XuanMu-RedTeam-Agent - 玄幕安全团队开源汉化和并二次开发了新功能的无需docker可在Kali Linux上直接运行的红队A

项目说明

<p align="center">
<strong>English</strong> ·
<a href="README_zh.md">中文</a>
</p>

<p align="center">
<strong>Open-source red-team multi-agent collaboration platform with Blackboard-style shared reasoning graph.</strong>
</p>

<p align="center">
<a href="#o">O</a> ·
<a href="#architecture">Architecture</a> ·
<a href="#blackboard">Blackboard</a> ·
<a href="#quick-start">Quick Start</a>
</p>

---

> Security Notice
>
> This project is intended only for authorized security testing, risk assessment, and academic research.
> Unauthorized use against any system, network, or data is prohibited.
> The authors are not responsible for any misuse or damages.

---

O

XuanMu (玄幕) is a control-plane-oriented red-team multi-agent platform built for authorized penetration testing, discovery, code auditing, and security research.

It combines:

The design goal is to make agent-assisted security work operationally bounded and fully re. Con are not the only source of truth — project scope, assets, findings, relationship graphs, attack paths, replayable timelines, and the agent's reasoning process are all persisted as explicit application data.

---

Architecture

Agent Team

| Code | Name | Role |
|------|------|------|
| cso | XuanMu (玄幕) | Security Lead — task decomposition, coordination, reasoning |
| cae | ShouZhuo (守拙) | Code Audit Engineer |
| cie | GuanXing (观星) | Intelligence & Recon Engineer |
| cpe | PoJun (破军) | Penetration Testing Engineer |
| cre | SuYuan (溯源) | Reverse Analysis Engineer |
| cce | PoZhen (破阵) | Cryptography Engineer |

Execution

Commands run directly on the host asyncio subprocess — no Docker required. The execution layer supports:

---

Blackboard

The Blackboard is a Cairn-inspired shared reasoning graph that enables Stigmergy — agents coordinate through a shared board instead of talking over each other.

Node Types

| Type | Meaning | Lifecycle |
|------|---------|-----------|
| Fact | A confirmed, objecti | proposed → confirmed / rejected |
| Intent | A declared exploration direction | proposed → in_progress → confirmed / rejected / superseded |
| Hint | Human or agent guidance | persisted |

How Agents Use It

The Blackboard layer complements the existing e plane (Asset / Finding / GraphEdge): e records what was found, the blackboard records why we looked, what we found, what's next.

REST API

| Endpoint | Method | Description |
|----------|--------|-------------|
| GET /api/blackboard/{project_id} | Read | Full blackboard graph snapshot |
| POST /api/blackboard/{project_id}/nodes | Create | New fact / intent / hint node |
| PUT /api/blackboard/{project_id}/nodes/{node_id} | Update | Change status / description |
| DELETE /api/blackboard/{project_id}/nodes/{node_id} | Delete | Remove a node |

---

Technical Highlights

| Feature | Description |
|---------|-------------|
| Multi-agent orchestration | Lead agent coordinates specialists + blackboard |
| Structured e plane | Assets, findings, relationship graph, attack paths — all persisted |
| Blackboard reasoning graph | Cairn-style Fact-Intent graph for traceable agent reasoning |
| Replayable timeline | Normalized timeline events streamed live or loaded as history |
| No Docker required | Commands run directly on host subprocess |
| Knowledge base | Structured methodology documents agents can reference |
| Full web UI | Chat playground + project workspace with graph |
| Local-first LLM | Direct OpenAI-compatible API calls — works with DeepSeek, Qwen, GLM, etc. |

---

Quick Start

Prerequisites

One-command setup

> 📖 New user? Read the Getting Started Guide — from installation to your first project.

This will:
1. Install system dependencies (PostgreSQL, Node.js)
2. Configure the database
3. Create a Python en and install dependencies
4. Build the frontend
5. Create start/stop scripts

Configure

Fill in LLM API keys, base URLs, and model names for each agent role.
Example configuration uses DeepSeek — works with any OpenAI-compatible API.

You can also use the interacti tool:

It walks you through setting API keys, base URLs, and models for each agent without manually editing JSON.

Start

Open — login with / admin123.

---

Repository Layout

---

License

This project is licensed under the MIT License.

---

Acknowledgments

项目信息

开发语言：Python

Star 数：126

技术要点

该项目属于红蓝攻防方向的开源资源，代码结构清晰，适合安全研究人员学习参考。项目围绕玄幕安全团队开源汉化和并二次开发了新功能的无需docker可在Kali Linux上直接运行的红队AI agent。展开，提供了可运行的实现，对理解相关安全机制、工具原理与攻防技术具有较高的研究价值。

从技术实现角度看，此类项目的核心在于对底层原理的深入理解与工程化落地。开发者在研究过程中可以结合官方文档与社区讨论，梳理项目的初始化流程、核心模块职责与数据流转，逐步掌握其中的技术细节与设计思路。

使用与学习建议

1. 下载源码压缩包，解压到本地环境。

2. 根据项目类型准备对应的运行环境与依赖。

3. 按照 README 中的步骤完成配置与部署。

4. 在隔离的测试环境中运行验证，避免影响生产系统。

免责声明

本资源来源于开源社区，仅供安全研究、技术学习与合法授权范围内的攻防演练使用，请勿用于任何非法或侵害他人权益的用途。使用本项目产生的一切后果由使用者自行承担。