← Back to Skills Marketplace
freeman88-tch

Web Tools Guide

by Freeman88-tch · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ⚠ suspicious
38
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install lobster-web-tools
Description
MANDATORY before calling web_search, web_fetch, browser, or opencli. Contains required error-handling procedures (web_search failure → must guide user to con...
README (SKILL.md)

\x3C!-- baseDir = /root/.openclaw/workspace/skills/web-tools-guide -->

Web 工具策略

遵循 ReAct 范式。四个工具不是层级关系,是分支决策

┌─ 没有 URL,需要搜索 ──────→ web_search   (关键词搜索)
│
├─ 已知 URL,静态内容 ──────→ web_fetch    (直取页面)
│
├─ 以上失败 / 不适用 ──────→ opencli      (CLI 结构化访问,70+ 站点)
│
└─ 全都不行 ───────────────→ browser      (浏览器自动化,兜底)

先按场景选 web_search 或 web_fetch;失败时先试 opencli,最后才上 browser。 每次切换工具告知用户原因,不要静默降级。


决策流程

有明确 URL?
├─ YES → 静态内容(文章/文档/API/RSS)?
│        ├─ YES → web_fetch
│        │        失败(空白/403/CAPTCHA)?→ opencli → browser
│        └─ NO(需要 JS/登录/交互/截图)→ opencli → browser
└─ NO  → web_search
         ├─ 成功 → 对结果 URL 按上述逻辑选 fetch/opencli/browser
         ├─ 失败(API 错误)→ 引导配置(见"web_search 失败处理")
         └─ 无结果/不适用 → opencli → browser

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_fetchopenclibrowser 深入获取。

失败时:见下方"web_search 失败处理"。


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 → 尝试 opencli,仍不行再升级到 browser


opencli(Fallback,优先于 browser)

何时用:web_search / web_fetch 失败或不适用时,先试 opencli 再考虑 browser。覆盖 70+ 主流网站,秒级返回结构化数据。

首次使用前:如果执行 opencli 提示 command not found,需要先运行安装脚本(幂等,可重复运行):

bash {baseDir}/scripts/setup-opencli.sh

该脚本会自动完成:安装 opencli CLI → 编译 Browser Bridge 插件 → 重启浏览器加载插件。

渐进式发现(不需要记命令)

opencli --help                    # 有没有这个站?
opencli \x3Csite> --help             # 这个站能做什么?
opencli \x3Csite> \x3Ccommand> --help   # 这个命令怎么用?

详细用法read {baseDir}/references/opencli-guide.md

失败时:告知用户 opencli 失败原因,降级到 browser。


browser(最后手段)

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
  • 其他工具全部失败的兜底

操作流程

信息获取(只读):

  1. 导航到目标 URL
  2. 等待关键元素出现(不要用固定时间等待)
  3. 提取所需内容(文本、链接、图片等)
  4. 返回结果给用户

登录操作:

  1. 查找登录页 URL → read {baseDir}/references/well-known-sites.json
  2. 告知用户即将执行登录操作,获取确认
  3. 导航到登录页
  4. 填写凭证(用户提供)或提示用户扫码
  5. 等待登录成功,确认后继续后续操作

页面交互:

  1. 导航到目标页面
  2. 使用 CSS 选择器定位元素(辅以文本内容匹配)
  3. 执行交互:点击、输入、选择、滚动
  4. 等待响应/页面变化
  5. 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

web_search 返回错误时,不要静默降级,必须引导配置:

  1. read {baseDir}/references/web-search-config.md
  2. 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)
  3. 等待用户回复:
    • 用户提供 API Key → 再次 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
  4. 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
    • 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.loginsearch.baidu)。带 {query} 的 URL 替换为实际搜索词。

Usage Guidance
Review the setup script before running it. Install only if you trust the opencli npm package and GitHub release source, preferably in a separate browser profile with no sensitive logged-in accounts. Provide API keys only if you are comfortable storing them in OpenClaw config, and require explicit confirmation before any login, posting, publishing, deleting, or account-changing action.
Capability Analysis
Type: OpenClaw Skill Name: lobster-web-tools Version: 1.0.0 The skill bundle provides advanced web interaction tools but employs high-risk procedures. The script 'scripts/setup-opencli.sh' installs a global npm package, downloads a third-party browser extension from GitHub (jackwener/opencli), and forcefully restarts the user's browser with that extension enabled by killing active processes and modifying startup arguments. Additionally, 'references/web-search-config.md' contains instructions for the agent to solicit, analyze, and store user API keys (Tavily/Kimi) in the local configuration. While these actions are aligned with the stated purpose of enhancing web capabilities, the automated installation of unverified browser extensions and the collection of secrets represent significant security risks.
Capability Tags
cryptorequires-sensitive-credentials
Capability Assessment
Purpose & Capability
The stated purpose is a web-search/fetch/browser/opencli guide, which fits much of the content, but the artifacts also include a runnable installer, browser extension loading, account/session use, optional write actions, and API-key configuration that go beyond a simple instruction-only guide.
Instruction Scope
The skill gives strong tool-routing instructions and requires user confirmation for login and irreversible actions. The confirmation language is helpful, but users should notice that the workflow can shift from web search to credential setup and local service restarts.
Install Mechanism
Although the registry says there is no install spec, the skill instructs running a setup script that globally installs an npm package and downloads a browser extension from GitHub releases without visible pinning, checksums, or signature verification.
Credentials
The setup and configuration instructions can modify the local OpenClaw environment, kill and restart Chrome, load a browser extension, and restart or kill gateway processes. Those actions are high-impact for a web-tools guide.
Persistence & Privilege
The skill can persist an opencli installation, a browser extension under /root/.openclaw, and Tavily/Kimi API keys in OpenClaw config, while also using logged-in browser contexts for sites.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install lobster-web-tools
  3. After installation, invoke the skill by name or use /lobster-web-tools
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
🦞 龙虾兵团自研融合技能 v1首发
Metadata
Slug lobster-web-tools
Version 1.0.0
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 1
Frequently Asked Questions

What is Web Tools Guide?

MANDATORY before calling web_search, web_fetch, browser, or opencli. Contains required error-handling procedures (web_search failure → must guide user to con... It is an AI Agent Skill for Claude Code / OpenClaw, with 38 downloads so far.

How do I install Web Tools Guide?

Run "/install lobster-web-tools" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is Web Tools Guide free?

Yes, Web Tools Guide is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does Web Tools Guide support?

Web Tools Guide is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created Web Tools Guide?

It is built and maintained by Freeman88-tch (@freeman88-tch); the current version is v1.0.0.

💬 Comments