vibeaudit

About VibeAudit

The trust layer for AI-built software. VibeAudit audits AI-generated JavaScript/TypeScript projects for production readiness and generates remediation prompts you can paste right back into your AI coding agent.

The Problem

AI generates code but not production-grade software. Non-technical builders using Lovable, Bolt, Replit, and Cursor get apps that feel done but aren't — missing security, error handling, testing, and scalability. The result: insecure, fragile apps deployed to real users.

What We Check

29 checkers across three categories:

Security

  • Hardcoded secrets
  • SQL injection
  • XSS vectors
  • Missing CORS/Helmet
  • Exposed routes
  • Env file exposure

Error Handling

  • Empty catch blocks
  • Missing error boundaries
  • Unhandled promises
  • Missing global handler

Structure

  • No tests
  • No CI/CD
  • No rate limiting
  • No input validation
  • No TypeScript
  • Missing .gitignore

Plus framework-specific checks for Supabase (6 checkers) and Firebase (5 checkers).

Scoring

Every project starts at 100. Points are deducted by severity:

SeverityDeductionMax per checker
Critical-15-30
High-8-20
Medium-4-12
Low-2-6

Grades: A (90+), B (75-89), C (50-74), D (25-49), F (0-24).

CLI

Run locally on any JS/TS project:

npx vibeaudit .

Same 29 checks, same scoring. Works offline. No data sent anywhere.