← 返回 Skills 市场
free-web-search
作者
ucsdzehualiu
· GitHub ↗
· v7.0.8
· MIT-0
2218
总下载
1
收藏
10
当前安装
21
版本数
在 OpenClaw 中安装
/install free-web-search
功能描述
基于Bing国内版的稳定联网搜索工具，中文环境深度优化，支持全文内容抓取，绕过常见反爬限制，返回结构化搜索结果。
使用说明 (SKILL.md)
free-web-search v14 联网搜索工具
基于 Playwright 浏览器实现的稳定搜索工具，意图识别 + 请求节流 + 结果质量评分 + 保留CSS修复。
v14 更新内容
- ✅ [关键修复] 保留CSS：之前拦截CSS导致Bing搜索结果标题文字丢失
- ✅ 意图识别+query改写：搜索质量差时自动改写query（城市游玩→景点推荐、今日价格→实时行情等）
- ✅ 改写仅在质量差时触发：先搜原始query，质量好就不改写，避免改写搞坏本来好的query
- ✅ 请求节流：两次Bing请求间隔≥3s，避免触发限流
- ✅ 限流检测+退避：0结果时递增等待重试，排除重试也0结果时停止
- ✅
--filter回退：过滤后为空自动回退到不过滤结果 - ✅ 单域名排除重试：最多2轮，结果更好才替换
- ✅ DuckDuckGo国内快速失败：10s超时×1次
- ✅
--no-rewrite：禁用query改写（调试用）
核心能力
- ✅ 中文环境深度优化：强制 Bing 返回中文结果
- ✅ 反爬检测绕过：多层反检测措施（stealth.js）
- ✅ 全文抓取：支持按需抓取目标网页的完整正文内容
- ✅ Headless 模式：服务器可用，无需显示器
【核心必读】搜索Query优化规范
搜索效果的好坏，90%取决于Query是否合理，请严格遵循以下规则生成搜索词：
一、黄金原则
- 简洁精准：只保留核心关键词，用2-5个核心词组合，禁止长句、反问句、口语化描述
- 限定明确：需要时效性/领域/地区内容时，必须加上对应的限定词
- 格式正确：使用中文关键词 + 英文/数字限定词，禁止特殊符号、无意义助词
二、正确示例 vs 错误示例
| 搜索场景 | 正确Query（推荐） | 错误Query（禁止） |
|---|---|---|
| 时效性新闻 | 2026年04月 美伊局势 最新 | 你能帮我查一下最近美国和伊朗之间发生了什么事吗 |
| 技术教程 | Python 异步编程 最佳实践 2026 | 我想学习一下Python的异步编程，有没有好的教程 |
| 知识科普 | 中国大型邮轮 花城号 出坞 最新消息 | 中国的那个大型邮轮花城号现在怎么样了 |
| 本地内容 | 广东东莞 今日天气 | 我现在在东莞，今天天气怎么样啊 |
| 官方信息 | 华为云 ModelArts 官方文档 | 华为云的那个ModelArts的官网在哪里，文档怎么看 |
参数说明
| 参数名 | 类型 | 说明 | 默认值 | 取值限制 |
|---|---|---|---|---|
query |
字符串 | 【必填】搜索关键词 | - | 不能为空 |
max |
整数 | 最多返回的搜索结果条数 | 10 | 1-20 |
full |
整数 | 抓取前N条结果的网页全文 | 0 | 0-5 |
engine |
字符串 | 搜索引擎选择 | bing | bing/duckduckgo/auto |
filter |
布尔 | 过滤低质量域名 | false | - |
使用示例
# 基础搜索
python scripts/web_search.py "经济新闻 今日" --max=10
# 抓取前3条结果的全文
python scripts/web_search.py "经济新闻 最新" --full=3
# 使用 auto 模式（Bing 结果不足时切换 DuckDuckGo）
python scripts/web_search.py "技术教程" --engine=auto
# 过滤知乎等低质量域名
python scripts/web_search.py "某个话题" --filter
常见问题
搜索返回空结果
- 检查网络连接（VPN 可能影响 Bing 国内版）
- 尝试
--engine=duckduckgo直接用 DuckDuckGo - 检查 Query 是否过于冗长或口语化
浏览器启动失败
pip install playwright && playwright install chromium
全文抓取失败
- 某些网站有强反爬限制
- 知乎等域名在全文抓取时自动跳过
结果集中在单一域名
- 脚本会自动检测并警告
[WARN] 结果集中在单一域名 - 解决方案：换用更具体的关键词，避免歧义词
搜索关键词避坑指南
| ❌ 避免 | ✅ 推荐 |
|---|---|
民生新闻 |
住房 医疗 就业 或 社会政策 百姓生活 |
经济新闻 |
财经政策 GDP 或 A股 沪指 |
长护险 |
长期护理保险 养老服务 |
服务器环境
- 脚本强制使用
headless=True，无需显示器 - 已添加服务器兼容的浏览器参数
安全使用建议
This skill appears to do what it says: a headless Playwright-based scraper optimized for Chinese/Bing CN search. Before installing or running it, consider the following:
- Review the full scripts (especially the remainder of scripts/web_search.py) in a text editor to confirm there is no hidden network exfiltration or unexpected remote endpoints (the provided snippet was large but truncated).
- The tool includes anti-bot evasion (stealth JS and aggressive browser flags). That is coherent with scraping but may violate target sites' terms of service or local law; ensure you have permission to scrape and that usage complies with policies.
- setup.sh and the Python script will download and install packages and Chromium via Playwright; run these in an isolated environment (container, VM) if you are concerned about changes to the host or supply-chain risks.
- The code will make network requests to target websites—do not provide any secrets or credentials to this tool.
- If you plan frequent/large-scale use, add your own rate-limiting, logging, and monitor network traffic; consider using explicit proxies and obey robots.txt/ToS.
If you want higher assurance, provide the full web_search.py file (untruncated) for a complete line-by-line review; that could change the confidence level.
功能分析
Type: OpenClaw Skill
Name: free-web-search
Version: 7.0.8
The skill bundle provides a functional web search tool using Playwright to scrape Bing and DuckDuckGo. The scripts (scripts/web_search.py and scripts/setup.sh) include standard scraping features such as anti-bot detection bypass (stealth.js), query rewriting for better search quality, and automated dependency installation. All high-risk capabilities, such as browser automation and full-text extraction, are directly aligned with the tool's stated purpose of providing structured search results, and no evidence of data exfiltration or malicious prompt injection was found.
能力评估
Purpose & Capability
Name/description (Bing CN search with full-page fetch and anti-crawl measures) match the included Python scraper and setup script. Required resources (Playwright, Chromium) are exactly what a headless web-scraper needs; no unrelated credentials, binaries, or config paths are requested.
Instruction Scope
SKILL.md and the tool entry point instruct the agent to run scripts/web_search.py and optionally run setup instructions; the instructions are focused on search/query handling and scraping. The code actively implements anti-detection measures (stealth JS, browser args) and full-page fetching; while coherent for robust scraping, this is functionally invasive (bypassing anti-bot protections) and may raise legal/ToS concerns. SKILL.md does not ask the agent to read unrelated files or exfiltrate credentials.
Install Mechanism
There is no platform-level install spec in the registry, but a provided scripts/setup.sh installs Python packages via pip and runs 'playwright install chromium'. That is a standard but networked install: it will download packages and browser binaries from upstream. The runtime script also attempts to pip-install playwright if missing. These steps write to disk and fetch code from package indexes and Playwright releases, which is expected for this functionality but worth reviewing before running.
Credentials
The skill requires no environment variables, credentials, or config paths. The code does not declare or request secrets. This is proportionate to a public-web scraping/search tool.
Persistence & Privilege
Skill flags are default (not always), it does not request permanent platform presence or modify other skills. It will install runtime dependencies locally if the user runs setup, but does not attempt to persist in the agent configuration or escalate privileges.
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install free-web-search - 安装完成后，直接呼叫该 Skill 的名称或使用
/free-web-search触发 - 根据 Skill 的参数说明提供必要输入，即可获得结构化输出
版本历史
v7.0.8
free-web-search v7.0.8
- No file changes detected in this release.
- All features, usage and parameters remain unchanged from the previous version.
v7.0.7
free-web-search v7.0.7
- 更新描述和参数细节，强调中文环境优化和全文抓取能力
- 明确仅支持 Bing（国内版）和 DuckDuckGo，可自动选择
- 增加 Query 优化规范，推荐简洁精准关键词，禁止长句/反问句
- trigger_keywords 新增“是什么”
- 优化工具参数说明，调整部分默认值和最大值限制
- 移除新闻专用参数，突出中文环境和结构化结果特点
v7.0.6
free-web-search v10 — 新增多搜索引擎、新闻模式、纯 Bash 安装脚本等特性。
- 支持 Bing/Google/DuckDuckGo 多引擎，自动切换（auto 模式）
- 新增 --news 新闻模式，仅限新闻源并屏蔽知乎/百科噪声
- 新增 scripts/setup.sh 一键安装脚本
- 搜索全面升级为 httpx+BeautifulSoup，全文采用 Playwright 渲染
- 参数简化，使用更灵活（--max, --full, --engine, --news, --filter）
- Query 规范和实例文档更精炼、明确
v7.0.5
- 版本号从 12 降回 7。
- 其余功能与文档内容无变化。
v7.0.4
free-web-search v12 introduces several server and search quality improvements:
- 新增结果质量检测，自动警告结果集中在单一域名
- 页面渲染/超时/重试时间更长，提升稳定性
- 新增域名黑名单和 --filter 参数，可过滤知乎等低质量网站
- 支持通过 engine 参数选择 bing、duckduckgo 或 auto
- 全文抓取时自动跳过黑名单域名
- 增强 headless 模式，优化服务器环境兼容性
v7.0.3
free-web-search v7.0.3
- 更新触发关键词，收窄触发范围（移除部分泛用词，添加更精准词）
- 精简和优化了 SKILL.md 文档内容，突出查询优化规则和常用场景
- 简化“参数说明”和“使用示例”，适配更广泛的用户需求
- 删除“双引擎降级”等特定特性描述，强调 Playwright 方案和中文环境优化
- 增强关于 Query 规范、过滤技巧的说明，便于生成高效的搜索关键词
v7.0.2
Version 8 is a major upgrade focused on precision, usability, and flexibility:
- Deeply optimized search for Chinese environment, with enforced Chinese Bing results and precise query structuring.
- Added many new trigger keywords for richer natural language query support.
- Now supports structural search result output and full content fetching with fine-grained control (max/full arguments).
- Implements automatic fallback to requests+BeautifulSoup mode if Playwright is unavailable.
- Comprehensive user and agent documentation for query optimization, when to trigger a search, and recommended parameter usage.
- Removed unnecessary files for a cleaner build.
v7.0.1
- No changes detected in this version.
- Version number remains at 7.0.
- All features and documentation are unchanged.
v7.0.0
Version 7.0 is a major update focusing on improved reliability and search accuracy.
- Switched to Bing as the search engine for more up-to-date and stable results
- Automatically appends the current date to queries to ensure the freshest information
- Search parameters clarified and simplified, with better descriptions
- Expanded documentation with practical tips for optimizing Chinese search queries
- Tool parameters standardized for easier integration
v3.0.0
/
v9.3.4
No changes detected in this version.
- No file modifications or updates were made to the skill.
- Functionality and documentation remain the same as the previous version.
v9.3.3
free-web-search introduces major enhancements for faster and higher-quality web searches.
- Added parallel fetching (multi-threaded), making page retrieval up to 3× faster
- Introduced smart filtering for deduplication and low-quality result removal
- New --simple parameter for concise, readable output
- Connection reuse implemented to reduce network overhead
- Improved error handling with automatic network retries
- Enhanced Chinese encoding detection and support
- Skill documentation updated with new parameters and usage examples
v9.3.2
free-web-search 9.3.2
- 新增：全面重构说明文档，明确技能核心能力为“搜索并获取网页全文内容”
- 工具接口更新：引入 web_search 工具，支持获取搜索引擎摘要与前N个页面的完整内容
- 支持中文搜索和常见网页编码，优化编码检测多层次方案，强制 Windows 下 UTF-8 输出，解决乱码
- 参数调整：引入 --full 选项自定义抓取完整页面数量，默认前3个
- 增强结果结构化输出，包括标题、链接、预览摘要及全文
v9.3.1
- Initial skill version released under the new name and structure.
- Added detailed execution workflow for web search: from intent analysis, search execution, to structured result presentation.
- Supports Chinese and English trigger keywords.
- Integrates with the `general_search` tool for live web queries.
- Emphasizes accuracy, transparency, and proper citation of sources in result output.
v9.3.0
- Added support for recency-boosted searches with the new --recent command-line option.
- Documentation updated to include instructions for using the --recent flag for fresher search results.
- Version bumped to 9.3.0.
v9.0.8
- Version bumped from 9.0.0 to 9.2.0 in SKILL.md.
- No code or documentation changes detected beyond the version update.
v1.0.7
Version 1.0.7
- No code or documentation changes detected in this release.
- Skill version updated without additional modifications.
v1.0.6
free-web-search v9.0.0 is a major update focused on speed, simplicity, and agent integration.
- Switched to fast, lightweight httpx-based engine (no more browser dependencies)
- Now optimized for Claude Code / OpenClaw agents with improved JSON output
- Extended documentation with clear usage guidance, sample queries, and agent workflow
- Enhanced error handling and structured output for easier integration
- Updated dependencies (removed playwright, added httpx, lxml)
v1.0.5
- No code or documentation changes detected in this release.
- Version number updated from 1.0.5 to 8.0.0.
v1.0.4
- Added support for multi-engine search with automatic fallback from DuckDuckGo to Bing when needed.
- Improved intelligent content extraction, filtering ads and redundant information for cleaner results.
- Enhanced error handling for more robust searches.
- Introduced structured output and sorting by title relevance and authoritative sources.
- Added command-line options for JSON output and customizable page count.
元数据
常见问题
free-web-search 是什么？
基于Bing国内版的稳定联网搜索工具，中文环境深度优化，支持全文内容抓取，绕过常见反爬限制，返回结构化搜索结果。 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 2218 次。
如何安装 free-web-search？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install free-web-search」即可一键安装，无需额外配置。
free-web-search 是免费的吗？
是的，free-web-search 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
free-web-search 支持哪些平台？
free-web-search 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。
谁开发了 free-web-search？
由 ucsdzehualiu（@ucsdzehualiu）开发并维护，当前版本 v7.0.8。
推荐 Skills