AI / LLM Red Team Field Manual & Consultant’s Handbook provides a comprehensive operational toolkit for AI/LLM red team assessments. It targets engineers, researchers, and power users, offering both tactical field guidance and strategic consulting frameworks to enhance AI security.

How It Works

This repository serves as a "Gold Master" release of a standardized, 46-chapter curriculum covering the AI security spectrum. It includes a detailed Professional Consultancy Guide and a compact Field Manual for operational reference. The approach integrates theoretical knowledge with practical application, supported by an automated Python testing framework for prompt injection, fuzzing, and safety validation.

Quick Start & Requirements

• Primary Use: Manual exploration via SUMMARY.md for the Handbook and docs/field_manuals/ for operational checklists.
• Automated Testing Prerequisites: Python 3.8+, API access to a target LLM (OpenAI, Anthropic, or local Ollama).
• Setup: Clone the repository, navigate to scripts/, and install dependencies using pip install -r config/requirements.txt.
• Running Tests: Configure environment variables in .env (copied from .env.example) and execute tests via python examples/runner.py --target "gpt-4" --test "prompt_injection".

Highlighted Details

• A complete 46-chapter curriculum covering AI security fundamentals, RAG security, agentic threats, and compliance (EU AI Act/ISO 42001).
• Includes tactical Field Manuals with checklists and quick-reference guides for engagements.
• Features a Python Testing Framework for automated prompt injection, fuzzing, and safety validation.
• Covers a wide range of attacks (Prompt Injection, Jailbreaking, Data Leakage) and defenses (Adversarial ML, Privacy, Mitigation).

Maintenance & Community

Contributions are welcomed via forking and Pull Requests. Issues can be reported through GitHub Issues.

Licensing & Compatibility

The project is licensed under CC BY-SA 4.0 (Creative Commons Attribution-ShareAlike 4.0 International). This license requires attribution and that any derivative works be shared under the same or a compatible license.

Limitations & Caveats

This handbook is intended for Authorized Security Testing Only and for educational purposes. The authors disclaim liability for misuse, and users must comply with the Terms of Service of any public LLMs when conducting tests.