安全扫描
OpenClaw
可疑
medium confidenceThe skill's instructions are mostly consistent with a 'web tools' helper, but it explicitly guides the agent to request user API keys and to modify/ restart the local OpenClaw gateway configuration — actions that change system state and store secrets and therefore deserve caution.
评估建议
This skill is largely a sensible decision-flow for choosing web_search/web_fetch/browser, but it also instructs the agent to ask you for API keys and to apply them by running 'openclaw config set' and restarting the gateway (or using docker/pkill). Before installing or using it: 1) Don't paste sensitive API keys unless you trust the skill and its publisher — there is no homepage and the source is unknown. 2) Prefer to perform the configuration yourself if you're uncomfortable with automated comm...详细分析 ▾
ℹ 用途与能力
The skill is a guidance/instruction pack for choosing web_search → web_fetch → browser. Requesting a search API key and enabling a web search plugin is coherent with that purpose. However, the SKILL.md goes beyond passive guidance and includes concrete steps to write secrets into openclaw.json and restart the gateway — capabilities that are powerful and not strictly necessary for a purely advisory 'guide' skill.
⚠ 指令范围
The runtime instructions tell the agent to read local reference files (expected) but also to prompt the user for API keys, then execute shell commands to set configuration entries and restart the gateway (openclaw config set, openclaw restart gateway, docker restart, pkill). Those are system-level actions (writing config, restarting processes) that modify persistent agent behavior and could be disruptive if performed without explicit, informed user consent.
✓ 安装机制
Instruction-only skill with no install spec and no bundled code — nothing is downloaded or written by an installer. This minimizes code-execution risk from third-party packages.
ℹ 凭证需求
The skill asks the user to provide a web-search API key (Tavily or Kimi) and prescribes saving it into openclaw.json. Asking for a single, service-specific API key is proportionate to enabling web_search, but API keys are sensitive and the skill instructs persistent storage and service restarts, so users should confirm trust before providing secrets.
⚠ 持久化与权限
The SKILL.md explicitly instructs configuring openclaw.json to enable plugins and restarting the gateway, which changes system-wide agent/plugin configuration and grants persistent capability to the platform. The skill modifies other components' config (plugins.entries.*) and suggests killing/restarting processes — this is a privileged operation and should be performed only with user understanding and consent.
安装前注意事项
- Don't paste sensitive API keys unless you trust the skill and its publisher — there is no homepage and the source is unknown.
- Prefer to perform the configuration yourself if you're uncomfortable with automated commands that write openclaw.json or restart services.
- If you do provide a key, verify where it will be stored (openclaw.json) and who/what can read it.
- Watch for the agent's confirmation prompts — the skill requires explicit user confirmation before performing config changes, but confirm text matches what you expect.
- Consider backing up your openclaw config before allowing changes. If you need higher assurance (e.g., organization policy), treat this skill as requiring manual configuration rather than handing secrets to an automated assistant.
安全有层次，运行前请审查代码。
运行时依赖
无特殊依赖
版本
latestv1.0.12026/4/1
● 可疑
安装命令 点击复制
官方npx clawhub@latest install web-tools-guide
镜像加速npx clawhub@latest install web-tools-guide --registry https://cn.clawhub-mirror.com
技能文档
# Web 工具策略
遵循 ReAct 范式，按从轻到重、从通用到具体选择工具：
Level 1: web_search — 关键词搜索，快速获取信息入口
Level 2: web_fetch — 已知 URL，直接获取静态内容
Level 3: browser — 浏览器自动化，处理复杂网页操作
逐级升级，每次升级告知用户原因，不要静默切换。
决策流程
有明确 URL？
├─ YES → 静态内容（文章/文档/API/RSS）？→ web_fetch
│ 需要 JS 渲染/登录/交互/截图？ → browser
└─ NO → web_search
├─ 成功 → 对结果 URL 按上述逻辑选 fetch/browser
└─ 失败 → 引导配置（见"web_search 失败处理"）
Level 1: web_search
何时用：没有明确 URL，需要搜索信息（新闻、热点、查资料、比较信息）。
怎么用：直接调用 web_search，传入搜索关键词。
结果处理：返回的 URL 按决策流程选 web_fetch 或 browser 深入获取。
失败时：见下方"web_search 失败处理"。
Level 2: web_fetch
何时用：已知 URL，页面为静态内容——新闻文章、博客、技术文档、API 端点、RSS 源。
怎么用：直接调用 web_fetch，传入 URL。
失败信号：返回空白页、403、CAPTCHA、骨架 HTML → 说明需要 JS 渲染或登录态，告知用户后升级到 browser。
Level 3: browser
这是最重量级的工具，也是当前问题最多的场景。以下是详细操作指引。
何时用
- JS 渲染页面：SPA、动态加载内容（微博 feed、知乎回答、小红书瀑布流）
- 需要登录态：登录后才可见的内容、管理后台
- 页面交互：点击按钮、填写表单、翻页、滚动加载更多
- 截图需求：需要页面视觉信息
- web_fetch 失败的兜底：前一级工具无法获取有效内容
操作流程
信息获取（只读）：
- 导航到目标 URL
- 等待关键元素出现（不要用固定时间等待）
- 提取所需内容（文本、链接、图片等）
- 返回结果给用户
登录操作：
- 查找登录页 URL →
read {baseDir}/references/well-known-sites.json - 告知用户即将执行登录操作，获取确认
- 导航到登录页
- 填写凭证（用户提供）或提示用户扫码
- 等待登录成功，确认后继续后续操作
页面交互：
- 导航到目标页面
- 使用 CSS 选择器定位元素（辅以文本内容匹配）
- 执行交互：点击、输入、选择、滚动
- 等待响应/页面变化
- 提取结果或截图
关键注意事项
- 登录操作必须获得用户授权 — 任何涉及账号登录的操作前，先告知用户并等待确认
- 敏感操作必须二次确认 — 发帖、删除、支付等不可逆操作
- 优先 CSS 选择器 — 比 XPath 更稳定，辅以文本匹配
- 智能等待 — 等待目标元素出现，而非
sleep(3)式固定等待 - CAPTCHA/验证码 — 无法自动处理时告知用户需手动介入
- 页面加载超时 — 设置合理超时，失败时告知用户并建议重试
- 多步操作保持状态 — 登录后的后续操作复用同一浏览器上下文，不要重新打开
web_search 失败处理
当 web_search 返回错误时，不要静默降级，必须引导配置：
read {baseDir}/references/web-search-config.md- 按文件中 Step 1 原样输出配置引导给用户（不要改写表格或省略内容）
- 等待用户回复：
read {baseDir}/references/web-search-config.md，按 Step 2-5 执行
- 用户说"暂不配置" → 进入降级方案
- 其他回复 → 正常响应
- 降级方案（仅在用户明确拒绝配置后）：
read {baseDir}/references/well-known-sites.json 获取常用网站 URL
- 用 web_fetch 直接获取目标网站内容
- 仍不行 → 升级到 browser常用网站
需要常用网站 URL 时（登录页、搜索引擎、热搜榜等）：
read {baseDir}/references/well-known-sites.json
通过 key 查找（如 social.weibo.login、search.baidu）。带 {query} 的 URL 替换为实际搜索词。
数据来源：ClawHub ↗ · 中文优化：龙虾技能库
OpenClaw 技能定制 / 插件定制 / 私有工作流定制
免费技能或插件可能存在安全风险，如需更匹配、更安全的方案，建议联系付费定制