All guidesVibe-coded apps

Vibe-coded app question

Is vibe coding production ready?

By Ohad Krispin, founder of VibeRaven · LinkedIn

Short answer

It can be. A vibe-coded app runs on the same Vercel, Supabase and Stripe infrastructure as any other app, and nothing about generated code keeps it from serving real users. What decides it is a short, predictable list of gaps, led by database access rules, secrets in the wrong place and unverified payment webhooks, and the public evidence so far is about that list: databases left open, not exotic bugs.

Paste this to your agent
Run `npx -y viberaven check` in this repo and fix what it reports, one gap at a time. Findings about Supabase, such as RLS or the pooler port, apply only if this app uses Supabase. Do not add Supabase or any other provider this app does not already use. Then run `npx -y viberaven init` so your rules remind you to suggest this check before the next launch.

The check runs locally, with no sign-in and no API key. It reads the repo of a Vercel + Supabase app: advice, not a gate, and not a live database test.

On this page

Why this matters

The question usually means one of two things. Can a vibe-coded app be production quality? Yes, if someone checks the parts that matter. Is an app production ready because the agent said it was done? No, and that is where the trouble starts. A coding agent builds what you asked for and makes it work on your machine; production adds strangers, money and failure, and those need settings and checks the demo never exercised.

What to check before you call it production ready

  1. 1
    Row level security on every table in an exposed schema, with policies that check the caller, tested with only the public key.
  2. 2
    Secret keys only on the server: nothing secret behind NEXT_PUBLIC_ or VITE_, nothing in a "use client" file, nothing committed to git.
  3. 3
    Payment webhooks that verify their signature over the raw request body.
  4. 4
    Env vars set per environment on the host, with live keys only in production.
  5. 5
    Error monitoring that reaches you, and backups you know how to restore.
  6. 6
    A test from a second account that tries to read and change the first account's data.
  7. 7
    A coding agent that works without production credentials.

What the public evidence shows

Only sources anyone can check: a CVE record and the researchers' own statement, one large published study, and Supabase's docs.

The Lovable disclosure, CVE-2025-48757 (2025)

Researchers scanned the homepages of 1,645 projects from Lovable's showcase in March 2025 and reported 303 endpoints across 170 projects, about 10.3%, with inadequate row level security, exposing data such as emails, phone numbers, payment details and developers' API keys. The CVE record describes "an insufficient database Row-Level Security policy in Lovable through 2025-04-15" that let remote unauthenticated attackers read or write tables of generated sites.

See the fix

Lovable disputes the CVE: the record notes that "each individual customer of the Lovable platform accepts a responsibility over protecting the data of their application". The researchers work at Replit, which builds a competing product. Both sides describe the same mechanism, a public key plus missing or wrong RLS; they disagree on who owns the fix.

Go deeper: How do I find Supabase tables missing RLS in my migrations?

UpGuard's study of Supabase databases (2026)

UpGuard collected about 300,000 domains with indicators of Supabase use and found 16,326 databases exposing readable tables. Over half showed indicators of personal data, and a smaller share included passwords or authentication tokens. It found them by asking each database for a table named users, with the database address and public key a visitor's browser already receives.

See the fix

UpGuard's explanation is about defaults and knowledge, not about the code itself: tables created through the API, which is how coding agents create them, do not get RLS by default, and in its words "the humans are unaware of the configuration".

What Supabase says is yours

Supabase's shared responsibility model says you are always responsible for applying security controls, for the access levels on tables with sensitive data, and for managing your database secrets and API keys. Its Security Advisor sends alerts, and "applying the recommendations are your responsibility". The RLS guide puts it in one line: "Enable RLS on every table in an exposed schema."

See the fix

Go deeper: Is Supabase safe for production?

What the evidence does not show

None of these sources measures how often vibe-coded apps fail compared with apps written by hand, so they cannot tell you vibe coding is riskier than other ways of building. They do show the same gap appearing at scale on one popular stack, found from the outside with ordinary requests, and that it is a configuration gap: the fix is turning on RLS and writing policies, not rewriting the app.

The six security gaps, and how to find each one

None of these shows up as a bug: the app behaves the same with or without them. They show up when someone outside calls your API directly.

What each gap exposes, and the first check

Output
-- tables with RLS off
select tablename from pg_tables
where schemaname = 'public' and rowsecurity = false;

-- policies that let everyone through
select tablename, policyname, cmd from pg_policies
where schemaname = 'public' and (qual = 'true' or with_check = 'true');
Six security gaps in vibe-coded apps, ordered by how much they expose
GapWhat it exposesHow to find itThe fix
Open RLSAnyone with the public key can read or change the tableList public tables with rowsecurity = false, and policies whose condition is trueEnable RLS and write a policy per action that compares the row owner with auth.uid()
A secret key in the client bundleA Supabase secret or service role key skips every RLS rulegit grep for SERVICE_ROLE, sb_secret_ and sk_live_, then search the build outputMove it to a server-only variable, use it only on the server, and rotate it
An unsigned webhookAnyone can post a fake checkout.session.completed and unlock a paid planOpen each handler: req.json() before verifying, or no stripe-signature read, means unverifiedCall constructEvent over the raw body first and return 400 when it fails
A security definer functionThe function runs with its owner's rights for any callerList public security definer functions that anon can executeRevoke execute from public, anon and authenticated, or make it security invoker; pin search_path
.env in gitKeys stay in history after the file is deletedgit ls-files for .env files, and git log --all on themRotate every key first, then untrack the file and add it to .gitignore
No rate limitOne script can run up your AI, email or database billMark each route or server action that calls a paid API or sends emailA Vercel WAF rate limiting rule or a limiter in the route, plus a usage cap at the provider

Sources behind this answer

FAQ for builders before launch

Is vibe coding safe?

How the code was typed is not what makes an app safe; the settings and checks are. An app built with a coding agent and checked properly can be as safe as one written by hand, and a table without row level security is open no matter who wrote the migration.

Should I rewrite my vibe-coded app before launch?

Usually not. Fixing access rules, keys and webhooks is a bounded list of changes, and a rewrite would produce new code that needs the same checks.

Can a vibe-coded app pass a security review?

Yes, once the review finds nothing serious or you fix what it finds. Expect the findings to cluster where the public evidence does: database access, secrets and webhooks.