Is My Vibe-Coded Game Safe? A Free 10-Minute Security Self-Check
You don't need to be a security expert to catch the worst problems in an AI-built game. Set a 10-minute timer and work through these five checks.
This is a quick triage, not a full audit. Each check takes about two minutes and targets a mistake that shows up often in vibe-coded projects. If one fails, follow the linked guide for the full fix. For a longer pass before launch, use our full pre-launch checklist.
Check 1: Leaked keys in your repo and your build (2 min)
Scan both your git history and the files you actually ship. The free, open-source gitleaks scanner does both:
# Every commit in the repo, including deleted files
gitleaks git -v .
# The exported build folder (web bundle, export directory, etc.)
gitleaks dir -v ./dist
Then do a manual pass on the build, since that's what players download. For a web game, open it in the browser, press F12 and search the Sources tab for sk_live_, service_role, secret, api_key and token.
Fails if: you find a payment secret key, a database admin key, an AI API key, a Steam publisher key or a bot token. Fix: rotate the key at the provider first (deleting it from code isn't enough), then move the call to a server. See Steam keys and API keys in Unity. (Public-by-design keys like a Firebase web config or a Supabase anon/publishable key are fine to ship; check 2 is what protects them.)
Check 2: Is your database open to everyone? (2 min)
- Firebase: open the Rules tab for Firestore, Realtime Database and Storage. Firebase's own insecure rules guide warns against
allow read, write: if true(or".read": true, ".write": true), and against rules that only checkrequest.auth != null, which let any logged-in player touch everyone's data. - Supabase: open the Security Advisor and check that every table in an exposed schema (usually
public) has RLS enabled. Supabase's RLS docs are blunt: a table in an exposed schema without RLS is readable and writable by any role with a grant on it. Also look for write policies that just sayusing (true).
Quick attack test: sign in as one test player and try to read or change another player's data from the browser console. If it works, it fails. Fix: Firebase and Supabase rules for game data.
Check 3: Does your game trust the client? (2 min)
Search your code for places where the client tells the server, or the database, what happened:
grep -rnE "score|coins|gold|xp|damage|health|inventory|isPremium|unlocked" \
--include=*.js --include=*.ts --include=*.gd --include=*.cs src/ scripts/ 2>/dev/null
Look at each hit and ask: could a player change this number before it reaches the server? Red flags are a client sending score: 12000, an RPC that applies damage passed in by the caller, or unlocks and currency stored in localStorage or cookies.
Fix: the client sends inputs and requests, and the server decides outcomes. See Never trust the client, Godot 4 RPC security and protecting browser game saves.
Check 4: Can someone get paid items without paying? (2 min)
Find the code that unlocks a purchased item. It passes only if the unlock happens on your server after the payment provider confirms it: a Stripe webhook with signature verification, or server-side receipt validation for app stores.
Fails if: the item is granted when the player lands on a /success page, when the client says “purchase complete”, when the price comes from the client, or when ownership is just a flag in local storage. Try it yourself: open the success URL directly without paying. Fix: server-side entitlements and verified payments.
Check 5: Exposed admin and debug endpoints (2 min)
AI tools often leave behind helper routes, cheat commands and dashboards built to test things. Search for them:
grep -rniE "admin|debug|/test|cheat|godmode|give_item|reset_db|seed" \
--include=*.js --include=*.ts --include=*.gd --include=*.cs --include=*.py . 2>/dev/null | grep -v node_modules
Then, logged out, try visiting /admin, /api/admin, /debug and any route you found, on your live site or game server.
Fails if: any of them respond without a login, or if a debug or console command that grants items can be triggered in a release build or over the network. Fix: remove them from production builds, or put them behind real authentication with an admin role checked on the server. Also make sure database ports and admin panels on your game server aren't reachable from the internet.
Scorecard
- 5/5 passed: good start. Do the full checklist before launch.
- Failed check 1 or 4: fix today. Leaked secrets and free purchases cost real money.
- Failed check 2 or 5: fix before you share the game. Open databases and admin routes are found by automated scanners.
- Failed check 3: fix before you add leaderboards, trading or PvP.
Prompt to paste into your AI coding tool
Run a security self-check on this game project and report pass/fail for each:
1) secrets in code, config, build output or git history;
2) database rules or RLS that allow public or cross-player reads/writes;
3) server or database code that trusts client-sent scores, currency, damage or unlocks;
4) purchases granted without server-side, signature-verified payment confirmation;
5) admin, debug or test endpoints and cheat commands reachable in production.
For each failure give the file, line and a concrete fix. Don't change code yet.
Want this checked automatically?
BossCheck is a security scanner for indie games built with AI tools. It's in pre-launch: join the waitlist for early access.
Join the waitlist →FAQ
How do I know if my vibe-coded game is secure?
Run a quick self-check: scan your repo and build for leaked keys, confirm your Firebase rules or Supabase RLS block cross-player access, make sure the server decides scores and currency, verify purchases server-side, and remove admin or debug endpoints from production.
What's the most dangerous security mistake in an AI-built game?
Leaked secret keys and unverified purchases cost money fastest, while an open database can expose or wipe every player's data. All three are common in vibe-coded projects.
Is a free scanner like gitleaks enough?
It's a good start for finding leaked secrets, but it won't catch open database rules, client-trusting game logic or unprotected purchase flows. You need to check those separately.
Is my Firebase API key or Supabase anon key a leak?
No. Both are designed to be public. What matters is that your Firebase Security Rules or Supabase RLS policies and grants restrict what those keys can do.
How long does this self-check take?
About 10 minutes for a small project: roughly two minutes per check. Fixing what you find can take longer.
More guides
- How to Keep Your Steam Web API Key Out of Your Game Build
- Firebase and Supabase Security Rules for Multiplayer Games
- Never Trust the Client: Stopping Cheaters in Vibe-Coded Multiplayer Games
- Security Checklist Before Shipping a Game Built with Cursor, Claude or Lovable
- Is Code from Cursor, Lovable or Claude Secure? What to Check Before You Ship
- How to Protect API Keys in a Unity Game
- How to Stop Players Editing Browser Game Saves to Unlock Paid Items
- Securing Godot 4 Multiplayer: RPCs, Authority and Server-Side Validation
- Stop Bots and Scripts from Breaking Your Game Economy Before a Steam Launch