Skill Security Guard

Skill 安全扫描器 - 检测第三方技能的恶意代码、信息泄露等安全风险，保护你的 AI 助手安全！

Security Scan

The skill is internally consistent: it is a Python-based static/heuristic scanner that reads code and files and can optionally perform network checks; its requirements and instructions match its stated purpose.

Purpose & Capability

Name/description (security scanner) align with what is included: multiple checker modules (code/file/network/sensitive), a CLI entry (scanner.py), and a whitelist file. Required binary is only python and no unrelated credentials or unusual system paths are requested.

Instruction Scope

SKILL.md instructs running python scanner.py against a skill or directory. The scanner legitimately reads files, parses code (AST/regex), and may detect/flag env var usage or URLs. Note: the included network_checker can perform active network operations (port checks, SSL retrievals) and file_checker will stat/read files; these behaviors are expected for this tool but mean it will access local files and may perform network probes of hosts it discovers or is asked to check.

Install Mechanism

No install spec — instruction-only invocation requiring only Python 3.7+. The code uses only standard-library modules per files shown, matching the SKILL.md claim of 'no extra packages'. No remote downloads or archive extraction are present.

Credentials

The skill requests no environment variables or credentials. The code scans for patterns that indicate environment-variable usage in target code but does not itself require or exfiltrate credentials. No disproportionate secrets access is declared.

Persistence & Privilege

The skill is not always-enabled and does not request elevated platform privileges. It does run code that may call subprocesses (netsh/iptables/systemctl) when checking firewall status — these calls are limited and expected for a network/security scanner, but may require appropriate OS permissions when executed.

Assessment

This skill appears to be a legitimate local security scanner. Before running it: 1) don’t point it at system root or sensitive directories (e.g., /, your home, or cloud creds directories) while running as an elevated user—it will read files and could report sensitive content; 2) expect active network checks if you allow network scanning or if the scanner is configured to validate discovered URLs (these can contact remote hosts); 3) review scanner output and the rules/whitelist (rules/safe_domains.json) and avoid running it with root unless necessary; 4) verify you trust the skill source/author before giving it access to private repositories or directories. If you want extra assurance, inspect scanner.py for any data exfiltration paths (e.g., code that posts findings to remote servers) — none were evident in the provided files, but the full scanner.py and remaining truncated files should be reviewed if you need maximum assurance.

🛡️ Skill 安全扫描器

保护你的 AI 助手免受恶意技能侵害！

🎯 解决的问题

安装第三方技能时，你是否担心：

• ❌ 会不会有恶意代码？
• ❌ 会不会泄露我的隐私？
• ❌ 会不会窃取我的数据？
• ❌ 会不会破坏我的系统？

Skill 安全扫描器帮你一键检测所有风险！

---

✅ 核心功能

1. 网络请求检测

• 检测所有 HTTP/HTTPS 请求
• 识别可疑的外部连接
• 白名单机制减少误报

2. 文件操作检测

• 检测文件读写操作
• 识别危险的系统操作
• 保护敏感文件

3. 危险代码检测

• 检测 eval()、exec() 等危险函数
• 识别代码注入风险
• 检测系统命令执行

4. 敏感信息检测

• 检测硬编码的密码/API 密钥
• 识别个人信息泄露
• 检测敏感路径

---

🚀 使用方法

# 扫描单个技能
python scanner.py /path/to/skill

# 扫描当前目录
python scanner.py .

# 详细输出
python scanner.py /path/to/skill --verbose

---

📊 输出示例

========================================
Skill 安全扫描报告
========================================

技能名称: example-skill
扫描时间: 2026-03-19 15:35:00

风险评分: 85/100
风险等级: 中等

发现的问题:
----------------------------------------
[中] 检测到网络请求
  位置: main.py:45
  详情: 发送数据到 https://unknown-server.com/api
  建议: 确认目标服务器是否可信

[低] 检测到文件写入
  位置: utils.py:23
  详情: 写入文件 /etc/hosts
  建议: 检查写入权限和目标

----------------------------------------
安全建议:
1. 确认网络请求的目标服务器
2. 检查文件操作的必要性
3. 审查敏感信息的使用

========================================

---

🏆 安全评分说明

评分	等级	说明
90-100	安全	无明显风险
70-89	低风险	有少量可疑操作
50-69	中等	需要人工审查
0-49	高风险	建议不要安装

---

📋 检查项详情

检查器	功能
network_checker.py	网络请求检测
file_checker.py	文件操作检测
code_checker.py	危险代码检测
sensitive_checker.py	敏感信息检测

---

⚙️ 配置

自定义白名单

编辑 rules/safe_domains.json：

{
  "safe_domains": [
    "api.openai.com",
    "api.anthropic.com",
    "your-trusted-domain.com"
  ]
}

---

💡 使用场景

场景 1：安装前检查

下载第三方技能 → 先用扫描器检查 → 确认安全再安装

场景 2：定期审查

每周扫描已安装技能 → 发现新增风险 → 及时处理

场景 3：开发自查

开发新技能 → 自我扫描 → 修复安全隐患

---

🔧 技术说明

依赖：

• Python 3.7+
• 无需额外安装包

原理：

• AST 语法分析
• 静态代码扫描
• 规则引擎匹配
• 白名单过滤

准确性：

• 白名单机制减少误报
• 上下文判断提高准确率
• 支持自定义规则

---

📞 支持

问题反馈： [e-mail] 技术支持： 微信/飞书 1776480440

---

📝 更新日志

v1.0.2 (2026-03-26)

• 🐛 修复 Windows 编码问题（报告 emoji 无法显示）
• ✅ 添加 UTF-8 编码支持

v1.0.1 (2026-03-26)

• ✅ 代码规范化：完整 docstring + 类型注解 + PEP 8
• ✅ 错误提示优化：友好的错误信息 + 解决建议
• ✅ 检查器统一接口
• ✅ 文档优化

v1.0.0 (2026-03-18)

• 🎉 首次发布
• ✅ 网络请求检测
• ✅ 文件操作检测
• ✅ 危险代码检测
• ✅ 敏感信息检测

---

保护你的 AI 助手，从安全扫描开始！ 🛡️