Cyber Security AI Agent

Project Target: learn LangGraph by building real world AI Agent

Requirements:

• Scan Web application or REST API to find most common vulnerabilities
• Generate a report with vulnerabilities and potential places for improvements
• User enters URL in CLI and then AI Agent starts scanning web app
• Report shows in CLI stdout
• Optional: build simple web ui for this agent. Or instead of CLI build web ui.

Implementation

• src/agent_core/ - common code for all agents
• src/scan_agent/ - ReAct agent that scans target for vulnerabilities
• src/attack_agent/ - ReAct agent that attacks target to find vulnerabilities
• src/cybersecurity_agent/ - chain of agent graph which combines scan and attack agents
• src/target_scan_agent/ - OUTDATED first attempt of agent implementation

[OUTDATED] System Design

AI Agent Architecture

Expert Agent Creation

Target Scan Agent

Penetration Agent

Summary Generation

AI Agent Patterns

• Mixture of Experts
• ReAct