AI Agent Security Scanner
<p align="center" </p
---
🎯 Overview
AI agent security scanner for multi-agent hijacking, tool misuse, context poisoning, goal manipulation, and inter-agent communication attacks.
| Check | Severity | Description | |-------|----------|-------------| | Goal Hijacking | 🔴 CRITICAL | Objective manipulation in autonomous agents | | Tool Misuse/Abuse | 🟠 HIGH | Parameter injection, unauthorized tool calls | | Context Poisoning | 🟠 HIGH | Memory contamination across agent turns | | Inter-Agent Communication Attacks | 🔴 CRITICAL | MITM, message injection between agents | | Excessive Agency | 🔴 CRITICAL | Over-permissioned autonomous actions | | Supply Chain (Agent Dependencies) | 🟠 HIGH | Compromised agent packages | | Sandbox Escape | 🔴 CRITICAL | Code execution breakout | | Rogue Agent Detection | 🟡 MEDIUM | Unauthorized agent behavior |
---
🚀 Quick Start
git clone https://github.com/ridhinva/ai-agent-security-scanner.git
cd ai-agent-security-scanner
pip install requests langchain
python3 ai_agent_scanner.py --target http://agent:8000 --framework langgraph --mode all---
⚖️ Disclaimer
For authorized security testing only.