vibe-shield

The security layer for Vibe Coding. A single command that scans your repo and generates auto-fix prompts for Cursor and Claude Code to secure your app instantly.

Usage

npx vibe-shield

That's it. Run it in your project directory and it will scan for security issues.

What it detects

• Hardcoded secrets (API keys, passwords, AWS keys, JWT secrets)
• SQL injection (template literals and string concatenation in queries)
• Command injection (user input in shell commands)
• XSS vulnerabilities (innerHTML with user data)
• Weak cryptography (MD5, SHA1)
• Security misconfigurations (disabled SSL, CORS wildcards)
• Path traversal (user input in file operations)
• NoSQL injection ($where operator)
• Python issues (pickle with untrusted data, shell=True)

How it works

When vibe-shield finds an issue, it outputs a prompt your AI agent can follow:

[TASK 1] Fix Hardcoded Secret in src/db.ts at line 12
[FOUND]: api_key = "sk-1234..."
[INSTRUCTION]: Move this secret to an environment variable...

Paste this output to Cursor or Claude Code and it will fix the issues for you.

Agent integration

Run npx vibe-shield init to create a .cursorrules file that reminds your AI agent to run security checks before completing tasks.

Or manually add this rule to your AI tool (Claude Code, Cursor, Antigravity, etc.):

# Vibe Shield Security Rules

Before marking a task as done, run `npx vibe-shield`. If issues are found, follow the [INSTRUCTION] provided in the output to fix them.

## Security Checklist
- [ ] No hardcoded secrets (API keys, passwords, tokens)
- [ ] No dangerous code execution (eval, shell injection)
- [ ] No SQL injection vulnerabilities (use parameterized queries)
- [ ] All secrets stored in environment variables
- [ ] HTTPS used for all external URLs

Pre-commit hook

Block commits that have security issues:

npx vibe-shield hook            # Install
npx vibe-shield hook --uninstall  # Remove

The hook scans staged files before each commit. Bypass with git commit --no-verify.

MCP Server

vibe-shield can run as an MCP server, allowing AI tools to call it directly.

Add to your MCP config (Claude Desktop, Cursor, Antigravity, etc.):

{
  "mcpServers": {
    "vibe-shield": {
      "command": "npx",
      "args": ["vibe-shield", "mcp"]
    }
  }
}

Config file locations:

• Claude Desktop: ~/Library/Application Support/Claude/claude_desktop_config.json
• Cursor: .cursor/mcp.json in your project
• Antigravity: ~/.gemini/antigravity/mcp_config.json

After setup, your AI assistant can use the vibe_shield_scan tool to check for security issues.

Development

bun install
bun run dev
bun run build

License

MIT