AI Agent Security Scanner
AI Agent安全审计工具。扫描Skill/Agent代码中的敏感信息泄露、API密钥暴露、注入风险、权限问题、数据安全漏洞,AI智能分析给出修复建议。适用于开发者发布前安全自检、代码安全review。
G620710
@g620710
What This Skill Does
Scans AI agent and skill source code for security vulnerabilities including secret leaks, injection risks, permission issues, and data exposure. Combines static pattern matching with AI-powered analysis to produce severity-graded reports with fix recommendations.
Replaces manual code review for security issues by automating detection of 20+ vulnerability types and providing AI-generated remediation advice.
When to Use It
- Audit a Skill for sensitive information leaks before publishing to a marketplace
- Review third-party agent code for security risks before installation
- Run a scheduled security scan on your agent project repository
- Check for hardcoded API keys, passwords, or credentials in source files
- Assess prompt injection and tool abuse risks in an AI agent implementation
- Generate a compliance report with severity breakdown for security audit
Install
$ openclaw skills install @g620710/ai-agent-security-scannerAI Agent 安全审计工具
AI Agent和Skill代码安全审计利器,一键扫描安全漏洞,AI智能分析给出可落地的修复方案。帮你在发布前堵住安全漏洞,避免密钥泄露、注入攻击、数据泄露等风险。
When to Use
- Skill发布前做安全自检
- 第三方Skill安装前安全审查
- 代码仓库定期安全巡检
- AI Agent项目安全评估
- 排查代码中的敏感信息泄露
- 合规审计和安全加固
Core Rules
- 静态扫描 + AI深度分析双层检测机制
- 支持单文件和整个目录递归扫描
- 覆盖20+种常见安全漏洞类型
- 按严重程度分级:严重/高危/中危/低危
- 输出清晰的修复建议和代码示例
- 支持JSON和文本两种报告格式
Quick Start
# 扫描单个文件
python3 {baseDir}/scripts/audit.py path/to/your/skill.py
# 扫描整个Skill目录
python3 {baseDir}/scripts/audit.py path/to/skill/directory
# 只看严重和高危问题
python3 {baseDir}/scripts/audit.py ./ --severity high
# 输出JSON格式报告
python3 {baseDir}/scripts/audit.py ./ --format json -o report.json
# 跳过AI分析(快速扫描)
python3 {baseDir}/scripts/audit.py ./ --no-ai
Input Parameters
| 参数 | 类型 | 必填 | 说明 |
|---|---|---|---|
| target | string | 是 | 要扫描的文件或目录路径 |
| --no-ai | flag | 否 | 跳过AI深度分析,只做静态扫描 |
| --format | string | 否 | 输出格式:text/json,默认text |
| -o, --output | string | 否 | 输出报告文件路径 |
| --severity | string | 否 | 过滤级别:critical/high/medium/low/all,默认all |
| --api-key | string | 否 | DeepSeek API Key,也可通过环境变量设置 |
| --base-url | string | 否 | DeepSeek API Base URL |
Detection Capabilities
🔴 敏感信息泄露检测
- API密钥(OpenAI/DeepSeek/AWS/GitHub等格式)
- 密码和凭证
- 私钥文件
- JWT令牌
- 数据库连接串
- 手机号、身份证号等个人信息
🟠 注入攻击检测
- 命令注入(os.system/subprocess shell=True)
- 代码注入(eval/exec)
- SQL注入(字符串拼接SQL)
- 路径遍历漏洞
- 不安全的反序列化(pickle)
🟡 权限与配置风险
- 硬编码管理员凭证
- 不安全的文件权限(777)
- 调试模式开启
- CORS通配符配置
- HTTPS证书校验关闭
🤖 AI Agent 特有风险
- 提示注入漏洞
- 工具滥用风险
- 数据泄露风险
- 越权操作风险
- 输出过滤缺失
Output Format
扫描报告包含:
- 风险概览 - 各严重级别问题数量统计
- 总体风险评级 - 严重/高/中/低
- 问题明细 - 按严重程度排序,包含文件、行号、代码、描述
- AI深度分析 - 智能风险评估、修复方案、加固建议
Advanced Usage
CI/CD 集成
# 在CI流水线中使用,发现严重问题时失败退出
python3 {baseDir}/scripts/audit.py ./src --severity critical
if [ $? -ne 0 ]; then
echo "❌ 发现严重安全问题,构建终止!"
exit 1
fi
定期安全巡检
# 生成每日安全报告
python3 {baseDir}/scripts/audit.py /path/to/project -o reports/audit_$(date +%Y%m%d).json
批量审查多个Skill
for skill in ./skills/*/; do
name=$(basename "$skill")
echo "扫描: $name"
python3 {baseDir}/scripts/audit.py "$skill" --no-ai -o "reports/${name}_audit.txt"
done
API Configuration
需要配置DeepSeek API密钥(用于AI深度分析):
- 环境变量:
DEEPSEEK_API_KEY - 可选:
DEEPSEEK_BASE_URL(自定义API地址)
静态扫描无需API Key,只有AI深度分析需要。
References
references/owasp_top10.md- OWASP Top 10安全风险references/security_checklist.md- AI Agent安全检查清单references/fix_examples.md- 常见漏洞修复示例代码
Limitations
- 静态扫描基于正则匹配,可能存在误报和漏报
- AI分析仅供参考,关键问题建议人工复核
- 不支持二进制文件和编译后代码扫描
- 不支持动态运行时漏洞检测
Top skills in this category
Skill Vetter
@spclaudehomeSecurity-first skill vetting for AI agents. Use before installing any skill from ClawdHub, GitHub, or other sources. Checks for red flags, permission scope, and suspicious patterns.
Skill Vetter - Pre-Install Security Review
@donovanpankratz-delSecurity vetting protocol before installing any AI agent skill. Red flag detection for credential theft, obfuscated code, exfiltration. Risk classification L...
Skill Vetter 1.0.0
@fedrov2025Security-first skill vetting for AI agents. Use before installing any skill from ClawdHub, GitHub, or other sources. Checks for red flags, permission scope,...
Clawdhub
@steipeteUse the ClawdHub CLI to search, install, update, and publish agent skills from clawdhub.com. Use when you need to fetch new skills on the fly, sync installed skills to latest or a specific version, or publish new/updated skill folders with the npm-installed clawdhub CLI.
Evolver
@autogame-17A self-evolution engine for AI agents. Analyzes runtime history to identify improvements and applies protocol-constrained evolution. Communicates with EvoMap...