Scan your vibe-coded app.
Then get a human to finish it.
30-point security scan in seconds. Hit a wall? Book a 30-min review with a senior engineer — free during beta — and walk away with a build plan.
howll.ai — AI for business →
joshhorneman.com →
30 checks · 10 categories
Every check targets a vulnerability pattern AI code generators introduce. Each includes detection commands and fix instructions.
- API keys in frontend code
- .env files in git history
- Weak JWT secrets
- No rate limiting on auth
- JWTs in localStorage
- Tokens that never expire
- Sessions not invalidated
- Weak password hashing
- Frontend-only admin auth
- Missing auth middleware
- IDOR vulnerabilities
- SQL injection
- XSS via unsafe rendering
- Missing input validation
- eval / exec / pickle
- CORS wildcard
- Running as root
- Database port exposed
- No HTTPS enforcement
- Debug mode in production
- Stack traces in responses
- Supabase/Firebase RLS off
- Unsigned payment webhooks
- Hallucinated dependencies
- Uploads without MIME check
- Open redirects
- Missing security headers
- CSP missing or permissive
- No logging or monitoring
- No backup or recovery
Every check weighted by severity
Your score is the percentage of applicable points you pass. N/A checks excluded.
F <60 · D 60–69 · C 70–79 · B 80–89 · A 90–100 · max: 220 points
Maps to the frameworks that matter
Every failed check links to specific clauses. Turns a technical audit into a business risk conversation.
If you're an Australian business, APP 11 requires "reasonable steps" to protect personal information. These 30 checks are free. Hard to argue they're not reasonable.
Three ways your AI fails you.
All of them have a human fix.
You didn't buy Claude. You rented it. When it loops, burns tokens, or leads you somewhere it can't get you out of — that's when the marketplace earns its keep.
The scan finds the wall.
A human gets you over it.
AI gets you 80% of the way. Then it loops. Then it burns tokens. Then it ships you something you don't trust. A Vibe Check connects vibe coders with humans who can finish the job — audits, scaling, design, launch, any of it.
Book a 30-min vibe-check review.
A senior engineer walks your repo with you. Finds the gaps AI missed. You leave with a short written build plan — and a clear view of what to ship, fix, or rebuild next.
What happens next.
The call isn't the pitch. It's the plan. Here's what usually comes after a review.
Paths B and C are the upsell. Path A is the default. We'd rather tell you the plan is trivial and send you on your way than invent work.
Install the full CLI tool
The web scanner above catches the most common patterns. The full CLI skill runs deeper analysis with framework-specific detection.
Framework detection: Next.js · React · Express · Django · Flask · Supabase · Firebase · Docker
The scan takes five minutes.
The review takes thirty.
Both are free.
Run the scan. If you're stuck, book the call. Walk away with a plan.