Suijin
A dual-mode, LangGraph-based security platform: a red-team agent that chains reconnaissance, exploitation, and reporting with terminal-verified exploits, and a blue-team agent that monitors live HTTP traffic and responds with deception, blocking, or patching.
Suijin is an autonomous security testing platform built on LangGraph with a ReAct architecture. Its red-team mode chains reconnaissance, exploitation, and reporting into a state machine, and requires findings to be terminal-verified proof-of-concept exploits or baseline/exploit diffs rather than model claims; a completion gate structurally refuses to let the run quit while surfaces remain untested. Its blue-team mode monitors live HTTP traffic, detects attacks, and responds with deception, blocking, or source patching. Both modes share one toolkit, knowledge base, and knowledge graph.
The project reports autonomous, no-human-in-the-loop benchmark results against Cybench's HTB Cyber Apocalypse 2024 web track and the NYU CTF, scored by exact flag match, and includes onboard labs with chain-verification tests for practice.
The README states the tool is intended for authorized security testing, education, and research only, and that unauthorized use against systems the user does not own or have written permission to test is illegal.