vibe-check-recon

Read-only public-surface security recon for AI-built apps.

You shipped your MVP with Cursor / Lovable / Bolt / v0 / Windsurf / Claude Code. Before more users find it, run this to catch what your build agent and hosting platform exposed by default.

npx vibe-check-recon https://my-app.com

What it finds

What it does NOT do

This is passive reconnaissance only. It reads what the target server already serves publicly. It does not:

• Send injection payloads (SQL / XSS / SSRF)
• Brute force or fuzz endpoints
• Attempt authentication
• Touch any data that isn't already public
• Generate load on the target

It's safe to run against your own apps and apps you have permission to test.

What this isn't

This is the 30-second public-surface check. It reads what your server already serves and flags the obvious classes of mistake (.git/ exposed, API keys in the bundle, public buckets, weak headers, etc).

It does NOT touch the inside of your application. The interesting bugs live there.

Free finding vs paid finding — concrete example

You build a Lovable app. It uses Supabase. You ship it. You run npx vibe-check-recon https://yourapp.com.

What the free recon finds:

[info] Possible supabase_url in client bundle https://abcd1234.supabase.co This is expected — Supabase anon keys + URLs are public by design when paired with correct Row Level Security policies.

That's a true positive but a low-impact one. It just confirms the stack. The real question is whether the RLS policies are actually correct.

What the paid Vibe Audit finds, on the same app:

[critical] RLS policy on bookings table grants select to authenticated role unconditionally create policy "Users can view bookings" on bookings for select to authenticated using (true); Any logged-in user (including freshly-signed-up attackers) can read every row in the bookings table, including other users' contact details, party sizes, and payment status. The intended policy was almost certainly using (auth.uid() = user_id). Verified by signing in as a test account and querying bookings directly via the anon key. Patch:

drop policy "Users can view bookings" on bookings;
create policy "Users can view their own bookings" on bookings
  for select to authenticated using (auth.uid() = user_id);

Same stack. Same recon output. Different layer.

The free tool finds public surface. The paid audit reads your actual code, validates intent against the schema, and writes you the patch. That's the diff.

If your app has users, payments, or accounts — get the Vibe Audit. $99 founder special, 12-hour turnaround, ranked written report covering auth, data exposure, DB security, payments, env config, performance, and UX red flags.

Install

# One-shot
npx vibe-check-recon https://my-app.com

# Or install
npm install -g vibe-check-recon
vibe-check-recon https://my-app.com

Requires Node 18+.

Usage

# Pretty output
npx vibe-check-recon https://my-app.com

# JSON for programmatic use
npx vibe-check-recon https://my-app.com --json

# Just the highest-severity finding
npx vibe-check-recon https://my-app.com --top

# Help
npx vibe-check-recon --help

Optional environment

# Enables Brave-search-powered subdomain discovery beyond crt.sh
export BRAVE_API_KEY=BSAxxx

Without BRAVE_API_KEY, recon still works. It just skips the Brave-augmented URL discovery and relies on the public Certificate Transparency log alone.

Output format

JSON shape:

{
  "ok": true,
  "url": "https://my-app.com",
  "scanned_at": "2026-05-08T20:42:00.000Z",
  "stack_signals": { "framework": "next.js", "db": "supabase", "server": "Vercel" },
  "contacts": { "x": "@founder", "email": "hi@my-app.com", "linkedin": "..." },
  "subdomains_found": 4,
  "bucket_urls_found": 0,
  "findings": [
    {
      "severity": "critical",
      "category": "data_exposure",
      "title": "Full .git/ directory exposed",
      "detail": "ref: refs/heads/main",
      "url": "https://my-app.com/.git/HEAD",
      "remediation": "Anyone can clone the entire repo via git-dumper. Add .git to your build-output ignore list and redeploy."
    }
  ]
}

Severity tiers:

• critical — source or secrets fully exposed; immediate disclosure warranted
• high — sensitive config served; needs fast fix
• medium — misconfig with credible attack surface
• low — best-practice gap; mention only when other findings exist
• info — stack signal, not a vulnerability

Origin

Built from the recon module that powers the manual Vibe Audit service. Open-sourced because the public-surface check should be free for every founder shipping an MVP — and because the things this can't catch are exactly what a paid audit is for.

If recon finds something you didn't expect, that's the moment to ask whether the rest of your code has the same shape of issue. That's the audit.

License

MIT. Use it on your apps. Use it on your clients' apps with their permission. Don't use it as a launching point for active attacks.

Author

Matt Turley — Fractional CTO, 20+ years. Audits AI-built apps before launch. uxcontinuum.com/vibe-audit.