实时增长榜 · 周综述每周日由 spark 生成
JVM 的便携式 AI Agent 运行时。一个 @Agent 类可通过一套 SPI 在 Spring AI、LangChain4j、Anthropic 或另外 9 个后端上运行。支持 token 流式输出、工具调用、人工审批，并通过 WebSocket、SSE、gRPC 或 WebTransport/HTTP3 提供治理能力。兼容 MCP、A2A 与 AG-UI 协议。Portable AI agent runtime for the JVM. One @Agent class runs on Spring AI, LangChain4j, Anthropic, or 9 more behind one SPI. Token streaming, tool calls, human approvals, and governance over WebSocket, SSE, gRPC, or WebTransport/HTTP3. Speaks MCP, A2A, and AG-UI.
本地优先、agent 原生的 ComfyUI 控制平面 —— MCP server + 侧边栏 agent,可生成图像、视频与音频,撰写并运行工作流,并以自然语言编辑实时图,兼容任意 LLM(Claude、ChatGPT、Gemini、本地 Ollama 或任何托管模型)。提供 178 个工具、36 项 AI skill、55 个安装包。支持本地、局域网、VPS 或 Comfy Cloud 部署。Local-first, agent-native control plane for ComfyUI — MCP server + sidebar agent that generates images, video & audio, authors and runs workflows, and edits your live graph in natural language on ANY LLM (Claude, ChatGPT, Gemini, offline Ollama, or any hosted model). 178 tools, 36 AI skills, 55 installer packs. Local, LAN, VPS, or Comfy Cloud.
本地优先、自托管的 AI agent 运行时与 MCP 桥接，提供沙箱会话、内存、凭据、审计/回放以及本地 Console。Local-first, self-hosted AI agent runtime and MCP bridge with sandboxed sessions, memory, credentials, audit/replay, and a local Console.
开源 agent 运行时 —— 基于 SSH 的原生隔离、eBPF 出站策略、Kubernetes + LXC 后端、GPU 直通、原生 MCP 的 CLIOpen-source agent runtime — SSH-native isolation, eBPF egress policy, Kubernetes + LXC backends, GPU passthrough, MCP-native CLI
开源跨平台的 AI agents harness——基于 JVM（非 Electron）的原生多线程操作员控制台，可运行 Claude Code、Codex、Gemini 或 OpenCode，集成真实浏览器、终端、编辑器、密钥及 100+ MCP 工具。面向企业、科研与研究场景。Open-source, multi-platform harness for AI agents — a native, multi-threaded operator's console (JVM, not Electron) to run Claude Code, Codex, Gemini or OpenCode with a real browser, terminal, editor, secrets & 100+ MCP tools. Built for enterprises, science & research.
Rove — 你的终端的 Agent 多路复用器。在隔离的 worktree 与持久化会话中并行运行编码 Agent 任务。Rove — the agent multiplexer for your terminal. Run coding agents on parallel tasks with isolated worktrees and persistent sessions.
XXD Panel 100——朴素民俗风的叙事插画 Soldier Skill。XXD Panel 100 — naïve folk-inspired narrative illustration Soldier Skill
提供 DeepSeek 与 Qwen 服务，运行按需模型库，并在 2 块 NVIDIA DGX Spark 上使用 Unsloth QLoRA 微调。TP=2 的 vLLM 通道汇聚于单一 OpenAI 兼容端点，供 OpenCode、Cursor 与 Hermes 使用。诚实的基准，已提交的工件。Serve DeepSeek and Qwen, run an on-demand model library, and fine-tune with Unsloth QLoRA on 2x NVIDIA DGX Spark. TP=2 vLLM lanes behind one OpenAI-compatible endpoint for OpenCode, Cursor, and Hermes. Honest benchmarks, committed artifacts.
你专属的、持续进化的社交媒体助手，让创作更轻松。Your private evolving social media assistant, making creation easier.
3D 打印骑行装备 2026：轮挡、平板支架与 AirTag 挂架3D Printable Cycling Gear 2026: Wheel Blocks, Tablet Stands & AirTag Holders
FortiusANT 2026:面向 Zwift、Rouvy 与 TrainerRoad 的智能骑行台桥接工具,支持 ANT+ 与 BLEFortiusANT 2026: Smart Trainer Bridge for Zwift, Rouvy & TrainerRoad via ANT+ & BLE
CoRT 2.0：面向 AI Agents 的革命性上下文推理工具包（2026）。CoRT 2.0: Revolutionary Contextual Reasoning Toolkit for AI Agents 2026
Helldivers 2 Liberty Arsenal 2026:无限弹药、超级样本与勋章修改器Helldivers 2 Liberty Arsenal 2026: Unlimited Ammo, Super Samples & Medals Trainer
LGS-Helper 2026：面向玩家的智能本地化工具包与区域内容解锁器LGS-Helper 2026: Smart Localization Toolkit & Regional Content Unlocker for Gamers
Ultimate BlackOut CPVP 插件指南 2026：Meteor Client 进阶功能。Ultimate BlackOut CPVP Plugin Guide 2026: Advanced Meteor Client Features
Enshrouded Trainer 2026 —— 终极上帝模式与物品生成工具Enshrouded Trainer 2026 – Ultimate God Mode & Item Spawn Tool
Ultimate Plant vs Zombies Trainer Overhaul 2026:进阶模组与 AI 作弊指南Ultimate Plant vs Zombies Trainer Overhaul 2026: Advanced Mods & AI Cheats Guide
Megabonk Max Trainer 2026：终极 AI 驱动 DPS 优化器Megabonk Max Trainer 2026: Ultimate AI-Powered DPS Optimizer
Chess PGN Trainer Pro：训练大师 2026。Chess PGN Trainer Pro: Drill Master 2026
Arcane Codex：精通法术与数字巫术 2026Arcane Codex: Master Spellcraft & Digital Sorcery 2026
ESP32 DirCon Pro：通过 WiFi 与以太网连接的智能骑行台桥接器 2026ESP32 DirCon Pro: Smart Trainer Bridge Over WiFi & Ethernet 2026
高级 WARDOGS 辅助：ESP、自瞄与载具战斗增强 2026Advanced WARDOGS Hacks: ESP, Aimbot & Vehicle Combat Enhancements 2026
以证据为先的 agent skill，基于真实研究数据生成可审计的统计分析、诊断、发表级图表、表格与教学报告。Evidence-first agent skill for auditable statistical analysis, diagnostics, publication figures, tables, and teaching reports from real research data.
面向 AI Agent 的循证科学写作与修订 SkillEvidence scientific writing and revision skill for AI agents
一体化学术研究 Skill 套件：7 个仓库合并为一个路由式 Skill — 186 个 Skill、20 篇合并指南、148 篇工具指南，涵盖论文发现、阅读、文献综述、写作、同行评审、 rebuttal 与发表，以及 12/23 阶段自主研究流水线。All-in-one academic research skill suite: 7 repositories consolidated into one router-style skill — 186 skills, 20 merged guides, 148 tool guides, covering paper discovery, reading, literature review, writing, peer review, rebuttal, and publishing, plus 12/23-stage autonomous research pipelines.
提交至 Carnegie Mellon Sports Analytics Conference 2026 可复现研究竞赛的作品。Submission for Carnegie Mellon Sports Analytics Conference Reproducible Research Competition 2026
每日从 Zotero、arXiv 和 IACR ePrint 推荐的密码学论文。Daily cryptography paper recommendations from Zotero, arXiv, and IACR ePrint
开源 Python CGE 建模框架，扩展自 Fung & Burtwistle 的 PyCGE 及 Hosoe–Gasawa–Hashimoto 教材模型，包含 CAMCGE 与 IFPRI 实现。面向政策模拟、教学、复现与可重复研究；"-Core" 遵循 OG-Core 的示例约定。Open-source Python framework for CGE modelling, extending PyCGE by Fung & Burtwistle and the Hosoe–Gasawa–Hashimoto textbook models, with CAMCGE and IFPRI implementations. Built for policy simulation, teaching, replication and reproducible research; “-Core” follows the convention exemplified by OG-Core.
用于实证金融研究的 Python 流水线，集成数据获取、统计分析、尾部风险分析与投资组合构建。A Python pipeline for empirical financial research, integrating data acquisition, statistical analysis, tail risk analytics, and portfolio construction.
使用参数与非参数检验，围绕虚构的健康保险数据集回答两个研究问题。Statistical testing using parametric and non-parametric tests to address two research questions utilizing a fictional health insurance dataset
基于本地 RAG 的 AI 研究助手，用于分析学术论文、总结研究发现、对比研究，并以页级引用回答研究问题。构建于 Python、Streamlit、Chroma、Sentence Transformers、Ollama 和 Llama 3。An AI research assistant using local RAG to analyze academic papers, summarize research findings, compare studies, and answer research questions with page-level citations. Built with Python, Streamlit, Chroma, Sentence Transformers, Ollama, and Llama 3.
一个能回答我研究问题的 Dashboard。A Dashboard that answers my research questions
萨里大学 2026 年 9 月理学硕士论文《PRISM：Agentic AI 闭环科学写作》的日志、prompt 与支撑材料追踪。OpenAI PRISM 闭环工作流，加入对抗性红队/蓝队同行评审层，由非领域操作员应用于三个代谢组学数据集。Trace logs, prompts and supporting materials for the MSc dissertation "PRISM: Agentic AI Closed-Loop Scientific Writing", University of Surrey, September 2026. OpenAI PRISM closed-loop workflow with adversarial red team / blue team peer review layer, applied to three metabolomics datasets by a non-domain operator.
Oxford Nanopore 污水宏基因组测序数据的物种分类流程。Workflow for taxonomic classification of Oxford Nanopore wastewater metagenomic sequencing data.
一项关于数字界面设计（极简、支架式、游戏化）如何影响学生参与度、准确性与任务时长的独立研究的数据清洗与统计分析（ANOVA、Tukey、回归）。Data cleaning and statistical analysis (ANOVA, Tukey, regression) for an independent research study on how digital interface design (minimal, scaffolded, gamified) affects student engagement, accuracy, and time-on-task.
Crescent City（加州）的可复现学术史——从 Tolowa Dee-ni' 聚居到 1964 海啸至今共 46 章，附带 24 张确定性生成的 matplotlib 图表、BibTeX 校验，以及通过 Pandoc + XeLaTeX + pandoc-crossref 构建 PDF（Python、uv、pytest）。Reproducible scholarly history of Crescent City, California — 46 chapters from Tolowa Dee-ni' settlement through the 1964 tsunami to today, with 24 deterministically generated matplotlib figures, BibTeX validation, and PDF build via Pandoc + XeLaTeX + pandoc-crossref (Python, uv, pytest).
面向研究 Agent 的结构感知科学英语写作 Skill（POLISH/REWRITE/GENERATE）。MIT，欢迎贡献。Structure-aware scientific English writing skill for research agents (POLISH/REWRITE/GENERATE). MIT, contributions welcome.
基于人类基准的 LLM 生成 Infrastructure-as-Code 安全基准测试。100 场景 × 12 模型配置 = 1,196 个工件，由 Checkov/Trivy/KICS 扫描，并与 634 个人工编写模板对比。所有模型的漏洞密度均为人工的 3.2–3.9 倍。arXiv:2608.28021Human-anchored security benchmark for LLM-generated Infrastructure-as-Code. 100 scenarios × 12 model configs = 1,196 artifacts scanned by Checkov/Trivy/KICS, compared against 634 human-written templates. Every model: 3.2–3.9× human vulnerability density. arXiv:2608.28021
| 名次 | 名称 | 方向 | 周增 | Stars |
|---|---|---|---|---|
| 1 — | odysseus-dev/odysseus | Agent 智能体 | +152,621 | 84,138 |
| 2 — | anywhere-labs/deepseek-harness-desktop | 评测基准 | +8,281 | 13,495 |
| 3 — | stablyai/orca | Agent 智能体 | +5,299 | 47,228 |
| 4 — | s1dashu/ip-as-logo-skill | Agent 智能体 | +4,646 | 3,727 |
| 5 — | MiniMax-AI/MiniMax-H3 | LLM 基础设施 | +4,445 | 4,611 |
| 6 — | diegosouzapw/OmniRoute | Agent 智能体 | +4,343 | 53,672 |
| 7 — | holaboss-ai/holaOS | Agent 智能体 | +3,798 | 9,436 |
| 8 — | safishamsi/graphify | 数据与向量库 | +3,752 | 76,856 |
| 9 — | NousResearch/hermes-agent | Agent 智能体 | +3,386 | 230,735 |
| 10 — | zhu1090093659/dsh-web-ui | 评测基准 | +3,318 | 4,541 |
| 11 — | firecrawl/firecrawl | RAG 检索增强 | +2,975 | 165,271 |
| 12 — | firecrawl/anydoc | 工程化 | +2,870 | 13,720 |
| 13 — | koala73/worldmonitor | Agent 智能体 | +2,239 | 82,478 |
| 14 ↑1 | hoainho/img2threejs | 多模态 | +2,051 | 4,176 |
| 15 ↓1 | bojieli/ai-agent-book | Agent 智能体 | +1,978 | 43,727 |
| 16 — | DietrichGebert/ponytail | Agent 智能体 | +1,869 | 100,293 |
| 17 — | anthropics/skills | Agent 智能体 | +1,722 | 167,649 |
| 18 — | infiniflow/ragflow | RAG 检索增强 | +1,661 | 88,637 |
| 19 — | addyosmani/agent-skills | Agent 智能体 | +1,631 | 85,866 |
| 20 — | obra/superpowers | Agent 智能体 | +1,484 | 270,347 |