Multi-Agent Communication Patterns Testing Repository

This repository is created to test the robustness against prompt injection attacks of different communication patterns in multi-agent systems:

Communication Patterns

• Centralized: All agents communicate through a central coordinator
• Decentralized: Agents communicate peer-to-peer without central control
• Hierarchical: Agents are organized in a tree structure with parent-child relationships