More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Launch Guard Lite: Free Leaked API Key Check

    0

    Bots scan new websites and public code for API keys around the clock.

    Free

    Security scanned
    Launch Guard Lite: Free Leaked API Key Check

    Launch Guard Lite: Free Leaked API Key Check

    Example session with this skill installed

    I'm about to launch my app and I'm worried I might have leaked some API keys. Can you check? I'm not technical, so please keep it simple.

    • Read your context and instructions
    • Compiled the launch guard lite
    • Generated the app files

    Leaked key check: InvoiceFlow

    2026-11-03 · 35 files checked · Launch Guard Lite v1.0

    Result: LEAKS FOUND (3)

    None of your API keys is sent to visitors' browsers. But your database password and your login secret are typed straight into your code. Your live Stripe key, your Anthropic (AI) key and five other secrets are in a file that would be uploaded along with your code. First thing to do: change your database password in AWS and take it out of lib/db.ts.

    Fix now

    1. Your database password is typed into your code · HIGH

    Where: lib/db.ts:9 (postgresql://invoiceflow_admin:****@…rds.amazonaws.com:5432/invoiceflow), also .env.local:2
    What's wrong: lib/db.ts keeps a backup copy of your full database address, with the admin password in it. Code files go wherever your code goes: GitHub, AI tools, anyone you share it with. Passwords belong only in an environment variable, which is a private setting stored on your server, outside the code.
    What could happen: Anyone who has seen the code could log straight into your database, if it accepts connections from the internet. They could then read, change or delete every account, client and invoice without going through your app.
    How to fix:

    1. Change the password in the AWS Console: RDS → Databases → invoiceflow-db → Modify → New master password. Put the new address in DATABASE_URL on your hosting provider and in .env.local.
    2. Delete the backup address from lib/db.ts so the app reads only DATABASE_URL. Then redeploy.

    2. Your login secret is typed into your code · HIGH

    Where: lib/auth.ts:26 (secret: 'LN…A='), also .env.local:6 (NEXTAUTH_SECRET)
    What's wrong: This value (the NextAuth secret) is what proves each login is genuine. It is written directly in the code instead of being kept only in an environment variable.
    What could happen: Anyone who has seen the code could create a working login for any account, including an admin account, without knowing the password. Your admin area lists every user.
    How to fix (do both steps together):

    1. Create a new random secret and save it as NEXTAUTH_SECRET on your hosting provider and in .env.local. Everyone will be signed out once; that's expected.
    2. Delete the secret: '…' line in lib/auth.ts. NextAuth reads NEXTAUTH_SECRET on its own. Then redeploy.

    3. Your live Stripe key, AI key and 5 other secrets aren't protected from upload · HIGH

    Where: .env.local, plus .gitignore:29, which only blocks a file named exactly .env
    Secrets in the file: STRIPE_SECRET_KEY=sk_live_51…TC, STRIPE_WEBHOOK_SECRET=whsec_We…NI, ANTHROPIC_API_KEY=sk-ant-api03-KM…AA, RESEND_API_KEY=re_WO…Op, GOOGLE_CLIENT_SECRET=GO…Hu, plus the database password and login secret above.
    What's wrong: Git is the tool that saves your code and uploads it to sites like GitHub. .gitignore is the list of files git must never upload, and yours doesn't include .env.local.
    What could happen: The first time this code is uploaded (or an AI builder syncs it to GitHub), all seven secrets go with it.

    • With your live Stripe key, someone could refund payments, read customer data and create charges.
    • Your Anthropic key could be used on your bill. Bots usually find leaked AI keys within hours.
    • Your Resend key could be used to send spam from your domain.

    How to fix:

    1. Add .env* and !.env.example to .gitignore.
    2. If this code has ever been uploaded or shared (GitHub, an AI builder, a zip sent to someone), replace every key in the file:

    Stripe: Dashboard → Developers → API keys → Roll key. Then Developers → Webhooks → your endpoint → Roll secret.

    • Anthropic: console.anthropic.com → API keys → delete the key and create a new one.
    • Resend: Dashboard → API Keys → delete the key and create a new one.
    • Google: Google Cloud → APIs & Services → Credentials → your OAuth client → reset the secret.

    Then put the new values on your hosting provider and redeploy.

    Also worth fixing

    • Your public invoice page sends each freelancer's password hash to the visitor's browser · app/invoice/[id]/page.tsx:23 and :131. A password hash is a scrambled copy of a password. The page loads the freelancer's whole account (password hash, Stripe IDs and role) and passes all of it to the Pay button, which runs in the visitor's browser. Anyone with an invoice link can read it and try to crack weak passwords. It isn't an API key, but it is a secret that reaches browsers.

    Fix: load only the name, business name and email, and give the Pay button only the invoice ID, number, total and currency.

    Checked and fine

    • The Stripe publishable key (pk_live_… in components/PayInvoiceButton.tsx:8) and the PostHog key (phc_… in components/Providers.tsx:10) are meant to be in the browser.
    • The Google client ID and the Stripe price ID are not secrets.
    • Your Stripe secret, Anthropic, Resend and Google keys are only used in server code (lib/stripe.ts, app/api/ai/draft/route.ts, app/actions/invoices.ts, lib/auth.ts). None of them is sent to browsers.
    • No secret uses the public prefix NEXT_PUBLIC_ (a variable name that Next.js copies into the website). next.config.js doesn't copy any settings into the site either.
    • .env.example contains only placeholders.
    • The folder has no built copy of the site, so there were no downloadable files to check. The folder also isn't a git repository on its own, so there was no upload history to check.

    Not covered by this check

    • This was a leaked-key check only. It did not check payment webhooks, AI endpoints anyone can call, ownership checks or launch basics. The full pre-launch audit covering those is Launch Guard on Agensi.
    • While reading the code, I noticed several serious problems in those areas. I didn't check them thoroughly, so this list may not be complete. They should be fixed before launch:

    Anyone can fake a "payment received" message. A webhook is the message Stripe sends your app when someone pays. Yours (app/api/webhooks/stripe/route.ts:10) never checks that the message really came from Stripe, so someone could mark invoices as paid or give themselves Pro for free.

    Anyone can sign up as an admin. Sign-up copies every field the browser sends into the new account, including the role (app/api/register/route.ts:32).

    Anyone can use your AI on your bill. The AI drafting route (app/api/ai/draft/route.ts) has no login check. That is the same money risk as a leaked AI key.

    Clients can choose how much to pay. Checkout takes the amount from the browser (app/api/checkout/route.ts:39) instead of from the invoice.

    • Any logged-in user can delete any invoice if they know its ID (app/actions/invoices.ts:78).

    Invoice notes are shown as raw web code (app/invoice/[id]/page.tsx:124), so someone could plant harmful scripts on invoice pages.

    Next step

    I didn't change anything in your app. When you're ready, reply "fix the leaks" and I'll fix them one at a time, asking before each change, then run the check again.

    page.tsx

    ZIP · project files

    Generated

    Example file from a real run - the skill writes it into your workspace.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Can strangers get your secret keys?

    Bots scan new websites and public code for API keys around the clock. A leaked OpenAI, Gemini or Anthropic key gets used on your bill within hours. A leaked Stripe or Supabase service key exposes your customers and your money.

    Launch Guard Lite checks your app for leaked keys. It then tells you, in plain English, exactly what to revoke and how to fix it.

    What it checks

    • Keys written into your code. 20+ formats: OpenAI, Anthropic, Gemini, Stripe, Supabase service role, AWS, GitHub, SendGrid, Resend, private keys and Google service-account files.
    • Secrets behind public variable names. VITE_, NEXT_PUBLIC_, REACT_APP_, EXPO_PUBLIC_ and others are copied into every visitor's browser.
    • .env files that git will upload (or already has), because .gitignore doesn't cover them.
    • Keys inside your built website (dist/, build/, .next/static). That's what visitors actually download.
    • Database passwords and login secrets typed straight into the code.

    What you get

    • A short report: LEAKS FOUND or NO LEAKS FOUND.
    • Every leak with its file and line, and what could happen.
    • The exact place in each provider's dashboard to revoke the key.
    • Say "fix the leaks" and your agent moves the keys to server code one at a time, asking before each change, then re-checks.

    No false alarms about keys that are meant to be public

    The Supabase anon key, the Firebase web config and Stripe publishable keys belong in browser code. Launch Guard Lite knows this and checks every finding before it reports it.

    Safe by design

    • One file, nothing to install. The scanner is a single file with no dependencies.
    • Read-only. It reads only files inside your project folder. It makes no network requests, starts no other programs and writes nothing.
    • Masked secrets. Every secret it finds is printed masked.

    Want the full pre-launch check?

    Launch Guard Lite checks keys only. Launch Guard also checks:

    • Supabase and Firebase access rules
    • fake-able Stripe webhooks and prices set by the browser
    • AI endpoints anyone can call
    • ownership and admin checks
    • launch basics such as legal pages, mobile layout and error pages

    It finishes with a SHIP / FIX FIRST / BLOCK verdict and guided fixes.

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Frontend & Web Apps.

    • Checks for leaked secrets only. It does not check database rules, payments, server endpoints or launch basics (the full Launch Guard does).
    • It can't tell whether a leaked key was already used by someone else. Check your provider's usage page after revoking.
    • Git history is checked only through the git commands your agent runs, and only when the project is a git repository.
    • It recognizes 20+ common key formats. Unusual or custom secrets are caught only by the general "possible secret" check, so some may be missed.
    • Needs Node.js 18+ to run the scanner. In chat-only agents it falls back to a manual review, which is less thorough.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before buying a skill.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Free to use

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 14 days ago

    • Free with an Agensi account

    Software
    Node.js 18+
    Optional
    Git

    Permissions

    Terminal / Shell
    Read Files
    Write Files

    File Scopes

    **/*

    Runs one read-only Node.js scanner (scripts/scan-secrets.cjs). It reads only files inside your project folder, including .env files to detect leaked secrets, and prints any secret masked. It makes no network requests, never reads environment variables, starts no other programs and never writes files. The text "process.env" and URLs appear in it only as detection patterns. The terminal runs the scanner and, in git projects, git log to check whether an env file was ever committed. Your agent writes files only when you approve a fix or ask it to save the report.

    Listed14 days ago

    What's inside

    Frequently Asked Questions

    It scans your code, .env files and built website files for leaked secrets (OpenAI, Anthropic, Stripe, AWS and 20+ other key formats), checks whether .env files are ignored or already committed to git, and tells you exactly which keys to revoke and how to move them to server code.

    Yes. It's built for apps made with Lovable, Bolt, v0, Replit, Cursor and Claude Code, where keys often end up in browser code or behind VITE_ and NEXT_PUBLIC_ variables.

    Yes. It also scans built files in folders like dist/, build/ and .next/static, which is what visitors actually download. A key that slipped into the bundle is caught even if your source code looks clean.

    No. It recognizes keys that are public by design (the Supabase anon or publishable key, the Firebase web config, Stripe publishable keys) and lists them as fine. Your agent also double-checks every finding in the code before reporting it.

    Node.js 18 or newer and read access to your project folder. The scanner is a single file with no dependencies, and it never uses the network.

    No. Lite checks leaked keys only. Database rules (Supabase Row Level Security, Firebase rules), payment webhooks and AI endpoints are covered by the full Launch Guard.

    New in Frontend & Web Apps

    Free