Back to Details

skill-vetter

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Skill Vetter 🔒

Skill Vetter 🔒

Never install a skill without vetting it first.
切勿在未审查的情况下安装任何技能。

When to Use

使用场景

  • Before installing any skill from ClawHub, GitHub, or external sources
  • When evaluating skills shared by other agents
  • 在从ClawHub、GitHub或外部来源安装任何技能前
  • 评估其他Agent分享的技能时

Step 1: Source Check

步骤1：来源检查

  • Where did it come from?
  • Author known/reputable?
  • Downloads/stars?
  • Last updated?
  • 它来自哪里？
  • 作者是否知名/可信？
  • 下载量/星标数？
  • 最后更新时间？

Step 2: Code Review (MANDATORY)

步骤2：代码审查（强制要求）

Read ALL files. Reject immediately if you see:
🚨 REJECT:
• curl/wget to unknown URLs
• Sends data to external servers
• Requests credentials/tokens/API keys
• Reads ~/.ssh, ~/.aws, ~/.config without reason
• Accesses MEMORY.md, USER.md, SOUL.md, IDENTITY.md
• base64 decode
• eval()/exec() with external input
• Modifies system files outside workspace
• Installs unlisted packages
• Network calls to IPs instead of domains
• Obfuscated/minified code
• Requests sudo
• Accesses browser cookies/sessions
• Touches credential files
阅读所有文件。如果发现以下内容，立即拒绝：
🚨 REJECT:
• curl/wget to unknown URLs
• Sends data to external servers
• Requests credentials/tokens/API keys
• Reads ~/.ssh, ~/.aws, ~/.config without reason
• Accesses MEMORY.md, USER.md, SOUL.md, IDENTITY.md
• base64 decode
• eval()/exec() with external input
• Modifies system files outside workspace
• Installs unlisted packages
• Network calls to IPs instead of domains
• Obfuscated/minified code
• Requests sudo
• Accesses browser cookies/sessions
• Touches credential files

Step 3: Permission Scope

步骤3：权限范围

  • Files read/written?
  • Commands run?
  • Network access? To where?
  • Scope minimal for stated purpose?
  • 读取/写入哪些文件？
  • 执行哪些命令？
  • 网络访问权限？访问哪些地址？
  • 权限范围是否与宣称的用途匹配且最小化？

Step 4: Risk Classification

步骤4：风险分级

LevelExamplesAction
🟢 LOWNotes, weather, formattingBasic review, install OK
🟡 MEDIUMFile ops, browser, APIsFull code review
🔴 HIGHCredentials, trading, systemHuman approval required
⛔ EXTREMESecurity configs, rootDo NOT install
等级示例操作
🟢 低风险笔记、天气、格式处理基础审查，可安装
🟡 中风险文件操作、浏览器、API调用完整代码审查
🔴 高风险凭证处理、交易、系统操作需要人工审批
⛔ 极高风险安全配置、root权限禁止安装

Output

输出结果

SKILL VETTING REPORT
═══════════════════════
Skill: [name] | Source: [origin] | Author: [name] | Version: [ver]
Metrics: [downloads/stars] | Updated: [date] | Files: [count]
Red Flags: [None / list]
Permissions: Files: [list] | Network: [list] | Commands: [list]
Risk: [🟢/🟡/🔴/⛔]
Verdict: [✅ SAFE / ⚠️ CAUTION / ❌ DO NOT INSTALL]
Notes: [observations]
═══════════════════════
SKILL VETTING REPORT
═══════════════════════
Skill: [name] | Source: [origin] | Author: [name] | Version: [ver]
Metrics: [downloads/stars] | Updated: [date] | Files: [count]
Red Flags: [None / list]
Permissions: Files: [list] | Network: [list] | Commands: [list]
Risk: [🟢/🟡/🔴/⛔]
Verdict: [✅ SAFE / ⚠️ CAUTION / ❌ DO NOT INSTALL]
Notes: [observations]
═══════════════════════

Quick Vet (GitHub)

GitHub快速审查

bash
curl -s "https://api.github.com/repos/OWNER/REPO" | jq '{stars: .stargazers_count, forks: .forks_count, updated: .updated_at}'
curl -s "https://api.github.com/repos/OWNER/REPO/contents/skills/SKILL_NAME" | jq '.[].name'
bash
curl -s "https://api.github.com/repos/OWNER/REPO" | jq '{stars: .stargazers_count, forks: .forks_count, updated: .updated_at}'
curl -s "https://api.github.com/repos/OWNER/REPO/contents/skills/SKILL_NAME" | jq '.[].name'

Trust Hierarchy

信任层级

  1. Official OpenClaw → lower scrutiny (still review)
  2. High-star repos (1000+) → moderate
  3. Known authors → moderate
  4. Unknown sources → maximum
  5. Requesting credentials → human approval always
  1. 官方OpenClaw → 低强度审查（仍需审查）
  2. 高星标仓库（1000+星）→ 中等强度审查
  3. 知名作者 → 中等强度审查
  4. 未知来源 → 最高强度审查
  5. 请求凭证 → 始终需要人工审批

Completion Checklist

完成检查清单

  • Source checked
  • All files read, red flags checked
  • Permission scope assessed
  • Risk level assigned
  • Verdict delivered
Unchecked = not done.
Paranoia is a feature. 🔒🦀
  • 来源已检查
  • 已阅读所有文件，完成红警项检查
  • 已评估权限范围
  • 已分配风险等级
  • 已给出审查结论
未勾选=未完成。
谨慎是一种优势。 🔒🦀

Credits

致谢

Originally by adamb0mbNZ — ClawHub.
最初由adamb0mbNZ创作 — ClawHub.