AI Agents

The complete AI agent architecture now lives in agents/README.md.

The platform uses an LLM-first agent model:

AgentContext
  -> Prompt Builder
  -> LLM Provider
  -> Structured AgentResponse

Agents are LLM-powered analytical modules that reason over the attack surface graph, vulnerability dataset, scan history, and risk statistics. They produce structured recommendations, summaries, attack path explanations, and operational suggestions.

Agent categories:

• Advisory Agents: Risk Agent, Attack Path Agent, Executive Summary Agent
• Operational Agents: Recon Agent, Pentest Agent

Operational agents produce recommendations only. Scan execution, scope changes, and other operational actions require approval and backend policy enforcement.