Skip to content

noamrazbuilds/vibe-check

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

/vibe-check -- Security Audit for Vibe-Coded Apps

A Claude Code slash command that scans any codebase for security vulnerabilities and suggests fixes. Built from 77+ real vulnerability findings across 6 production applications, expanded with coverage for the full vibe coding ecosystem.

Install

Option A: Direct file install (simplest)

Copy the command file to your global Claude Code commands directory:

# One-liner install
curl -o ~/.claude/commands/vibe-check.md \
  https://raw.githubusercontent.com/noamrazbuilds/vibe-check/main/commands/vibe-check.md

Or clone and copy:

git clone https://github.com/noamrazbuilds/vibe-check.git
cp vibe-check/commands/vibe-check.md ~/.claude/commands/

After installing, use it as /vibe-check in any project.

Option B: Test without installing

claude --plugin-dir ./vibe-check

Usage

/vibe-check              # Standard scan (Phases 1-6)
/vibe-check quick        # Secrets + critical vulns only (Phases 1-2)
/vibe-check full         # Comprehensive audit (all 8 phases)

What it checks

Phases

Phase Focus Modes
1 Secret detection -- API keys (OpenAI, Stripe, Supabase, Firebase, Clerk, AWS), tokens, passwords, database connection strings, client-side auth secrets all
2 Critical vulnerabilities -- auth bypass, SQL/command injection, database security gaps (RLS, Security Rules, RLS recursion detection), webhook signature bypass, privilege escalation, client-side auth logic all
3 OWASP Top 10 -- XSS, IDOR, SSRF, CORS misconfiguration, sensitive data exposure, NoSQL injection, prompt injection standard, full
4 Input validation -- length limits, file uploads, rate limiting, weak randomness, redirect validation, postMessage origin checks standard, full
5 Dependencies -- npm audit, unused packages, hallucinated packages ("slopsquatting"), deprecated libraries, security headers, SRI standard, full
6 Stack-specific checks -- 11 conditional modules based on detected tech stack (see below) standard, full
7 API contract & type safety -- schema vs interface mismatches, unsafe type assertions, missing runtime validation full
8 Production readiness -- dev bypasses, debug logging, test functions, source maps, environment parity full

Auto-detected stacks (40+ technologies)

The skill auto-detects your tech stack from package.json, config files, and directory structure, then runs targeted checks.

Frontend: Next.js, React, SvelteKit, Nuxt, Astro, Remix, Angular, Express, Fastify, Django, Flask, Rails

BaaS / Database: Supabase, Firebase, Convex, Prisma, Drizzle, MongoDB/Mongoose, PlanetScale, Neon, PostgreSQL RLS (auto-detected from migration files)

Auth: Clerk, Auth.js/NextAuth, Better Auth, Supabase Auth, Firebase Auth, Lucia (flags as deprecated)

Payments: Stripe (webhook verification, price tampering, key exposure), Lemon Squeezy, Paddle

AI/LLM: OpenAI, Anthropic, Vercel AI SDK, Replicate, Google Gemini -- checks for prompt injection, API key proxying, cost protection

Deployment: Vercel, Netlify, Cloudflare Workers/Pages, Railway, Fly.io, Docker, GitHub Actions

Other: Chrome Extensions, Resend, UploadThing, Zod

Stack-specific check modules (Phase 6)

Module Activates when Key checks
Supabase / PostgreSQL RLS @supabase/supabase-js, supabase/ dir, or SQL files with CREATE POLICY RLS gaps, USING(true), RLS recursion/cycle detection, SECURITY DEFINER audit, RLS policy testing, getUser() vs getSession(), service role key exposure, client-side mutations
Firebase firebase in deps or firebase.json Security Rules audit (allow write: if true), Admin SDK in client code, Firestore write validation
Convex convex in deps or convex/ dir Auth in queries/mutations, public vs internal functions, arg validators
Clerk @clerk in deps Middleware coverage, server-side verification, webhook signature checking
Auth.js next-auth or @auth/core in deps Secret configuration, callback security, CSRF protection
Stripe stripe in deps Webhook signature verification, price tampering, API key exposure, idempotency
Next.js next in deps or config Server Action auth, API route auth, middleware coverage, security headers
SvelteKit @sveltejs/kit in deps or config hooks.server.ts auth, server vs client data loading, form action auth
Nuxt nuxt in deps or config Server route auth, runtime config secrets, middleware
Chrome Extension manifest.json with manifest_version Hardcoded credentials, CSP, postMessage validation, permissions
AI/LLM OpenAI, Anthropic, or AI SDK in deps API key proxying, prompt injection, cost protection, output safety

Output format

The skill produces a structured markdown report:

  • Summary table with finding counts by severity (CRITICAL / HIGH / MEDIUM / LOW)
  • Findings with file path, line number, description, exploit scenario, confidence level, and fix recommendation
  • Positive observations to acknowledge good security practices and avoid alarm fatigue
  • Priority action items table with effort estimates

Each finding includes a proven fix pattern with code examples.

How it was built

This skill was distilled from 19 security documents across 6 production applications:

  • 77+ real vulnerability findings catalogued by severity
  • Search patterns proven to find real bugs -- every grep pattern in the skill caught an actual vulnerability
  • Fix templates from real remediations -- webhook verification, DOMPurify, SSRF protection, rate limiting, SECURITY DEFINER templates (including RLS cycle-breaker pattern), RLS policy testing, Firebase Security Rules, auth check patterns for Clerk/Auth.js/Supabase/Firebase
  • Vibe coding anti-patterns from Wiz Research, Vidoc Security, and Kaspersky -- client-side auth, exposed API keys, slopsquatting, missing input validation

Contributing

Found a vulnerability pattern that should be included? Open an issue or PR. The goal is to cover the most common security mistakes across the tools developers actually use.

License

MIT -- Copyright (c) 2026 Noam Raz / Pleasant Secret Labs

About

Claude Code /vibe-check slash command — security audit built for vibe coders. 8 phases, 40+ tech stack detection, 77+ real vulnerability patterns.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors