GUIDES

    Vibe Coding Security Checklist: 12 Checks To Run Before You Launch

    A plain-English vibe coding security checklist for Lovable, Bolt, Replit and Cursor apps. Run these 12 checks before launch and fix what you find.

    LetMeCheck team · 4 min read
    Vibe Coding Security Checklist: 12 Checks To Run Before You Launch
    IN THIS POST

    Your AI-built app works. Before real users arrive, here are 12 checks that catch the most common security holes, written for people who do not read code every day.

    You prompted, the AI built, and the app works. Sign-ups save, the dashboard loads and the preview looks great. Then a friend asks, "Is it safe to put real users on this?" and the room goes quiet.

    Here is a calmer answer. This vibe coding security checklist has 12 checks, grouped into three areas, for apps built with Lovable, Bolt, Replit, Cursor or Claude Code. You do not need to be a developer to run most of them. Block out an afternoon, work from top to bottom and you will know where you stand.

    Why AI-Built Apps Need A Second Look

    AI builders are very good at making things work. Locking things down is a different job, and unless you ask for it, it often gets skipped. Most problems are not clever attacks. They are doors left open by default: a key pasted into the wrong file, a database table anyone can read, a page that hides its data behind a login screen but not behind a login check.

    Several of these checks mirror guidance the platforms publish themselves. Replit's security checklist, for example, tells builders to keep API credentials on the server, check permissions on every action and protect every endpoint. If you want a head start, our guide to running your first scan shows how to check a live URL with no sign-up, and you can use this list to dig into whatever it finds.

    Lock Down Your Keys And Database (Checks 1 To 4)

    1. Keep secret keys out of browser code. Anything the browser downloads can be read by anyone with developer tools open. Search your front-end files for words like key, secret and token. If a real secret turns up, move it to server-side code and replace the key. Our glossary entry on API key exposure explains the risk.
    2. Make sure your .env file is not on the web. Type yoursite.com/.env into a browser. Anything other than a not-found page means your secrets are public, and every value in the file needs replacing.
    3. Switch on row level security for every table. If you use Supabase, each table needs RLS enabled and a policy that matches who should see the data. Without it, the public key in your app can read and edit rows. Here is what row level security means in practice.
    4. Keep the service role key on the server. This key skips every database rule. If it appears in browser code, treat it as leaked. Your front end should only ever use the public key.

    Protect Logins And User Data (Checks 5 To 8)

    1. Check logins on every private endpoint. Hiding a page behind a sign-in screen is not enough if the address that returns the data answers anyone. Call your API addresses while signed out and see what comes back. This gap is common in Cursor-built apps, where routes are generated without checks.
    2. Test that users only see their own data. Create two accounts, add records to the first, then try to open them from the second by changing the number in the address bar. If it works, you have an insecure direct object reference, one of the most common forms of broken access control.
    3. Use proper sign-in and slow down guessing. Rely on the sign-in system your platform provides instead of one the AI invented, and confirm repeated wrong passwords get blocked.
    4. Verify admin rights on the server. Admin pages should check for an admin role on the server, not just hide the button. Anything a user can edit in their own browser is no place to decide permissions.

    Harden Your Code And Launch Setup (Checks 9 To 12)

    1. Use HTTPS and basic security headers. Your site should load over HTTPS only and send headers such as X-Frame-Options and X-Content-Type-Options. Most hosts make this a setting, not a coding job.
    2. Update and review your dependencies. AI tools sometimes suggest outdated packages, or names that do not exist and that attackers can claim. Run an audit and remove anything you do not recognise. Our write-up of the September 2025 npm attacks shows how this goes wrong.
    3. Validate everything users type. Every form field, search box and URL parameter is a way in. Validate on the server and use parameterised queries, which closes off SQL injection and cross-site scripting.
    4. Set spending caps and friendly errors. If your app calls a paid AI or email service, set a hard monthly cap so a leaked key or a bot cannot run up a bill. Make sure error pages show users a plain message, not database details.

    The 12 Checks At A Glance

    \#
    Check
    Quick Test
    1
    No secrets in front-end code
    Search browser files for "key", "secret", "token"
    2
    No exposed .env file
    Visit /.env on your live site
    3
    RLS on every table
    Look for warnings in the Supabase table list
    4
    Service role key server-side only
    Search front-end code for "service_role"
    5
    Endpoints check login
    Call API addresses while signed out
    6
    Users only see their data
    Use two accounts and swap record IDs
    7
    Sign-in is hardened
    Try repeated wrong passwords
    8
    Admin checked on the server
    Open admin routes as a normal user
    9
    HTTPS and headers
    Load the http:// version of your site
    10
    Dependencies current and real
    Run an audit and review the list
    11
    Input validated
    Paste odd characters into every form
    12
    Spending caps and clean errors
    Check the provider dashboard and trigger an error

    Fix What You Find, Then Check Again

    Do not panic and do not rebuild. Fix the most serious item first, which is almost always a leaked key or an open table, and then retest. The AI tool that built the app can usually write the fix if you describe the problem precisely. Our guide to reading a scan report explains how issues are ranked and how to use the fix prompts.

    If one item is beyond you, an hour of a developer's time is cheap compared with a leak. Run through this list once before launch and again after big changes. It gets faster each round, and your users will never know how close it was.

    References

    Does your app have any of these problems?

    Paste your website for a free quick check. About a minute, and nothing on your site is changed.