Firecrawl：把网页抓取与搜索写进 Agent 的 Web 数据 Skill
Firecrawl（@user_b22a52ba/firecrawl，v2.0.0）是 SkillHub 上的 data-analysis 社区技能，将 Firecrawl 的 Scrape、Search、Map、Crawl、Monitor 及结构化 extract 等能力写入 SKILL.md，供 Agent 在网页抓取与搜索场景按统一约定调用。需配置 FIRECRAWL_API_KEY，通过 skillhub install 或安装文档装入各客户端 skills 目录。本文基于官方 Skill 文档与 SkillHub 页面整理安装、端点与典型 Python 示例，并说明异步 Crawl、计费与反爬等注意点。
Read MoreWeb访问技能：为 AI Agent 统一联网与浏览器 CDP 操作指引
Web访问技能（@user_3c6cb52e/web-access-png）是 SkillHub 社区版 1.0.0 技能包，用 SKILL.md 规范 Agent 在 WebSearch、WebFetch、curl 与浏览器 CDP 之间的选型，并给出 CDP Proxy（localhost:3456）调用、本地书签历史检索与子 Agent 分治等约定。本文基于官方页面与 skillhub 安装包核对功能与安装命令，说明 Cursor/Codex 等 skills 目录配置，并提示当前下载包可能仅含指引文件、脚本需以商店或 GitHub 仓库更新为准。
Read More网页抓取工具箱：标准库实现的 Agent 网页采集 Skill
网页抓取工具箱（@user_f28fdad8/oc-web-fetch）是 SkillHub 社区 Skill，用四个 Python 脚本完成单 URL 抓取、批量抓取、元素提取与 HTML 转 Markdown，仅依赖标准库、无需 API Key。文章说明其定位与四类能力、通过 skillhub install 安装到 Cursor 等客户端 skills 目录的方法，并给出官方文档中的 JSON 入参示例与适用边界（无 JS 渲染、简化选择器等）。
Read More文章获取爬虫：从政务与新闻站点采集 AI 与数字化政策资讯
文章获取爬虫（@user_6ed3129a/capoocat，v1.0.0）是 SkillHub 社区 data-analysis 类 Skill，无需 API Key。面向中国政府与新闻网站，按人工智能、数字化、信息化等关键词抓取并排除招标公示类噪声，输出 original_articles、article_links 与 crawled_links.json。可通过 skillhub install 安装到 Cursor 等 skills 目录，或在对话中让 Agent 按场景触发；亦可 pip 安装 requests、beautifulsoup4、lxml 后运行 crawler_skill.py。
Read More公众号抓取 skill：从链接抓取正文并写入飞书文档
公众号抓取 skill（@user_5223609c/scrapling-article-fetch）是 SkillHub 社区版 1.0.0 技能，用 Scrapling 与 html2text 从 URL 提取可读正文（含图），按优先级选择器定位并按字符数截断，再写入飞书文档并返回链接，适合 mp.weixin.qq.com 及一般文章链接的快速验收。本文依据官方页面说明安装方式（skillhub install 与 AI 安装提示词）、核心能力与适用注意，并给出 Cursor、Claude Code、Codex 等 skills 目录参考。
Read MoreBrowserSkill 浏览器自启动器：让 bsk 抓取任务自己拉起 Edge
BrowserSkill（bsk）抓取与自动化要求本机 Edge 已开且扩展连上 daemon，无人值守任务常因 browsers connected 0 中断。SkillHub 社区 Skill「BrowserSkill 浏览器自启动器」（@user_656c2e70/bsk-browser-launcher，v1.0.0）通过 launch_browser.py 做 bsk status 自检、按需启动 Edge、轮询等待连接，并以退出码 0/3/4 约定就绪、超时与建会话失败；可选 --start-session 输出 SESSION=id。文章说明安装命令、典型 bsk 会话骨架及 Windows 路径依赖。
Read More网页抓取与内容理解：多策略级联的网页采集与理解 Skill
SkillHub 社区 Skill「网页抓取与内容理解」（@user_f12a44b7/self-dev-web-scraper，v1.0.7）面向公开网页的正文与元数据采集。它从轻量 HTTP 起步，按需回退到 HTML 解析与 Playwright 渲染，并集成新闻检测、样板移除、结构化元数据提取与 LLM 实体识别。安装方式为按 skillhub.md 安装坐标包；通过「抓取网页」「批量爬取」等自然语言触发。不适用登录站（需自配 Cookie/Session）、API 文档站与大文件下载。
Read More自适应网络工具向导·上网助手：按意图路由网页任务
网络搜索工具·上网助手（@laoxi/adaptive-web-router-shangwangzhushou，v1.0.2）是劳希在 SkillHub 发布的 ai-agent 类技能。它根据意图、能力、风险与成本，将搜索、读 URL、热点新闻、JS 渲染、登录页、截图与网页交互等任务路由到 web_search、web_fetch、opencli、browser 或宿主等价工具，并强调最小权限、有限重试、进度说明与来源核验，支持从网页工具故障中恢复。不处理纯本地文件，也不自动安装软件。安装可参考 skillhub.cn/install/skillhub.md，使用 skillhub install 并指定 Agent 的 skills 目录。
Read Morefirecrawl-build-interact: Assembling Multi-Step Firecrawl /interact Browser Workflows in Business Code
firecrawl-build-interact is an official Firecrawl Agent Skill used to assemble Firecrawl /interact workflows in business code. When /scrape cannot handle logins, pagination, form filling, infinite scrolling, or JavaScript single-page applications, it encapsulates patterns such as steps, waits, and authenticated navigation, allowing multi-step browser operations to be treated as a single call. The installation command is npx skills add with the target repository and skill name. It is suitable for scenarios such as authenticated scraping, pagination collection, and multi-step form processing.
Read Moredsh-wsl-playwright: Scraping Page Titles and Body Content with Headless Playwright on WSL
The DeepSeek Harness (DSH) plugin `dsh-wsl-playwright` is designed for running headless Playwright browsers in a WSL environment, addressing environment isolation and permission issues associated with running browsers on Windows. It relies on system-installed npx and Node.js and does not include browser binaries. Its core features include `pw_status` for environment checks and `pw_fetch` for retrieving a webpage's title and main content. Users can install and use it with simple commands, while being mindful of DSH version requirements and avoiding conflicts with interactive browsers.
Read Morebetter-crawler-4-agent
better-crawler-4-agent is an MCP server designed for web scraping challenges facing intelligent agents. It uses real browser rendering to address issues where HTTP scraping returns empty page shells or is blocked by anti-bot measures. The plugin provides three tools: single-page scraping, concurrent scraping (supporting 20 concurrent tasks), and status checking. It adopts a tiered fallback strategy that prioritizes browser rendering and automatically falls back to trafilatura or httpx on failure. The plugin is compatible with clients such as DSH, Claude Code, and Cursor, but users should pin the Playwright version and respect concurrency limits.
Read Moredsh-tavily-pool: Tavily-Powered DeepSeek Harness Multi-Key Search Plugin
The DeepSeek Harness plugin `dsh-tavily-pool` addresses rate-limiting issues caused by relying on a single API key for search. Built on Tavily, it ensures high availability for search and scraping through multi-key pool management, balance-aware rotation, and automatic failover. Its core features include bulk key management, call monitoring, and standalone toggles, with support for plain-text scraping and parameter configuration. It is easy to install, has no runtime dependencies, and is suitable for developers who require high-concurrency search capabilities.
Read Moredsh-web-fetch-enhanced: Whitelisting Non-Public Addresses for DeepSeek Harness Web Fetch
`dsh-web-fetch-enhanced` is a security-enhanced plugin for DeepSeek Harness, designed to address the issue that the native `web_fetch` rejects non-public addresses by default and therefore cannot access trusted intranet services. By default, the plugin rejects non-public targets and allows only explicitly matched CIDRs or domain allowlists. Core features include compatibility with the native tool, instant configuration through the web interface, a dual-layer allowlist mechanism (CIDRs + domains), DNS validation, per-hop redirect re-verification, and resource limit controls. Users can configure allowed network ranges and domains via the DSH Web interface. After installation, the Agent can securely read intranet documents or access private services, while the plugin strictly restricts the request scope, disallows Cookies, URL credentials, and excessive allowlisting, and automatically injects an authorization statement for the model.
Read Morexby-fetch: DSH Web Content Scraping Plugin
`xby-fetch` is an MCP server plugin for DeepSeek Harness (DSH) agents, addressing the lack of internet connectivity in earlier versions. It supports converting web page HTML to Markdown for easier content processing. After installation, an API key must be set via a command to gain access. The plugin retrieves content from a specified URL using the `fetch` command, enabling the agent to obtain the latest information. Note that users must register on the Xiaobenyang official website to obtain the API key, and it is recommended to review the source code and license before installation.
Read Morezoahdev/dsh-firecrawl: Web Scraping and Search via the Firecrawl API
`zoahdev/dsh-firecrawl` is a community plugin for DeepSeek Harness that serves as a bridge to Firecrawl, providing web scraping and search capabilities. It converts URLs to Markdown via API and supports complex pages. To install it, use `dsh plugin --profile web add dsh-firecrawl` and configure the `apiKey`. You can retrieve content by calling tools such as `firecrawl_scrape` and `firecrawl_search`. Before use, note that a valid API key is required, it relies on specific versions, it does not support streaming, and it is an unofficial project.
Read Moredsh-cloudflare-browser-run: Provide real browser crawling capabilities for DeepSeek Harness
dsh-cloudflare-browser-run is a DSH plugin maintained by RealAlexandreAI that integrates Cloudflare Browser Run's headless Chrome into DSH, addressing the limitation of standard HTTP scraping in handling JavaScript-rendered pages. The plugin provides three tools: browse (returns clean markdown by default, with optional screenshots or PDF), screenshot, and pdf. Its design features include real browser access and SSRF protection.
Read Moredsh-web-fetch: CDP Rendering + Tavily Extraction, Dual-source Web Scraping Plugin
dsh-web-fetch is an MIT-licensed internet plugin for DeepSeek Harness (DSH) developed by runfali. It solves two pain points in web scraping: pages requiring real rendering, such as SPAs, and...
Read Moredshsearch-multi: Multi-engine Web Search and Web Scraping Plugin for DSH
dshsearch-multi is a DSH (DeepSeek Harness) plugin developed by abcdream-Lary, licensed under the MIT License, aiming to provide unified multi-engine web search and content extraction capabilities for agents. The plugin is registered via the standard Provider interface, requiring no modification to the DSH source code, and supports five major engines: AnySearch, Bing, DuckDuckGo, Tavily, and Exa. Notably, the first three do not require an API Key, while the latter two require configured keys; if a search fails, it automatically switches to available engines as a fallback. The plugin also provides `web.fetch` web scraping functionality and deterministic query intent recognition without invoking an LLM. Users can configure it via the DSH Web settings page or `settings.yaml`...
Read Moredsh-scrape-webpage: Web Scraping and Statistical Analysis Plugin
`dsh-scrape-webpage` is a Host-only, zero-dependency, MIT-licensed plugin within the DeepSeek Harness ecosystem, designed to automate webpage content scraping and lightweight analysis. The plugin supports HTTP/HTTPS protocols and can extract titles, descriptions, body text, H1-H6 structures, and links (up to 100), along with outputting statistical information such as character count, reading time, language tendencies, and the top 30 high-frequency keywords. Its core advantage lies in the integration of image downloading and image recognition extension features. The plugin can download up to six page content images to the local environment and supports connecting independent image recognition plugins through the `scrape.imageAnalyzer` service interface, enabling the附加 output of visual analysis results for images. In terms of security, it runs by default within a sandbox environment, and if HTTPS encounters issues...
Read MoreIntegrate Web Search, Crawling and Browser Interaction into Application Code with firecrawl-build
`firecrawl-build` is the official application integration entry Skill of Firecrawl, located in the `firecrawl/skills` repository. It targets scenarios where products, agents or workflows need to obtain webpage data through code. It first clarifies what data to fetch and how to fetch it, then routes the requirements to `/scrape`, `/search` or `/interact`, instead of mixing "help me search a page in the current session" with "integrate Firecrawl into the backend". This article checks the installation commands, environment variables, SDK, sub-skill division, smoke test, and the boundaries with CLI/Workflow skills against the SKILL.md, repository README and official documentation.
Read More# 翻译结果： firecrawl-cli: Let AI Agents Scrape, Search, Crawl and Map Web Pages via CLI
firecrawl-cli is the official Firecrawl Agent Skill (located in the directory skills/firecrawl-cli, with the frontmatter named firecrawl). It enables assistants such as Cursor, Claude Code, and Codex to search, scrape, map, crawl and interact with web pages via the Firecrawl CLI, with results saved in .firecrawl/ in Markdown/JSON format. Based on the official SKILL.md and CLI documentation, this article explains the installation and authentication procedures, the workflow of search→scrape→map→crawl→interact, as well as key notes including credits and rate limits for users without an API key.
Read More