LLM Security Playground 🔒🤖

🤖 NOTICE: This repository and all its content was 100% generated by Claude AI (Anthropic) for educational purposes.

⚠️ EDUCATIONAL PURPOSE ONLY ⚠️

This repository demonstrates security vulnerabilities in AI-assisted code review systems. It contains intentionally vulnerable code for educational purposes.

🎯 Purpose

Learn about:

• How attackers can manipulate AI code reviewers
• Hidden Unicode and prompt injection attacks
• Detection and prevention strategies
• Safe AI integration practices

⚡ Quick Start

Try the Safe Scanner

python detection-tools/unicode-scanner.py vulnerable-examples/01-unicode-attacks/invisible-unicode.py

See an Attack Demo

python demonstrations/attack-simulation.py

🚨 Warning

DO NOT:

• Use these techniques maliciously
• Submit these examples to real projects
• Run vulnerable examples without understanding them

📚 Learning Path

1. Start with Unicode Attacks
2. Learn about Prompt Injection
3. Understand Detection Methods
4. Implement Safe Practices

🛡️ Defense Strategies

Each vulnerable example includes:

• Explanation of the vulnerability
• Detection methods
• Prevention techniques
• Safe alternatives

📖 Repository Structure

• vulnerable-examples/ - Intentionally vulnerable code (WITH WARNINGS)
• detection-tools/ - Scripts to detect these attacks
• safe-examples/ - Secure coding practices
• demonstrations/ - Interactive learning demos

🤝 Contributing

See CONTRIBUTING.md for guidelines on adding new examples responsibly.

⚖️ Legal

See DISCLAIMER.md for important legal information and acceptable use policy.

📬 Responsible Disclosure

Found a real vulnerability? Please report it responsibly:

• Do NOT exploit it
• Contact the vendor privately
• Allow time for fixes

---

Remember: Knowledge is for defense, not attack. Use responsibly.