deploy — and fix what it finds while it’s still cheap to fix.

Local analysis · No repository access required