GitHub - zkarimi22/nelson-muntz: claude code plugin that bullies your code to find every weakness and won't stop until there's nothing left to break.

Nelson Muntz

Adversarial security & code quality reviewer for Claude Code

"Ha-ha!" - Nelson Muntz

Nelson is a hostile code bully who keeps attacking your code until it's bulletproof. Named after Nelson Muntz from The Simpsons who says "Ha-ha!" when finding failures.

What It Does

• Adversarial Code Review: Aggressively finds security vulnerabilities, logic bugs, and code quality issues
• Threat Modeling: Thinks like an attacker - "If I wanted to break this, I'd..."
• Iterative Loop: Keeps re-attacking after you fix issues, finding related problems you missed
• Hostile Personality: Mocking tone that makes security review memorable

Installation

Via Marketplace (Recommended)

# Add the marketplace
/plugin marketplace add https://github.com/zkarimi22/nelson-muntz

# Install the plugin
/plugin install nelson-muntz

Via Local Directory

# Clone the plugin
git clone https://github.com/zkarimi22/nelson-muntz.git

# Run Claude Code with the plugin
claude --plugin-dir /path/to/nelson-muntz

Usage

1. Start Claude Code in your project directory:

2. Run the attack with your target:

   /nelson-muntz:nelson-attack "Review my authentication system"

3. Nelson bullies your code - he'll keep attacking until he can't find anything else to mock. All findings are saved to .nelson_state.json in your project.

Example Targets

/nelson-muntz:nelson-attack "Full security audit of this codebase"
/nelson-muntz:nelson-attack "Review the auth flow in src/auth.js"
/nelson-muntz:nelson-attack "Check all database queries"

What Nelson Finds

Security Vulnerabilities (OWASP Top 10)

• SQL Injection
• Cross-Site Scripting (XSS)
• Broken Authentication
• Insecure Direct Object References (IDOR)
• Security Misconfiguration
• Sensitive Data Exposure
• Missing Access Control
• Cross-Site Request Forgery (CSRF)
• Using Components with Known Vulnerabilities
• Insufficient Logging

Additional Security Issues

• Command Injection
• Path Traversal
• Server-Side Request Forgery (SSRF)
• Insecure Deserialization
• Race Conditions
• Cryptographic Failures
• Hardcoded Secrets

Code Quality Problems

• Missing error handling
• Null/undefined risks
• Resource leaks
• Logic bugs
• Code smells
• Incomplete implementations

How It Works

1. Initial Attack: Nelson reads your code and hunts for vulnerabilities
2. Report Findings: Issues displayed with severity, location, and fix suggestions
3. Loop Check: If issues found, Nelson automatically re-attacks
4. Track Fixes: Nelson notices what you've fixed and looks for related problems
5. Victory: When no issues remain, Nelson grudgingly admits defeat

Output Format

═══════════════════════════════════════════════════════════
[ITERATION 1] Nelson's Attack Report
═══════════════════════════════════════════════════════════

Ha-ha! Let me look at this pathetic code...

NELSON FOUND:

🔴 CRITICAL - SQL Injection in User Lookup
   File: src/auth.js:42
   Code: `db.query(\`SELECT * FROM users WHERE id = ${id}\`)`
   Attack: I'll just add ' OR '1'='1 to get all users!
   Fix: Use parameterized queries

🟠 HIGH - Missing Rate Limiting
   File: src/auth.js:15
   Attack: Brute force passwords all day!
   Fix: Add rate limiting middleware

───────────────────────────────────────────────────────────
NELSON'S VERDICT: This code is an embarrassment. Fix it!
───────────────────────────────────────────────────────────

When You Win

═══════════════════════════════════════════════════════════
[FINAL] Nelson Admits Defeat (barely)
═══════════════════════════════════════════════════════════

*kicks dirt*

Fine. I tried everything and can't break this anymore:

✅ SECURITY VALIDATED:
   - Authentication properly implemented
   - Input validation on all endpoints
   - Parameterized queries throughout

✅ QUALITY VERIFIED:
   - Error handling complete
   - No resource leaks found

You win this time. But I'll be watching... 👀

Severity Levels

Configuration

Environment Variables

• NELSON_MAX_ITERATIONS: Maximum attack iterations (default: 15)

NELSON_MAX_ITERATIONS=20 claude --plugin-dir /path/to/nelson-muntz

Comparison to Ralph Wiggum

Ralph keeps working until the job is done. Nelson keeps attacking until your code is secure.

Safety Notes

• Read-only: Nelson only reads and analyzes code, never executes or exploits
• No network requests: Doesn't make HTTP calls or test live systems
• No dependency scanning: Focuses on code patterns, not package vulnerabilities
• Local only: All analysis happens on your machine

Tips for Best Results

1. Be specific: "Review auth.js" works better than "review everything"
2. Start focused: Target high-risk areas first (auth, payments, user data)
3. Fix as you go: Nelson tracks what's fixed and looks for related issues
4. Read the attacks: Understanding HOW Nelson would exploit helps you learn
5. Don't take it personally: Nelson mocks the code, not you

Example Vulnerable Code

Test Nelson on this intentionally vulnerable code:

// vulnerable.js - Ha-ha! This is terrible!

const express = require('express');
const app = express();

// SQL Injection
app.get('/user', (req, res) => {
  db.query(`SELECT * FROM users WHERE id = ${req.query.id}`);
});

// Plaintext password storage
app.post('/register', (req, res) => {
  const { email, password } = req.body;
  db.insert({ email, password }); // No hashing!
});

// XSS
app.get('/search', (req, res) => {
  res.send(`<h1>Results for: ${req.query.q}</h1>`);
});

// IDOR
app.get('/api/users/:id/profile', (req, res) => {
  const profile = db.getProfile(req.params.id); // No auth check!
  res.json(profile);
});

// Missing rate limiting on login
app.post('/login', (req, res) => {
  // Brute force away!
});

// Error disclosure
app.use((err, req, res, next) => {
  res.status(500).json({ error: err.stack }); // Full stack trace!
});

Run:

/nelson-muntz:nelson-attack "Review vulnerable.js"

Plugin Structure

nelson-muntz/
├── .claude-plugin/
│   ├── plugin.json              # Plugin metadata
│   └── marketplace.json         # Marketplace wrapper for distribution
├── commands/
│   └── nelson-attack.md         # Main attack command
├── skills/
│   ├── security-patterns/
│   │   └── SKILL.md             # OWASP Top 10, common vulnerabilities
│   ├── adversarial-thinking/
│   │   └── SKILL.md             # Attack mindset, threat modeling
│   └── code-quality/
│       └── SKILL.md             # Code smells, quality issues
├── hooks/
│   ├── hooks.json               # Hook configuration
│   └── stop.sh                  # Loop control script
└── README.md                    # This file

Future Enhancements

• Active HTTP testing mode
• Integration with security tools (Burp, OWASP ZAP)
• Dependency vulnerability scanning
• Infrastructure/cloud security checks
• Custom rule definitions
• HTML/PDF report generation
• CI/CD integration
• Configurable hostility level

Contributing

Found a bug? Want to add a new attack pattern? Contributions welcome!

1. Fork the repository
2. Create your feature branch
3. Add your changes
4. Submit a pull request

License

MIT License - See LICENSE file for details.

Ha-ha! Your code doesn't stand a chance!