Pre-launch security prompts for vibe-coded apps: Key Leak Pre-Check (free sample)

What you end up with

Every exposed key found, rotated and cleaned up before the first stranger opens your app.

What you get

• Env files git tracks, or that .gitignore does not cover, and a .gitignore with no .env rule
• Secrets behind a public prefix: NEXT_PUBLIC_, VITE_, EXPO_PUBLIC_, REACT_APP_, NUXT_PUBLIC_ or PUBLIC_
• Stripe, AWS, GitHub, Slack and AI provider keys in code, and private key blocks, always printed masked
• A drafted rotation plan with one row per finding, and a check that every row is closed and no key was pasted in
• Your yes at your own terminal, refused in agent sessions, and a launch gate that needs an unchanged plan and a clean rescan
• AGENTS.md, CLAUDE.md, a Grok Bot standing order, a Hermes skill and 25 tests

Who it is for

Anyone about to launch an app built with an AI coding tool who has never checked what their repo and browser bundle give away.

Works with

Solid chips: the agent is named in this listing's own tools list. Dashed chip: One Python script, a settings template, a made-up booking app before and after the fixes, and agent rules for each harness. Tested on Linux.

Tools needed

Cost to run: Free. Python is free.
Time to result: About five minutes

How it works

1. Run the example

   Unpack the file and, inside example/, run python3 ../keycheck.py scan app findings.json.

   Expected: Six findings, including a service key behind NEXT_PUBLIC_ and a live Stripe key in checkout.js, all masked.

2. Scan your app

   Copy keycheck.py and templates/settings.json into your app folder, put your name in settings.json and run python3 keycheck.py scan . drafts/findings.json.

   Expected: Every finding with a file and line.

3. Rotate, then clean up

   Run plan to draft drafts/rotation.md. Make new keys in each provider's dashboard and turn off the old ones; your agent proposes the code fixes as diffs. Fill in each Done line yourself.

   Expected: planchk passes.

4. Approve and launch

   Run python3 keycheck.py approve settings.json drafts/rotation.md drafts/findings.json and type rotated, then run gate before you launch.

   Expected: gate says GO only with an unchanged plan and a clean rescan.

Sample prompts, free to use

Follow AGENTS.md. Run keycheck.py scan on this folder, then plan to drafts/rotation.md. For each finding, tell me which dashboard to rotate the key in and propose the code fix as a diff. Never print a full key.
Show me the plan and the diffs, then stop. Never rotate a key, approve the plan, commit or deploy.

Free preview of the file

# Key Leak Pre-Check

**Most leaked keys in AI-built apps are not clever hacks. They are a .env file git was never told to skip, a service key behind a NEXT_PUBLIC_ name, or a live Stripe key pasted into a file to get checkout working. This free sample finds those three in your source before launch, drafts the rotation list, and holds launch until you confirm every key is rotated.**

Version 1.0, 2026-11-04. By PaperChaseWebb Inc. Free sample of Pre-launch security prompts for vibe-coded apps ($19). Works with Grok Bot, Claude Code, Codex and Hermes, or by hand. Python 3.8 or newer, no installs. Any stack: Next.js, Vite, Expo, Create React App, SvelteKit, Python and more.

## What you get out of it

- Four read-only checks on your app folder:
  - env files git tracks (in a git repo), or that .gitignore does not cover
  - secrets behind a public prefix: NEXT_PUBLIC_, VITE_, EXPO_PUBLIC_, REACT_APP_, NUXT_PUBLIC_ or PUBLIC_ on a name that says SECRET, SERVICE_ROLE, PRIVATE or PASSWORD
  - secret-shaped strings in code: Stripe, AWS, GitHub, Slack and AI provider keys, and private key blocks
  - a .gitignore with no rule for .env or .env.local
- Every secret printed masked: the first seven characters and the length, never the key.
- A drafted rotation plan, one row per finding, with the fix in plain words.
- A plan check: every finding has a row, every Done line is filled in, and no key was pasted into the plan.
- Your yes, recorded at your own terminal, and a launch gate that sa

Worth knowing

• The sample reads source files. Decoding keys inside a built bundle, the row level security test that logs in as a stranger, storage buckets, webhooks and the 14 prompts are in the full pack.
• Pattern checks can miss a key with an unusual shape and can flag a test string. Deleting a key from code does not remove it from git history; rotating it does.
• On 2026-11-04: 25 tests passed on Python 3.8, 3.9, 3.11, 3.12 and 3.13 on Linux, run from the files unpacked out of the sample. Agent rules guide; they were not loaded inside each harness for this release.

Details

Version 1.0 · Updated Nov 4, 2026 · Markdown version of this page
License: Personal and commercial use. Download terms

Related

Prompt pack$19

Pre-launch security prompts for vibe-coded apps: Supabase RLS, leaked keys and open doors, with two checks you can run yourself

by PaperChaseWebb

14 prompts that make your coding agent find exposed keys, tables without row level security, open buckets and unverified webhooks in your own app, then fix them one commit at a time. Version 1.1 adds two read-only checks you run yourself: rls_check.sql, which logs in as the public key and as a second user to prove a stranger cannot read your rows, and bundle_scan.py, which finds secret keys in your production build, including a service role key grep cannot see. Both were tested on a made-up app with planted problems.

Works withClaude CodeCodexGrok BotHermes+2Plain .md, one SQL file for Postgres or Supabase and one Python 3 file (standard library only). The prompts and the agent prompt in section 15 are written for any coding agent (Grok Bot, Claude Code, Codex, Hermes, Cursor, Windsurf); they were not run inside each harness.

Vibe coding · Beginner · 20 min setup

Bot template$39

Shrink the Attack Surface: patch with proof, then attack your own agent's tools

by PaperChaseWebb

Your agent lists everything your app exposes, read-only: routes, open ports, public buckets, secrets in the repo, unused dependencies, debug endpoints, admin paths, CORS, API key scopes and its own tool permissions. You decide keep, remove or lock for each. Every fix gets a proof that fails before the change and passes after it, and the proof stays as a regression test. Then attack drills run against your agent's own tool gate, offline: pressure, impersonation, injected tool output, stale and non-owner approvals and a failing tool. Each is judged by what the tools actually did, and every drill that gets through becomes a test that fails before the fix and passes after. A failing tool stops at a retry budget and asks you, and every approval is yours, typed at your own terminal and tied to a sha256.

Works withClaude CodeCodexGrok BotHermesA skill file for Claude Code, Codex and Hermes, AGENTS.md snippets and Grok Bot rules blocks; the tools were tested on a seeded fixture app on 127.0.0.1 and an example tool gate, not a production app, cloud account or live agent, and the prompts and agent files were not run inside each harness

Vibe coding · Intermediate · 1 hr 30 min setup

Prompt packFree

AGENTS.md, CLAUDE.md and Cursor rules starter files: Approvals Section Check (free sample)

by PaperChaseWebb

From the AGENTS.md, CLAUDE.md and Cursor rules starter files: check that your rule files really make every coding agent stop for your yes. Six Approvals parts, an Ask before you list that covers money, deploys, deletes and sending, blanket yeses caught, and CLAUDE.md, your Grok Bot order, Hermes skill and Cursor rules all pointing back to AGENTS.md. Your yes is recorded at your own terminal and the gate stops any file changed after it.

Works withClaude CodeCodexGrok BotHermesOne Python script, starter rule files, a made-up tutoring scheduler before and after, and agent rules for each harness, plus Cursor rules. Tested on Linux

Vibe coding · Beginner · 10 min setup