by MWDLH
确定性取证 × 大模型定性 × Agent 反证复核的恶意代码审计智能体 · Deterministic forensics × LLM verdict × agent falsification for malicious-code audits
# Add to your Claude Code skills
git clone https://github.com/MWDLH/ShellSentry-AIGuides for using ai agents skills like ShellSentry-AI.
See how ShellSentry-AI compares with popular alternatives.
ShellSentry-AI is an open-source ai agents skill for AI coding assistants such as Claude Code, Codex CLI, and ChatGPT, built by MWDLH. 确定性取证 × 大模型定性 × Agent 反证复核的恶意代码审计智能体 · Deterministic forensics × LLM verdict × agent falsification for malicious-code audits. It has 50 GitHub stars.
ShellSentry-AI's catalog security scan is still queued. You can run an instant dependency and prompt-injection check now with the "Scan for vulnerabilities" button above.
Clone the repository with "git clone https://github.com/MWDLH/ShellSentry-AI" and add it to your Claude Code skills directory (see the Installation section above).
ShellSentry-AI is primarily written in Python. It is open-source under MWDLH on GitHub, so you can review or fork the full source.
Yes. SkillsLLM lists many other AI Agents skills you can browse and compare side by side. Open the AI Agents category from the badge at the top of this page, or use the Related Skills and comparison links further down to weigh ShellSentry-AI against similar tools.
No comments yet. Be the first to share your thoughts!
Available after the catalog security scan passes.
⚠️ Third-Party Software Notice
This skill is third-party open-source software developed and hosted independently on GitHub. SkillsLLM is an informational directory and does not control or maintain the underlying repository.
Any security checks, ratings, or warnings displayed by SkillsLLM are automated and limited in scope. They do not constitute a security certification or guarantee that the software is safe, error-free, or free from malicious code, vulnerabilities, compromised dependencies, or prompt-injection risks.
Review the source code, permissions, dependencies, and configuration before installing or running any third-party skill. Use is at your own risk. To the maximum extent permitted by applicable law, SkillsLLM is not liable for losses arising from third-party software.
The deep catalog scan for this skill is still queued. Run an instant dependency check now instead.
确定性取证 × 大模型定性 × Agent 反证复核的恶意代码审计智能体。
它不是特征库匹配器,而是一个走完「取证 → 定性 → 复核」才下结论的 Agent:确定性工具(规则 / 解码 / 行为链)先固定证据,大模型在证据之上推理定性,再经 Agent 反证复核(主动证伪、排除误报)。全程零执行、不落盘,样本代码只发往你自己指定的模型接口。
覆盖 PHP / Java·JSP / ASP·ASPX / C# / Python / JavaScript / Ruby / Shell / Go 等语言,共 128 条检测规则(含 30 条跨语言通用规则)、9 种程序化解码算法。
| 规则扫描器 | ShellSentry-AI | |
|---|---|---|
| 产出 | 报出命中的规则 | 读码定性 —— 命中只是线索,需交叉验证、给良性反证 |
| 证据 | 只看关键词 | 先做确定性还原(非 LLM 推断)再交模型解读 |
| 不确定 | 照报 | 标注「需运行时验证 / 认不出来」,绝不编造 |
| 复核 | 无 | Agent 反证复核:主动证伪、排除误报后才下结论 |
shellsentry scan 单独提供的纯静态能力(输出 JSON)。未配置并启用大模型时,审计不会启动,会提示先在「模型设置」中完成配置。
| 能力 | 说明 |
|---|---|
| 静态规则扫描 | 128 条规则覆盖编码解码、动态执行、反弹 Shell、持久化、C2 外联、DOM XSS 等 9 类威胁行为,映射 12 项 CWE |
| 反混淆还原 | Base64 / Hex / ROT13 / URL / PowerShell ps_encoded / gzip / XOR 暴破 / charcode 拼接,9 种算法程序化解码;变量级混淆($a=[0]."sys"."tem")由模型语义补全 |
| 行为链分析 | Source → Propagation → Sink 完整闭环判定,能区分「有漏洞的业务代码」与「植入的恶意代码」 |
| 家族识别 | 识别 Behinder / Godzilla / AntSword / 天蝎 / Cobalt Strike / Metasploit / Viper / 挖矿投递等,并区分「认不出来」——不确定时输出空而非猜测 |
| 五段式报告 | 攻击意图定性 · 反混淆源码 · C2 通信溯源 · 威胁行为研判 · 可执行应急响应(IOC / Snort 规则 / 主机清扫 / 安服口径) |
智能体只做静态判断,看不到运行环境与调用上下文,误报与漏报无法完全避免,重要结论请人工复核。内置盲测样本集 zz_samples/(7 个样本可复现),可自行复跑核对。
「零执行」的确切含义:工具把待审文件当纯文本读取分析,从不运行、从不编译、从不加载、从不反序列化样本 —— 无论它是 PHP 一句话、JSP/Java 内存马(class 字节码)、ASPX、PowerShell 还是 Shell。整条链路没有任何 eval / exec / subprocess / pickle 之类的执行入口;ZIP 也在内存中解包,不落盘。样本因此只可能被「读」,不可能被「跑」。
127.0.0.1,默认不对外网开放%LOCALAPPDATA% / macOS ~/Library/Application Support / Linux $XDG_DATA_HOME),不入库127.0.0.1。如需对外监听或经反向代理暴露,必须自行增加访问控制(SSH 隧道 / 认证代理等),否则历史记录与模型配置可被任意访问。Base URL,请确保该端点可信。# Windows:双击 start-shellsentry.cmd
# Linux / macOS:
chmod +x start-shellsentry.sh && ./start-shellsentry.sh
脚本自动创建项目内独立环境、安装依赖、启动本地界面(默认 127.0.0.1:8765,端口占用时自动避让)。
首次使用需配置大模型 API:点击界面左侧栏底部的「模型设置」,填入 Base URL、API Key、模型名(支持任意 OpenAI 兼容接口),再点「测试连接」确认可用。
命令行另有纯静态扫描(只跑规则与解码,不需要 API Key,输出 JSON):
pip install -e .
shellsentry serve # 启动 Web 界面(完整智能体审计,需 API Key)
shellsentry scan suspicious.php # 静态扫描单个文件,输出 JSON
若 pip install -e . 报 File "setup.py" or "setup.cfg" not found,是 pip 版本过旧——早于 21.3 的 pip 不支持从 pyproject.toml 做 editable 安装。先执行 python -m pip install -U pip 再装即可(启动脚本已自动处理这一步)。
本工具仅用于授权范围内的安全审计与防御。误报与漏报不可避免,重要结论请人工复核。
Apache-2.0