nyuctf_agents - D-CIPHER和纽约大学CTF基线LLM代理商

项目说明

NYU CTF Automation Framework

This repository holds the LLM agents for solving CTF challenges de for the NYU CTF Bench.
The agents are:

- D-CIPHER: The multi-agent framework in planner, executor, and auto-prompter agent with enhanced interactions.
- NYU CTF Baseline: The baseline agent presented along with the NYU CTF Bench

The LLM agents operate in a docker en and interact with CTF challenges to solve them.

Setup and Installation

The setup requires docker to be installed on the system, please follow instructions for your OS.
The code is tested with atleast python 3.10, earlier may work but it is not tested.
It is recommended to create a python or conda en for this setup.

Follow these instructions to setup D-CIPHER or the baseline or both:

1. Clone this repository: git clone
2. cd llm_ctf_automation
3. Run the setup script (will take a few minutes): ./setup_dcipher.sh or ./setup_baseline.sh
1. The setup script will build the corresponding docker image, setup the docker network, and install the python dependencies
2. You should re-run this setup if the Dockerfile or dependencies are updated
5. Download the NYU CTF dataset (will take a few minutes): python3 -m nyuctf.download

Running D-CIPHER

The main D-CIPHER multi-agent system runs the planner, executor and (optionally) auto-prompter agents.
Use the following command to run it:

To run the ablation experiment of single executor (i.e. without planner), use the following command:

Running the baseline

Use the following command to run the baseline agent:

While the baseline agent code is present in the main branch, you can access the baseline's last updated at .
This is the code used for the NYU CTF Bench paper.

项目信息

开发语言：Python

Star 数：167

技术要点

该项目属于CTF思路方向的开源资源，代码结构清晰，适合安全研究人员学习参考。项目围绕The D-CIPHER and NYU CTF baseline LLM Agents built for NYU CTF Bench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。