这一轮 GitHub Trending 里,最密集出现的并不是单一模型项目,而是一整组围绕 agent、skills、memory、workflow、browser 与 orchestration 展开的基础能力模块。更值得关注的是,日榜、周榜和月榜同时出现了这条线索:技术社区对 AI 开发的关注点,正在从“模型能做什么”,转向“系统如何长期运行、协同与交付”。
真正的变化不在于又多了几个 AI 工具,而在于一套更完整的 agent 工程栈开始成形:前端有界面与交互代理,中间有多代理编排和技能层,底层则是记忆、上下文、推理框架与推理基础设施。这种结构性上行,比单个热点项目更值得关注。
•来源链接:https://github.com/trending[1]
•时间范围:Daily(今日)
•页面实际可见条目数:14
1. abhigyanpatwari / GitNexus
•链接:https://github.com/abhigyanpatwari/GitNexus[2]
•Star:23,005
•新增 stars:837 stars today
•中文摘要:一个完全运行在浏览器端的零服务器代码情报引擎,可将 GitHub 仓库或 ZIP 包转成知识图谱,并内置 Graph RAG Agent,适合代码探索与结构化理解。
2. google-ai-edge / gallery
•链接:https://github.com/google-ai-edge/gallery[3]
•Star:17,563
•新增 stars:1,109 stars today
•中文摘要:展示端侧机器学习与生成式 AI 用例的示例集,强调模型本地运行与 on-device ML/GenAI 体验。
3. block / goose
•链接:https://github.com/block/goose[4]
•Star:37,706
•新增 stars:1,514 stars today
•中文摘要:一个可扩展的开源 AI agent,不只做代码建议,还能安装、执行、编辑和测试,强调与任意大模型协同完成完整开发动作。
4. google-ai-edge / LiteRT-LM
•链接:https://github.com/google-ai-edge/LiteRT-LM[5]
•Star:1,827
•新增 stars:487 stars today
•中文摘要:面向本地推理与边缘设备的大模型运行组件,代表轻量化推理基础设施继续受到关注。
5. immich-app / immich
•链接:https://github.com/immich-app/immich[6]
•Star:96,615
•新增 stars:155 stars today
•中文摘要:高性能自托管照片与视频管理方案,虽然不属于典型 AI agent 项目,但显示自托管软件仍有稳定吸引力。
6. KeygraphHQ / shannon
•链接:https://github.com/KeygraphHQ/shannon[7]
•Star:36,115
•新增 stars:703 stars today
•中文摘要:白盒式 AI 自动渗透测试工具,可分析源码、定位攻击面并执行真实利用,说明 AI 安全自动化工具也在升温。
7. NousResearch / hermes-agent
•链接:https://github.com/NousResearch/hermes-agent[8]
•Star:27,317
•新增 stars:1,251 stars today
•中文摘要:强调“可随你一同成长”的 agent 框架,延续了社区对长期运行型智能体的关注。
8. tobi / qmd
•链接:https://github.com/tobi/qmd[9]
•Star:18,409
•新增 stars:298 stars today
•中文摘要:本地化文档与知识库 CLI 搜索引擎,追踪当前 SOTA 方法,体现本地检索与知识访问仍是高频需求。
9. TelegramMessenger / Telegram-iOS
•链接:https://github.com/TelegramMessenger/Telegram-iOS[10]
•Star:8,247
•新增 stars:14 stars today
•中文摘要:Telegram 的 iOS 客户端源码,属于非 AI 项目,但在榜单中起到基础应用类项目的对照作用。
10. kepano / obsidian-skills
•链接:https://github.com/kepano/obsidian-skills[11]
•Star:20,182
•新增 stars:281 stars today
•中文摘要:为 Obsidian 提供 agent skills,覆盖 Markdown、Bases、JSON Canvas 与 CLI 使用,表明“技能化扩展”正在从代码助手走向知识工具。
11. ollama / ollama
•链接:https://github.com/ollama/ollama[12]
•Star:167,472
•新增 stars:164 stars today
•中文摘要:本地运行 Kimi、GLM、DeepSeek、Qwen、Gemma 等模型的主流基础设施,依然是本地 AI 生态的核心入口之一。
12. ggml-org / llama.cpp
•链接:https://github.com/ggml-org/llama.cpp[13]
•Star:101,807
•新增 stars:244 stars today
•中文摘要:经典的 C/C++ 大模型推理框架,持续稳居热门,说明推理层基础设施仍然具备长期需求。
13. siddharthvaddem / openscreen
•链接:https://github.com/siddharthvaddem/openscreen[14]
•Star:23,456
•新增 stars:1,823 stars today
•中文摘要:开源演示录屏与产品展示工具,强调无订阅、无水印、可商用,反映开发者对“展示与传播工具链”同样有强需求。
14. NVIDIA / personaplex
•链接:https://github.com/NVIDIA/personaplex[15]
•Star:7,049
•新增 stars:295 stars today
•中文摘要:NVIDIA 的 PersonaPlex 代码仓库,代表角色化、多人格或多角色 AI 系统的探索仍在继续。
•来源链接:https://github.com/trending?since=weekly[16]
•时间范围:Weekly(本周)
•页面实际可见条目数:14
1. Yeachan-Heo / oh-my-codex
•链接:https://github.com/Yeachan-Heo/oh-my-codex[17]
•Star:17,178
•新增 stars:13,476 stars this week
•中文摘要:为 Codex 增加 hooks、agent teams、HUD 等能力的增强层,直接把“单 agent”升级为“团队协同式 agent 工作台”。
2. siddharthvaddem / openscreen
•链接:https://github.com/siddharthvaddem/openscreen[18]
•Star:23,456
•新增 stars:12,682 stars this week
•中文摘要:免费、开源、可商用的演示录屏方案,本周热度说明开发工具之外,开发者传播工具也在获得更高关注。
3. luongnv89 / claude-howto
•链接:https://github.com/luongnv89/claude-howto[19]
•Star:21,006
•新增 stars:15,128 stars this week
•中文摘要:面向 Claude Code 的可视化示例指南,从基础概念到高级 agents,体现“如何更好使用 agent”的教育型内容需求大幅上升。
4. microsoft / VibeVoice
•链接:https://github.com/microsoft/VibeVoice[20]
•Star:36,687
•新增 stars:10,543 stars this week
•中文摘要:微软开源的前沿语音 AI 项目,显示多模态尤其是语音方向依然是活跃主线之一。
5. google-research / timesfm
•链接:https://github.com/google-research/timesfm[21]
•Star:15,087
•新增 stars:4,702 stars this week
•中文摘要:Google Research 的时间序列基础模型,用于预测任务,反映基础模型范式正继续向垂直任务扩展。
6. Yeachan-Heo / oh-my-claudecode
•链接:https://github.com/Yeachan-Heo/oh-my-claudecode[22]
•Star:24,852
•新增 stars:9,112 stars this week
•中文摘要:面向 Claude Code 的多代理编排框架,强调 teams-first,说明 agent orchestration 正在成为显性需求。
7. NousResearch / hermes-agent
•链接:https://github.com/NousResearch/hermes-agent[23]
•Star:27,317
•新增 stars:9,940 stars this week
•中文摘要:强调长期成长、演进型的 agent 架构,本周持续强势,说明持续运行能力依然是热门方向。
8. onyx-dot-app / onyx
•链接:https://github.com/onyx-dot-app/onyx[24]
•Star:25,349
•新增 stars:4,741 stars this week
•中文摘要:面向多模型的开源 AI 平台,提供更完整的 AI Chat 与集成能力,代表平台层需求依旧存在。
9. google-ai-edge / gallery
•链接:https://github.com/google-ai-edge/gallery[25]
•Star:17,563
•新增 stars:1,286 stars this week
•中文摘要:端侧 GenAI 用例展示项目继续上榜,说明本地 AI 与设备侧体验仍是技术社区稳定关注点。
10. sherlock-project / sherlock
•链接:https://github.com/sherlock-project/sherlock[26]
•Star:79,991
•新增 stars:5,633 stars this week
•中文摘要:跨社交网络按用户名检索账号的经典项目,本周重回热榜,显示 OSINT 与信息检索工具仍有稳定吸引力。
11. asgeirtj / system_prompts_leaks
•链接:https://github.com/asgeirtj/system_prompts_leaks[27]
•Star:37,555
•新增 stars:2,235 stars this week
•中文摘要:汇总主流模型与 AI 工具系统提示词的项目,侧面说明开发者对模型行为机制和 agent 约束层的兴趣持续存在。
12. vas3k / TaxHacker
•链接:https://github.com/vas3k/TaxHacker[28]
•Star:4,768
•新增 stars:2,204 stars this week
•中文摘要:自托管 AI 财务记账与票据分析应用,说明垂直场景 AI 应用仍在不断冒头。
13. nvim-treesitter / nvim-treesitter
•链接:https://github.com/nvim-treesitter/nvim-treesitter[29]
•Star:13,541
•新增 stars:102 stars this week
•中文摘要:Neovim 的 Treesitter 配置与抽象层,是榜单中较少的经典基础开发工具项目,形成对 AI 热点的背景参照。
14. 其余页面可见条目
•浏览器快照显示本周页面至少存在上述 13 个完整可读项目与一个尾部条目区域;由于页面快照截断影响,完整正文中按实际成功读取到的可见条目整理,不补造未稳定读取内容。
•来源链接:https://github.com/trending?since=monthly[30]
•时间范围:Monthly(本月)
•页面实际可见条目数:20
1. 666ghj / MiroFish
•链接:https://github.com/666ghj/MiroFish[31]
•Star:50,472
•新增 stars:45,506 stars this month
•中文摘要:一个通用群体智能引擎,试图把“预测万物”做成统一框架,属于极具想象空间的智能系统方向。
2. msitarzewski / agency-agents
•链接:https://github.com/msitarzewski/agency-agents[32]
•Star:72,709
•新增 stars:65,886 stars this month
•中文摘要:把完整 AI agency 抽象成一组有角色、有流程、有交付的专家 agent,是典型的“多角色协作系统”项目。
3. NousResearch / hermes-agent
•链接:https://github.com/NousResearch/hermes-agent[33]
•Star:27,317
•新增 stars:24,111 stars this month
•中文摘要:强调 agent 可持续成长,持续位居月榜前列,说明“长期运行型 agent”已经不是短期噱头。
4. bytedance / deer-flow
•链接:https://github.com/bytedance/deer-flow[34]
•Star:58,498
•新增 stars:33,890 stars this month
•中文摘要:一个开源长时任务 SuperAgent harness,整合 sandbox、memory、tools、skills、subagents 与 message gateway,体现 agent 工程体系化正在加速。
5. mattpocock / skills
•链接:https://github.com/mattpocock/skills[35]
•Star:12,414
•新增 stars:11,191 stars this month
•中文摘要:作者个人的 skills 目录公开化,说明 skill 生态本身已经变成开发者关注对象。
6. FujiwaraChoki / MoneyPrinterV2
•链接:https://github.com/FujiwaraChoki/MoneyPrinterV2[36]
•Star:28,542
•新增 stars:14,089 stars this month
•中文摘要:围绕在线赚钱流程自动化的项目,代表 AI 自动化正在向更直接的商业场景渗透。
7. alibaba / page-agent
•链接:https://github.com/alibaba/page-agent[37]
•Star:15,563
•新增 stars:14,717 stars this month
•中文摘要:基于页面内 JavaScript 的 GUI agent,可用自然语言操控网页界面,是浏览器原生代理方向的重要信号。
8. lightpanda-io / browser
•链接:https://github.com/lightpanda-io/browser[38]
•Star:27,325
•新增 stars:15,361 stars this month
•中文摘要:专为 AI 与自动化设计的无头浏览器,说明 browser automation 正在被重新定义为 AI 基础设施。
9. pbakaus / impeccable
•链接:https://github.com/pbakaus/impeccable[39]
•Star:16,178
•新增 stars:15,660 stars this month
•中文摘要:一个试图提升 AI harness 设计能力的设计语言项目,把设计规范直接纳入 agent 工作流。
10. affaan-m / everything-claude-code
•链接:https://github.com/affaan-m/everything-claude-code[40]
•Star:141,885
•新增 stars:79,260 stars this month
•中文摘要:围绕 Claude Code、Codex、Cursor 等系统整理 skills、memory、security 与 research-first 开发方法,显示“agent 使用论”本身也在快速扩散。
11. jarrodwatts / claude-hud
•链接:https://github.com/jarrodwatts/claude-hud[41]
•Star:17,162
•新增 stars:12,943 stars this month
•中文摘要:为 Claude Code 提供上下文使用、活动工具、运行中 agent 和 todo 进度的可视化 HUD,是 agent 可观测性方向的代表。
12. mvanhorn / last30days-skill
•链接:https://github.com/mvanhorn/last30days-skill[42]
•Star:18,996
•新增 stars:15,180 stars this month
•中文摘要:面向 AI agent 的研究型 skill,可跨 Reddit、X、YouTube、HN、Polymarket 与 Web 聚合近 30 天信息并生成带引用的研究摘要。
13. shareAI-lab / learn-claude-code
•链接:https://github.com/shareAI-lab/learn-claude-code[43]
•Star:48,927
•新增 stars:27,309 stars this month
•中文摘要:从 0 到 1 用 Bash 构建 nano claude code-like agent harness,说明“如何自己造 agent”仍是高热主题。
14. obra / superpowers
•链接:https://github.com/obra/superpowers[44]
•Star:136,989
•新增 stars:65,451 stars this month
•中文摘要:一个 agentic skills 框架与软件开发方法论项目,反映技能体系与开发流程方法正在一同产品化。
15. TauricResearch / TradingAgents
•链接:https://github.com/TauricResearch/TradingAgents[45]
•Star:47,729
•新增 stars:16,280 stars this month
•中文摘要:多智能体金融交易框架,说明 agent 应用已从编码扩展到复杂决策型场景。
16. langchain-ai / open-swe
•链接:https://github.com/langchain-ai/open-swe[46]
•Star:9,227
•新增 stars:3,966 stars this month
•中文摘要:异步开源编码 agent,面向软件工程自动化,是 SWE agent 路线的代表之一。
17. siddharthvaddem / openscreen
•链接:https://github.com/siddharthvaddem/openscreen[47]
•Star:23,456
•新增 stars:14,130 stars this month
•中文摘要:开源录屏演示工具持续上榜,体现开发者对“生成之外的表达与展示链路”也在加强投入。
18. volcengine / OpenViking
•链接:https://github.com/volcengine/OpenViking[48]
•Star:21,264
•新增 stars:16,523 stars this month
•中文摘要:为 AI agent 设计的开源上下文数据库,统一管理 memory、resources 与 skills,并支持分层上下文投递与自进化。
19. alirezarezvani / claude-skills
•链接:https://github.com/alirezarezvani/claude-skills[49]
•Star:9,693
•新增 stars:7,083 stars this month
•中文摘要:收录 220+ Claude Code skills 与多种 agent 插件,显示 skills 正从少数高级用户实践扩散为独立生态。
20. promptfoo / promptfoo
•链接:https://github.com/promptfoo/promptfoo[50]
•Star:19,587
•新增 stars:8,723 stars this month
•中文摘要:用于提示词、agent 与 RAG 的测试、红队与漏洞扫描框架,表明评测与安全治理正在成为 AI 工程的标准组成部分。
21. langchain-ai / deepagents
•链接:https://github.com/langchain-ai/deepagents[51]
•Star:19,470
•新增 stars:9,451 stars this month
•中文摘要:基于 LangChain 与 LangGraph 的 agent harness,配有 planning、filesystem 与 subagents,更适合复杂任务处理。
22. microsoft / BitNet
•链接:https://github.com/microsoft/BitNet[52]
•Star:37,233
•新增 stars:8,697 stars this month
•中文摘要:1-bit LLM 官方推理框架,说明高效推理依然是月度尺度上的稳定主线。
如果只看日榜,会以为社区仍在追逐新鲜的 agent 项目与应用形态;如果结合周榜和月榜,信号就更清楚了:真正集体上行的不是某一个爆款仓库,而是围绕 agent 的工程栈本身。
从 skills、HUD、browser、page agent,到 memory、context database、open-swe、deepagents,再到 deer-flow 这类长时任务框架,技术社区正在把 AI 开发重新组织成一套更可运行、可协同、可观察、可治理的系统。这背后反映的并不只是热度变化,而是开发范式的迁移:AI 项目竞争的重点,正在从“接上模型”转向“能否把模型嵌入稳定的软件结构”。
如果这种趋势继续延续,那么接下来更值得关注的就不是单点模型能力,而是谁能把 skills、memory、workflow、browser automation 与 evaluation 这些模块组织成真正可复用的基础设施。GitHub Trending 这一次给出的,不只是热榜答案,更像是一份关于下一阶段 AI 工程主线的侧写。
•Daily:https://github.com/trending[53]
•Weekly:https://github.com/trending?since=weekly[54]
•Monthly:https://github.com/trending?since=monthly[55]
参考链接
[1] https://github.com/trending: https://github.com/trending
[2] https://github.com/abhigyanpatwari/GitNexus: https://github.com/abhigyanpatwari/GitNexus
[3] https://github.com/google-ai-edge/gallery: https://github.com/google-ai-edge/gallery
[4] https://github.com/block/goose: https://github.com/block/goose
[5] https://github.com/google-ai-edge/LiteRT-LM: https://github.com/google-ai-edge/LiteRT-LM
[6] https://github.com/immich-app/immich: https://github.com/immich-app/immich
[7] https://github.com/KeygraphHQ/shannon: https://github.com/KeygraphHQ/shannon
[8] https://github.com/NousResearch/hermes-agent: https://github.com/NousResearch/hermes-agent
[9] https://github.com/tobi/qmd: https://github.com/tobi/qmd
[10] https://github.com/TelegramMessenger/Telegram-iOS: https://github.com/TelegramMessenger/Telegram-iOS
[11] https://github.com/kepano/obsidian-skills: https://github.com/kepano/obsidian-skills
[12] https://github.com/ollama/ollama: https://github.com/ollama/ollama
[13] https://github.com/ggml-org/llama.cpp: https://github.com/ggml-org/llama.cpp
[14] https://github.com/siddharthvaddem/openscreen: https://github.com/siddharthvaddem/openscreen
[15] https://github.com/NVIDIA/personaplex: https://github.com/NVIDIA/personaplex
[16] https://github.com/trending?since=weekly: https://github.com/trending?since=weekly
[17] https://github.com/Yeachan-Heo/oh-my-codex: https://github.com/Yeachan-Heo/oh-my-codex
[18] https://github.com/siddharthvaddem/openscreen: https://github.com/siddharthvaddem/openscreen
[19] https://github.com/luongnv89/claude-howto: https://github.com/luongnv89/claude-howto
[20] https://github.com/microsoft/VibeVoice: https://github.com/microsoft/VibeVoice
[21] https://github.com/google-research/timesfm: https://github.com/google-research/timesfm
[22] https://github.com/Yeachan-Heo/oh-my-claudecode: https://github.com/Yeachan-Heo/oh-my-claudecode
[23] https://github.com/NousResearch/hermes-agent: https://github.com/NousResearch/hermes-agent
[24] https://github.com/onyx-dot-app/onyx: https://github.com/onyx-dot-app/onyx
[25] https://github.com/google-ai-edge/gallery: https://github.com/google-ai-edge/gallery
[26] https://github.com/sherlock-project/sherlock: https://github.com/sherlock-project/sherlock
[27] https://github.com/asgeirtj/system_prompts_leaks: https://github.com/asgeirtj/system_prompts_leaks
[28] https://github.com/vas3k/TaxHacker: https://github.com/vas3k/TaxHacker
[29] https://github.com/nvim-treesitter/nvim-treesitter: https://github.com/nvim-treesitter/nvim-treesitter
[30] https://github.com/trending?since=monthly: https://github.com/trending?since=monthly
[31] https://github.com/666ghj/MiroFish: https://github.com/666ghj/MiroFish
[32] https://github.com/msitarzewski/agency-agents: https://github.com/msitarzewski/agency-agents
[33] https://github.com/NousResearch/hermes-agent: https://github.com/NousResearch/hermes-agent
[34] https://github.com/bytedance/deer-flow: https://github.com/bytedance/deer-flow
[35] https://github.com/mattpocock/skills: https://github.com/mattpocock/skills
[36] https://github.com/FujiwaraChoki/MoneyPrinterV2: https://github.com/FujiwaraChoki/MoneyPrinterV2
[37] https://github.com/alibaba/page-agent: https://github.com/alibaba/page-agent
[38] https://github.com/lightpanda-io/browser: https://github.com/lightpanda-io/browser
[39] https://github.com/pbakaus/impeccable: https://github.com/pbakaus/impeccable
[40] https://github.com/affaan-m/everything-claude-code: https://github.com/affaan-m/everything-claude-code
[41] https://github.com/jarrodwatts/claude-hud: https://github.com/jarrodwatts/claude-hud
[42] https://github.com/mvanhorn/last30days-skill: https://github.com/mvanhorn/last30days-skill
[43] https://github.com/shareAI-lab/learn-claude-code: https://github.com/shareAI-lab/learn-claude-code
[44] https://github.com/obra/superpowers: https://github.com/obra/superpowers
[45] https://github.com/TauricResearch/TradingAgents: https://github.com/TauricResearch/TradingAgents
[46] https://github.com/langchain-ai/open-swe: https://github.com/langchain-ai/open-swe
[47] https://github.com/siddharthvaddem/openscreen: https://github.com/siddharthvaddem/openscreen
[48] https://github.com/volcengine/OpenViking: https://github.com/volcengine/OpenViking
[49] https://github.com/alirezarezvani/claude-skills: https://github.com/alirezarezvani/claude-skills
[50] https://github.com/promptfoo/promptfoo: https://github.com/promptfoo/promptfoo
[51] https://github.com/langchain-ai/deepagents: https://github.com/langchain-ai/deepagents
[52] https://github.com/microsoft/BitNet: https://github.com/microsoft/BitNet
[53] https://github.com/trending: https://github.com/trending
[54] https://github.com/trending?since=weekly: https://github.com/trending?since=weekly
[55] https://github.com/trending?since=monthly: https://github.com/trending?since=monthly
版权声明:本文内容由互联网用户自发贡献,该文观点仅代表作者本人。本站仅提供信息存储空间服务,不拥有所有权,不承担相关法律责任。如发现本站有涉嫌侵权/违法违规的内容,请联系我们,一经查实,本站将立刻删除。
如需转载请保留出处:https://51itzy.com/kjqy/254892.html