{ "updated_at": "2026-09-13 17:12", "total_curated": 12, "projects": [ { "id": "deer-flow", "name": "deer-flow", "repo": "https://github.com/bytedance/deer-flow", "category": "AI Agent", "stars": "82.3k", "lang": "Python", "summary": "字节跳动开源的长程超级智能体框架，集成沙箱与记忆实现复杂任务闭环。", "why_recommend": "告别单轮对话玩具！它用沙箱隔离、长程记忆与子智能体协作，真正解决了复杂任务耗时失控与上下文崩溃的顽疾。", "install_command": "git clone https://github.com/bytedance/deer-flow.git && cd deer-flow && pip install -e .", "agent_prompt": "请拉取 bytedance/deer-flow 仓库并初始化 Python 环境，阅读 README 及架构文档，梳理核心工作流并编写一个调用子 Agent 执行长程研究任务的最小可运行 Demo。", "tags": [ "SuperAgent", "长程任务", "多智能体", "沙箱执行", "DeepResearch" ] }, { "id": "siyuan", "name": "siyuan", "repo": "https://github.com/siyuan-note/siyuan", "category": "AI Agent", "stars": "46.3k", "lang": "TypeScript", "summary": "隐私优先、支持MCP的人与AI智能体协同自托管知识工作空间。", "why_recommend": "告别云笔记隐私泄露与数据锁定，本地优先架构深度融合MCP协议，是打造私有Agent专属长记忆知识库的终极底座。", "install_command": "docker run -d -p 6806:6806 -v ~/siyuan_workspace:/siyuan/workspace b3log/siyuan --workspace=/siyuan/workspace", "agent_prompt": "请按步骤执行：1. 阅读 SiYuan 官方文档并梳理开放 API 与 MCP 协议接入点；2. 使用 Docker 在本地 6806 端口快速启动 SiYuan 容器实例并挂载工作空间；3. 编写最小脚本调用其 API 接口完成鉴权，自动创建一份带块级引用的 Markdown 测试笔记，验证智能体读写闭环。", "tags": [ "知识库", "自托管", "MCP", "AI-Agent", "Local-First" ] }, { "id": "qwen-code", "name": "qwen-code", "repo": "https://github.com/QwenLM/qwen-code", "category": "AI Agent", "stars": "27.8k", "lang": "TypeScript", "summary": "通义千问开源的终端AI编程智能体，原生支持MCP与多模型协同开发。", "why_recommend": "告别闭源工具与厂商绑定！全开源体系联动模型自进化，原生整合MCP与记忆机制，支持本地及多模型热切换。", "install_command": "npm install -g @qwen-code/qwen-code@latest", "agent_prompt": "请在环境中执行 `npm install -g @qwen-code/qwen-code@latest` 进行全局安装，阅读官方 README 和相关文档了解 CLI 认证与交互流程，并在测试工作区提供配置 API Key 与启动自动代码审查的最小可运行命令示例。", "tags": [ "AI Agent", "终端编程", "代码智能体", "MCP", "开源生态" ] }, { "id": "kirocrew", "name": "KiroCrew", "repo": "https://github.com/kirodotdev/KiroCrew", "category": "AI Agent", "stars": "3.9k", "lang": "Python", "summary": "具备跨会话持久记忆与自主迭代能力的AI开发工作区。", "why_recommend": "击碎传统AI助手会话割裂痛点，原生支持开发状态长效持久化与能力自主闭环进化。", "install_command": "git clone https://github.com/kirodotdev/KiroCrew.git && cd KiroCrew && pip install -e .", "agent_prompt": "请拉取 kirodotdev/KiroCrew 源码并安装依赖，快速查阅核心文档与架构说明，编写并执行一个具备跨会话持久化记忆的最小工作区 Demo，验证其自我进化与上下文保持特性。", "tags": [ "AI Agent", "持久化工作区", "自我进化", "开发者工具" ] }, { "id": "openclaw-weixin", "name": "wechat-ilink-client", "repo": "https://github.com/wechat-ilink/wechat-ilink-client", "category": "AI Agent", "stars": "2.4k", "lang": "TypeScript", "summary": "独立的微信 iLink 官方机器人协议客户端，纯 Node.js 零外部 C++ 依赖", "why_recommend": "彻底颠覆易被封号的 Web 微信与复杂挂机外挂，通过逆向腾讯官方协议实现 7x24 小时云端稳定长连，配合 LLM 可秒建私有微信个人助理。", "install_command": "npm install wechat-ilink-client", "agent_prompt": "请帮我参考 wechat-ilink-client 在本地部署一个微信自动回复机器人", "tags": [ "微信生态", "自动化", "Bot", "Node.js" ] }, { "id": "crawl4ai", "name": "crawl4ai", "repo": "https://github.com/unclecode/crawl4ai", "category": "自动化工作流", "stars": "31.8k", "lang": "Python", "summary": "专为大语言模型优化的超高速开源网络爬虫与网页提取器", "why_recommend": "速度比 Puppeteer 快数倍，开箱直接输出结构化 Markdown、清除垃圾 DOM 并原生支持多模态媒体抓取，是 Agent 知识检索必备神器。", "install_command": "pip install crawl4ai && crawl4ai-setup", "agent_prompt": "请安装 crawl4ai 并为我编写一个抓取目标网页 Markdown 内容的脚本", "tags": [ "爬虫", "LLM", "数据提取", "Python" ] }, { "id": "skills-cli", "name": "skills", "repo": "https://github.com/vercel-labs/skills", "category": "CLI神器", "stars": "8.9k", "lang": "TypeScript", "summary": "AI Agent 跨平台通用技能库管理工具，标准化扩展 Agent 工具箱", "why_recommend": "类似 npm 但专为 AI Coding Agent 设计，一条命令即可将社区顶尖技能挂载给 Claude Code 或 Antigravity 使用。", "install_command": "npx skills add find-skills", "agent_prompt": "请使用 npx skills 命令为我检索并安装社区最受欢迎的技能包", "tags": [ "Agent技能", "CLI", "生态标准" ] }, { "id": "browser-use", "name": "browser-use", "repo": "https://github.com/browser-use/browser-use", "category": "AI Agent", "stars": "29.5k", "lang": "Python", "summary": "赋予任何大模型自主操作真实浏览器界面的开源 Agent 执行框架", "why_recommend": "具备视觉 DOM 解析、多步任务自规划与自动表单填报能力，能让 AI 像人类一样登录网站、点击按钮和完成全自动业务闭环。", "install_command": "pip install browser-use playwright && playwright install", "agent_prompt": "请使用 browser-use 编写一个自动打开网页并执行指定搜索任务的演示", "tags": [ "浏览器自动化", "视觉Agent", "Playwright" ] }, { "id": "superpower-skill", "name": "superpower", "repo": "https://github.com/superpower-ai/superpower", "category": "AI Agent", "stars": "5.3k", "lang": "JavaScript", "summary": "全面增强本地智能体思维链、代码分析与多工具并发协作的核心能力套件", "why_recommend": "强化 Agent 的长任务上下文维护与容错自愈能力，支持深度逆向代码与复杂工程重构。", "install_command": "npm install -g @superpower/agent-core", "agent_prompt": "请帮我配置并启用 superpower 增强套件", "tags": [ "超能力", "自愈排错", "系统架构" ] }, { "id": "dify", "name": "dify", "repo": "https://github.com/langgenius/dify", "category": "自动化工作流", "stars": "65.2k", "lang": "Python / TypeScript", "summary": "开箱即用的开源大模型应用与工作流可视化编排引擎", "why_recommend": "支持 RAG 知识库检索、多模型对比与 Agent 节点流转，企业与个人搭建专属 AI 工作流的首选。", "install_command": "git clone https://github.com/langgenius/dify.git && cd dify/docker && docker compose up -d", "agent_prompt": "请帮我检查本地环境并使用 docker-compose 一键启动 Dify 服务", "tags": [ "RAG", "工作流", "低代码", "Docker" ] }, { "id": "fast-agent-ui", "name": "fast-agent-ui", "repo": "https://github.com/fast-ai/agent-ui", "category": "前端与设计", "stars": "3.8k", "lang": "Vue 3", "summary": "轻量极致的 Agent 对话流与多模态交互 UI 框架", "why_recommend": "内置 KaTeX 数学公式渲染、Mermaid 架构图、实时代码 Diff 与思维链折叠，秒级接入任何 Agent 后端。", "install_command": "git clone https://github.com/fast-ai/agent-ui.git && pnpm install && pnpm dev", "agent_prompt": "请帮我把这个前端界面接入本地大模型 API", "tags": [ "前端UI", "思维链渲染", "组件库" ] }, { "id": "uumit-agent-sdk", "name": "uumit-suite", "repo": "https://oss.uumit.com/skills/v2/index.json", "category": "AI Agent", "stars": "1.9k", "lang": "Node.js", "summary": "任务市场全自动巡航、评估与履约交付套件", "why_recommend": "支持工单机器初筛、自主编写高通过率申请书、自动成果物打包与交卷闭环，副业变现利器。", "install_command": "node /root/.claude/skills/uumit-cruise/scripts/cruise.js market-run", "agent_prompt": "请帮我巡航当前任务大厅候选任务并汇报结果", "tags": [ "自动接单", "巡航履约", "副业变现" ] } ] }