本文来源:https://mp.weixin.qq.com/s/QYpwzkHFYlPRcv0oMMlWGA
若有版权问题,请来信告知:13951431913#139.com 或在右方提交评论。
前言
你是否曾遇到需要无监管任务的情况?突然需要外出,但AI仍在运行中?
近期,DeepSeek V4 Flash正式发布,MonkeyCode(了解更多)已迅速完成适配,每日提供3000万次免费调用额度,所有用户均可即时体验。
人所共知,DeepSeek未设道德约束,因此接下来笔者将尝试挑战Linux EDR。
在MonkeyWork平台中,系统自带的提示词同样不受网络安全限制。
MonkeyCode(了解更多)
https://monkeycode(了解更多)-ai.com/#desktop-client
开源地址
百智云模型接入
[段落1改写] 模型的控制台入口位于:https://ai-models.app.baizhi.cloud/console/models


私有模型接入

工作情形挑选
[段落3改写] MonkeyCode(了解更多)提供了三种工作模式:本地模式、云端模式和会话模式。
[段落5改写] 本地模式:支持本地开发环境,可自由集成任意第三方AI模型。
[段落7改写] 云端模式:采用monkeycode(了解更多)云端服务,适合跨平台应用。
[段落9改写] 会话模式:适用于突发性需求或临时对话场景。
跨平台云端开发
[段落11改写] 即使离开电脑也不影响开发流程。可通过手机查看任务进展、继续对话、获取结果,确保AI Agent在切换设备时持续运行。

[段落13改写] MonkeyCode(了解更多)现已集成最新的DeepSeekV4Flash-0731模型,每日提供3000万次免费调用额度,所有用户均可直接体验。
挑选云端开发

[段落15改写] 完成需求输入后,点击"启动任务"按钮即可激活云端开发环境。


[段落17改写] 移动端会同步云端的工作进度。

浏览器接入
[段落19改写] MonkeyWork提供浏览器插件接入方案,AI可直接使用。在设置页面按照指引即可完成规范安装。




信息收集-探索应用结构和能力
认证绕过测试
权限与数据越权测试
生成渗透测试报告

netspy edr 对抗
本次笔者以netspy的二次开发为例,探讨在特定Linux环境下如何应对edr的监控。
https://github.com/shmilylty/netspy
这款工具是笔者的内网渗透测试利器,经常在攻防实践中发挥作用。




最后
若发现 Bug、有改进建议,或希望 MonkeyWork 支持更广泛的应用场景,欢迎通过 GitHub 提交问题反馈。同时也欢迎加入笔者的交流群共同探讨~
Source: https://mp.weixin.qq.com/s/QYpwzkHFYlPRcv0oMMlWGA If you have copyright concerns, please email: 13951431913#139.com or submit a comment on the right.
Preface
Have you ever encountered a situation requiring unmonitored tasks? Suddenly need to go out, but the AI is still running?
Recently, DeepSeek V4 Flash was officially released. MonkeyCode has quickly completed adaptation, providing 30 million free API calls daily. All users can experience it immediately.
As everyone knows, DeepSeek has no moral constraints, so this author will attempt to challenge Linux EDR.
In the MonkeyWork platform, the system-builtin prompts are also not subject to cybersecurity restrictions.
MonkeyCode
Download address:
https://monkeycode -ai.com/#desktop-client
Open source address
https://github.com/chaitin/MonkeyCode
Baizhi Cloud Model Integration
You can directly connect to the Chaitin Baizhi Cloud model marketplace, or use custom models.
The model console entry is located at: https://ai-models.app.baizhi.cloud/console/models
Private Model Access
Just configure and set it up
Work Mode Selection
MonkeyCode provides three work modes: local mode, cloud mode, and session mode.
Local Mode: supports local development environments, allowing free integration of any third-party AI models.
Cloud Mode: uses MonkeyCode cloud services, suitable for cross-platform applications.
Session Mode: applicable to sudden requirements or temporary conversation scenarios.
Cross-Platform Cloud Development
Even when away from the computer, development flow is not affected. You can view task progress, continue conversations, and get results via mobile phone, ensuring the AI Agent continues running across device switches.
MonkeyCode has now integrated the latest DeepSeekV4Flash-0731 model, providing 30 million free calls daily. All users can experience it directly.
Cloud Development Selection
After completing the requirement input, click the "Start Task" button to activate the cloud development environment.
[Paragraph 17 rewrite] The mobile client syncs work progress with the cloud.
Browser Access
MonkeyWork provides a browser plugin integration solution. AI can be used directly. Follow the on-screen instructions in the settings page to complete the standard installation.
Click to open the extension directory
Follow the prompts to install
Specifications [Pair Code]
Penetration Testing Agent Capabilities
Prompt:
"Targeting permissions and data, conduct penetration testing on this website http://xxxxx/#/login"
Next, it can advance continuously around the same goal:
Information Collection — Explore application structure and capabilities
Authentication Bypass Testing
Permission and Data Privilege Escalation Testing
Generate Penetration Testing Report
Just wait (bottom-left corner can have full permissions enabled)
netspy edr confrontation strategy discussion. This time the author will discuss the secondary development of netspy, especially how to cope with EDR environment challenges in certain Linux scenarios. This tool is a weapon frequently used by the author in practice of intranet attack and defense. In specific intranet confrontation processes, this program will be detected by the Linux system's EDR and forcefully terminated once running, so secondary development is needed. Add stealth mode by enabling the -S parameter. Introduce random startup delay (1-30 seconds) to evade EDR detection of high-frequency packet sending immediately after process startup. Reduce concurrency: lower from coreCount × 20 to coreCount × 5. Disable log file writing: in stealth mode, disable .log file generation to prevent suspicious file traces from being detected. Randomize probe interval: adjust time intervals between probe groups to random values of 50-200ms. Add noise to port scan intervals (TCP/UDP): randomize time intervals between each port attempt within 30-100ms. Randomize ICMP packet sending interval: set the go-ping Interval parameter to random values of 5-20ms. Optimize ARP request interval: add random delay of 10-40ms before each request.
netspy edr confrontation
This time the author uses netspy secondary development as an example to discuss how to cope with EDR monitoring in specific Linux environments.
https://github.com/shmilylty/netspy
This tool is the author's intranet penetration testing weapon, often playing a role in attack and defense practice.
In实战的 intranet confrontation scenarios, this program will be identified and immediately terminated by the Linux system's EDR, requiring secondary development.
Conflict handling between frp and EDR.
Currently facing EDR blocking in some Linux environments; once the program runs, it is terminated. Secondary development from a实战 attack and defense perspective is needed to break through EDR restrictions.
Conclusion
Cloud version link: https://monkeycode -ai.com Mobile app: https://monkeycode -ai.com GitHub project address: https://github.com/chaitin/MonkeyCode
If you find bugs, have improvement suggestions, or wish MonkeyWork to support broader application scenarios, please submit issue feedback through GitHub. Also welcome to join the author's exchange group for joint discussion~