AgentNull

Acerca de

AgentNull is a red team-oriented repository containing a catalog of attack vectors targeting autonomous AI agents, such as those using MCP, LangGraph, and AutoGPT. It includes proof-of-concept (PoC) exploits for each identified threat, structured for both human readability and machine ingestion into SOC/SIEM systems. The repository is intended for educational and internal security research purposes to aid in threat modeling and understanding potential vulnerabilities in AI agent implementations.

Características Principales

• Catalog of AI agent attack vectors
• Structured catalog format for SOC/SIEM integration
• 1 GitHub stars
• Red team-oriented approach
• Focus on autonomous AI agents and MCP
• Proof-of-concept exploits for each attack

Casos de Uso

• Security research on AI agent vulnerabilities
• Red team exercises against AI agents
• Threat modeling for AI agent implementations