ThreatCode Resources

A collection of open security research resources for AI/LLM safety, built by ThreatCode Information Security.

Projects

resources/context — AI Code Security Anti-Patterns

Comprehensive security anti-pattern references distilled from 150+ sources to help LLMs generate safer code. Covers 25+ patterns across breadth and depth versions.

• ANTI_PATTERNS_BREADTH.md — ~65K tokens, 25+ patterns
• ANTI_PATTERNS_DEPTH.md — ~100K tokens, 7 critical patterns
• Interactive Website

resources/taxonomy — Prompt Injection Taxonomy

An interactive, searchable classification of prompt injection and LLM attacks (172 nodes across 4 pillars), built for cybersecurity auditors and penetration testers moving into AI red teaming.

• Interactive Website
• JSON Data
• GitHub Pages

Quick Stats

Metric	Value
Sources Analyzed	150+
Anti-Patterns Covered	25+
Taxonomy Nodes	172
CVEs Documented	40+
License	CC BY 4.0

Structure

resources/
├── context/          # AI Code Security Anti-Patterns
│   ├── README.md
│   ├── index.html    # Standalone landing page
│   ├── ANTI_PATTERNS_BREADTH.md
│   └── ANTI_PATTERNS_DEPTH.md
└── taxonomy/         # Prompt Injection Taxonomy
    ├── README.md
    ├── CHANGELOG.md
    ├── docs/
    │   ├── index.html
    │   ├── styles.css
    │   ├── app.js
    │   └── data/
    │       ├── taxonomy.json
    │       └── taxonomy.js

Getting Started

1. Clone the repository

   git clone https://github.com/threatcode/resources.git

2. View the websites — Open context/index.html or taxonomy/docs/index.html in a browser, or deploy to GitHub Pages.

3. Use the data — Include the markdown files in your AI system prompt, or use the JSON data programmatically.

Contributing

We welcome contributions! See CONTRIBUTING.md for guidelines.

License

This work is licensed under the Creative Commons Attribution 4.0 International (CC BY 4.0) license.

Research Sources

This project synthesizes findings from 150+ sources across CVE databases, academic research, security blogs, developer forums, and GitHub security advisories.

---

Built by ThreatCode Information Security · Jason Haddix