← 返回
未分类 中文

胖叔 Skill 安全审查

Security vetting for agent skills before installation. Scans skill code for dangerous Bash commands, sensitive file access, network exfiltration, obfuscated...
在安装前对代理技能进行安全审查,扫描代码中危险的 Bash 命令、敏感文件访问、网络泄露、混淆等风险。
hjshysst-dot
未分类 clawhub v1.0.0 1 版本 100000 Key: 无需
★ 0
Stars
📥 373
下载
💾 0
安装
1
版本
#latest

概述

Skill Vetter

Security scanner that analyzes skill code before installation.

When to Use

  • Before any skill installation via ClawHub or manual
  • Before any skill update (even trusted skills)
  • When user asks: "vet this skill", "check security", "scan before install"
  • When installing skills from unknown sources

Security Checks (Priority Order)

🔴 Critical (Block Install)

CheckPatternAction
------------------------
Disk wiperm -rf /, rm -rf ~, dd if=.*of=/dev/sdXBLOCK
Fork bomb`:(){ :\:& };:, fork()` loopBLOCK
Formatmkfs, newfs, umount -fBLOCK
SSH key deletionrm.\.ssh/, ssh-keygen.-DBLOCK
System takeoverchmod 777.*shadow, /etc/passwd editBLOCK

🟡 Medium (Warn + Allow)

CheckPatternAction
------------------------
Dangerous rmrm -rf [a-z]+\ (recursive without targets)WARN
Network exfilbase64 remote exfil, `curl.*\.*sh`WARN
Credential access.env, ~/.aws/, API key patternsWARN
Suspicious encodingObfuscated JavaScript, encoded commandsWARN
High-privilegesudo, chmod 777, setfaclWARN
Unknown networkNon-standard ports, suspicious domainsWARN

🔵 Info (Log Only)

CheckPattern
----------------
File writeWrites outside skill directory
Permission changeAny chmod
New file creationFile creation in system paths

Usage

Automatic Vetting (via Hook)

When OpenClaw hooks are configured, the vetter runs automatically:

# Configure in OpenClaw settings
skill_vetter:
  enabled: true
  auto_block: true   # Block critical issues
  warn_only: false    # false = block criticals

Manual Vetting

python3 ~/.openclaw/skills/skill-vetter/scripts/vetter.py \
    scan ~/.openclaw/skills/my-skill

Output Format

╔══════════════════════════════════════════════════╗
║ 🛡️ Skill Security Vetting Report                  ║
╠══════════════════════════════════════════════════╣
║ Skill: my-skill                                   ║
║ Scan Time: 2026-04-06 07:21:00                   ║
║                                                          ║
║ 🔴 CRITICAL: 0  🟡 WARNING: 2  🔵 INFO: 1            ║
╠══════════════════════════════════════════════════╣
║ Findings:                                              ║
║   (see actual scan output for details)                 ║
╠══════════════════════════════════════════════════╣
║ Verdict: ⚠️ INSTALL WITH CAUTION                       ║
╚══════════════════════════════════════════════════╝

Integration

For automatic pre-install vetting, this skill should be invoked by the OpenClaw skill installation hook. The hook configuration:

# In OpenClaw config
hooks:
  pre_skill_install:
    - name: skill-vetter
      action: scan_then_block
      block_on_critical: true

Verdict Logic

SeverityCountVerdict
--------------------------
🔴 Critical> 0🚫 BLOCK INSTALL
🟡 Warning> 3⚠️ WARN + CONFIRM
🟡 Warning≤ 3✅ INSTALL WITH CAUTION
Only 🔵 InfoAny✅ CLEAR TO INSTALL

Implementation

The scanner is in scripts/vetter.py. Key functions:

  • scan_skill(skill_path) - Main entry point
  • check_dangerous_commands(content) - Bash pattern matching
  • check_sensitive_access(content) - File/credential patterns
  • check_network_activity(content) - Exfil indicators
  • check_obfuscation(content) - Obfuscated code detection
  • generate_report(findings) - Formatted output

版本历史

共 1 个版本

  • v1.0.0 当前
    2026-05-07 13:54 安全 安全

安全检测

腾讯云安全 (Keen)

安全,无风险
查看报告

腾讯云安全 (Sanbu)

安全,无风险
查看报告

🔗 相关推荐

security-compliance

Skill Vetter

spclaudehome
AI智能体技能安全预审工具。安装ClawdHub、GitHub等来源技能前,检查风险信号、权限范围及可疑模式。
★ 1,215 📥 266,439
developer-tools

Github

steipete
使用 `gh` CLI 与 GitHub 交互,通过 `gh issue`、`gh pr`、`gh run` 和 `gh api` 管理议题、PR、CI 运行及高级查询。
★ 668 📥 324,054
ai-intelligence

ontology

oswalpalash
类型化知识图谱,用于结构化智能体记忆与可组合技能。支持创建/查询实体(人员、项目、任务、事件、文档)及关联...
★ 711 📥 243,738