[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-Galaxy-Dawn--claude-scholar":3,"tool-Galaxy-Dawn--claude-scholar":64},[4,17,27,35,48,56],{"id":5,"name":6,"github_repo":7,"description_zh":8,"stars":9,"difficulty_score":10,"last_commit_at":11,"category_tags":12,"status":16},3808,"stable-diffusion-webui","AUTOMATIC1111\u002Fstable-diffusion-webui","stable-diffusion-webui 是一个基于 Gradio 构建的网页版操作界面，旨在让用户能够轻松地在本地运行和使用强大的 Stable Diffusion 图像生成模型。它解决了原始模型依赖命令行、操作门槛高且功能分散的痛点，将复杂的 AI 绘图流程整合进一个直观易用的图形化平台。\n\n无论是希望快速上手的普通创作者、需要精细控制画面细节的设计师，还是想要深入探索模型潜力的开发者与研究人员，都能从中获益。其核心亮点在于极高的功能丰富度：不仅支持文生图、图生图、局部重绘（Inpainting）和外绘（Outpainting）等基础模式，还独创了注意力机制调整、提示词矩阵、负向提示词以及“高清修复”等高级功能。此外，它内置了 GFPGAN 和 CodeFormer 等人脸修复工具，支持多种神经网络放大算法，并允许用户通过插件系统无限扩展能力。即使是显存有限的设备，stable-diffusion-webui 也提供了相应的优化选项，让高质量的 AI 艺术创作变得触手可及。",162132,3,"2026-04-05T11:01:52",[13,14,15],"开发框架","图像","Agent","ready",{"id":18,"name":19,"github_repo":20,"description_zh":21,"stars":22,"difficulty_score":23,"last_commit_at":24,"category_tags":25,"status":16},1381,"everything-claude-code","affaan-m\u002Feverything-claude-code","everything-claude-code 是一套专为 AI 编程助手（如 Claude Code、Codex、Cursor 等）打造的高性能优化系统。它不仅仅是一组配置文件，而是一个经过长期实战打磨的完整框架，旨在解决 AI 代理在实际开发中面临的效率低下、记忆丢失、安全隐患及缺乏持续学习能力等核心痛点。\n\n通过引入技能模块化、直觉增强、记忆持久化机制以及内置的安全扫描功能，everything-claude-code 能显著提升 AI 在复杂任务中的表现，帮助开发者构建更稳定、更智能的生产级 AI 代理。其独特的“研究优先”开发理念和针对 Token 消耗的优化策略，使得模型响应更快、成本更低，同时有效防御潜在的攻击向量。\n\n这套工具特别适合软件开发者、AI 研究人员以及希望深度定制 AI 工作流的技术团队使用。无论您是在构建大型代码库，还是需要 AI 协助进行安全审计与自动化测试，everything-claude-code 都能提供强大的底层支持。作为一个曾荣获 Anthropic 黑客大奖的开源项目，它融合了多语言支持与丰富的实战钩子（hooks），让 AI 真正成长为懂上",138956,2,"2026-04-05T11:33:21",[13,15,26],"语言模型",{"id":28,"name":29,"github_repo":30,"description_zh":31,"stars":32,"difficulty_score":23,"last_commit_at":33,"category_tags":34,"status":16},2271,"ComfyUI","Comfy-Org\u002FComfyUI","ComfyUI 是一款功能强大且高度模块化的视觉 AI 引擎，专为设计和执行复杂的 Stable Diffusion 图像生成流程而打造。它摒弃了传统的代码编写模式，采用直观的节点式流程图界面，让用户通过连接不同的功能模块即可构建个性化的生成管线。\n\n这一设计巧妙解决了高级 AI 绘图工作流配置复杂、灵活性不足的痛点。用户无需具备编程背景，也能自由组合模型、调整参数并实时预览效果，轻松实现从基础文生图到多步骤高清修复等各类复杂任务。ComfyUI 拥有极佳的兼容性，不仅支持 Windows、macOS 和 Linux 全平台，还广泛适配 NVIDIA、AMD、Intel 及苹果 Silicon 等多种硬件架构，并率先支持 SDXL、Flux、SD3 等前沿模型。\n\n无论是希望深入探索算法潜力的研究人员和开发者，还是追求极致创作自由度的设计师与资深 AI 绘画爱好者，ComfyUI 都能提供强大的支持。其独特的模块化架构允许社区不断扩展新功能，使其成为当前最灵活、生态最丰富的开源扩散模型工具之一，帮助用户将创意高效转化为现实。",107662,"2026-04-03T11:11:01",[13,14,15],{"id":36,"name":37,"github_repo":38,"description_zh":39,"stars":40,"difficulty_score":23,"last_commit_at":41,"category_tags":42,"status":16},2268,"ML-For-Beginners","microsoft\u002FML-For-Beginners","ML-For-Beginners 是由微软推出的一套系统化机器学习入门课程，旨在帮助零基础用户轻松掌握经典机器学习知识。这套课程将学习路径规划为 12 周，包含 26 节精炼课程和 52 道配套测验，内容涵盖从基础概念到实际应用的完整流程，有效解决了初学者面对庞大知识体系时无从下手、缺乏结构化指导的痛点。\n\n无论是希望转型的开发者、需要补充算法背景的研究人员，还是对人工智能充满好奇的普通爱好者，都能从中受益。课程不仅提供了清晰的理论讲解，还强调动手实践，让用户在循序渐进中建立扎实的技能基础。其独特的亮点在于强大的多语言支持，通过自动化机制提供了包括简体中文在内的 50 多种语言版本，极大地降低了全球不同背景用户的学习门槛。此外，项目采用开源协作模式，社区活跃且内容持续更新，确保学习者能获取前沿且准确的技术资讯。如果你正寻找一条清晰、友好且专业的机器学习入门之路，ML-For-Beginners 将是理想的起点。",84991,"2026-04-05T10:45:23",[14,43,44,45,15,46,26,13,47],"数据工具","视频","插件","其他","音频",{"id":49,"name":50,"github_repo":51,"description_zh":52,"stars":53,"difficulty_score":10,"last_commit_at":54,"category_tags":55,"status":16},3128,"ragflow","infiniflow\u002Fragflow","RAGFlow 是一款领先的开源检索增强生成（RAG）引擎，旨在为大语言模型构建更精准、可靠的上下文层。它巧妙地将前沿的 RAG 技术与智能体（Agent）能力相结合，不仅支持从各类文档中高效提取知识，还能让模型基于这些知识进行逻辑推理和任务执行。\n\n在大模型应用中，幻觉问题和知识滞后是常见痛点。RAGFlow 通过深度解析复杂文档结构（如表格、图表及混合排版），显著提升了信息检索的准确度，从而有效减少模型“胡编乱造”的现象，确保回答既有据可依又具备时效性。其内置的智能体机制更进一步，使系统不仅能回答问题，还能自主规划步骤解决复杂问题。\n\n这款工具特别适合开发者、企业技术团队以及 AI 研究人员使用。无论是希望快速搭建私有知识库问答系统，还是致力于探索大模型在垂直领域落地的创新者，都能从中受益。RAGFlow 提供了可视化的工作流编排界面和灵活的 API 接口，既降低了非算法背景用户的上手门槛，也满足了专业开发者对系统深度定制的需求。作为基于 Apache 2.0 协议开源的项目，它正成为连接通用大模型与行业专有知识之间的重要桥梁。",77062,"2026-04-04T04:44:48",[15,14,13,26,46],{"id":57,"name":58,"github_repo":59,"description_zh":60,"stars":61,"difficulty_score":10,"last_commit_at":62,"category_tags":63,"status":16},519,"PaddleOCR","PaddlePaddle\u002FPaddleOCR","PaddleOCR 是一款基于百度飞桨框架开发的高性能开源光学字符识别工具包。它的核心能力是将图片、PDF 等文档中的文字提取出来，转换成计算机可读取的结构化数据，让机器真正“看懂”图文内容。\n\n面对海量纸质或电子文档，PaddleOCR 解决了人工录入效率低、数字化成本高的问题。尤其在人工智能领域，它扮演着连接图像与大型语言模型（LLM）的桥梁角色，能将视觉信息直接转化为文本输入，助力智能问答、文档分析等应用场景落地。\n\nPaddleOCR 适合开发者、算法研究人员以及有文档自动化需求的普通用户。其技术优势十分明显：不仅支持全球 100 多种语言的识别，还能在 Windows、Linux、macOS 等多个系统上运行，并灵活适配 CPU、GPU、NPU 等各类硬件。作为一个轻量级且社区活跃的开源项目，PaddleOCR 既能满足快速集成的需求，也能支撑前沿的视觉语言研究，是处理文字识别任务的理想选择。",74913,"2026-04-05T10:44:17",[26,14,13,46],{"id":65,"github_repo":66,"name":67,"description_en":68,"description_zh":69,"ai_summary_zh":69,"readme_en":70,"readme_zh":71,"quickstart_zh":72,"use_case_zh":73,"hero_image_url":74,"owner_login":75,"owner_name":76,"owner_avatar_url":77,"owner_bio":78,"owner_company":79,"owner_location":80,"owner_email":81,"owner_twitter":80,"owner_website":80,"owner_url":82,"languages":83,"stars":100,"forks":101,"last_commit_at":102,"license":80,"difficulty_score":23,"env_os":103,"env_gpu":104,"env_ram":104,"env_deps":105,"category_tags":114,"github_topics":115,"view_count":23,"oss_zip_url":80,"oss_zip_packed_at":80,"status":16,"created_at":127,"updated_at":128,"faqs":129,"releases":130},3284,"Galaxy-Dawn\u002Fclaude-scholar","claude-scholar","Semi-automated research assistant for academic research and software development. Supports Claude Code, OpenCode, and Codex CLI across ideation, coding, experiments, writing, and publication.","claude-scholar 是一款专为学术研究和软件开发设计的半自动化科研助手，尤其适合计算机科学与人工智能领域的研究者。它旨在解决科研全流程中碎片化任务繁琐、上下文管理困难以及文献与代码割裂的痛点，能够辅助用户完成从创意构思、文献综述、代码编写、实验运行到论文撰写及项目知识管理的完整闭环。\n\n该工具的核心亮点在于其广泛的兼容性与灵活的工作流支持。它不仅完美适配 Claude Code，还通过独立的分支分别支持 OpenAI 的 Codex CLI 和 OpenCode，让用户可以根据偏好自由选择底层驱动。在技术细节上，claude-scholar 集成了智能的 Zotero 文献导入与去重工作流，实现了参考文献的高效管理；同时提供安装脚本，能安全地备份并合并本地配置，避免覆盖用户现有的个性化设置。此外，它还支持与 Obsidian 等知识管理工具联动，帮助研究者构建结构化的项目知识库。无论是需要快速验证算法原型的开发者，还是致力于发表高质量论文的科研人员，claude-scholar 都能通过标准化的技能库与代理机制，显著提升科研效率，让使用者更专注于核心创新。","\u003Cdiv align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FGalaxy-Dawn_claude-scholar_readme_dc32474688cf.png\" alt=\"Claude Scholar Logo\" width=\"100%\"\u002F>\n\n  \u003Cp>\n    \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FGalaxy-Dawn\u002Fclaude-scholar\u002Fstargazers\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FGalaxy-Dawn\u002Fclaude-scholar?style=flat-square&color=yellow\" alt=\"Stars\"\u002F>\u003C\u002Fa>\n    \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FGalaxy-Dawn\u002Fclaude-scholar\u002Fnetwork\u002Fmembers\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fforks\u002FGalaxy-Dawn\u002Fclaude-scholar?style=flat-square\" alt=\"Forks\"\u002F>\u003C\u002Fa>\n    \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Flast-commit\u002FGalaxy-Dawn\u002Fclaude-scholar?style=flat-square\" alt=\"Last Commit\"\u002F>\n    \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-MIT-green?style=flat-square\" alt=\"License\"\u002F>\n    \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FClaude_Code-Compatible-blueviolet?style=flat-square\" alt=\"Claude Code\"\u002F>\n    \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FCodex_CLI-Compatible-blue?style=flat-square\" alt=\"Codex CLI\"\u002F>\n    \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FOpenCode-Compatible-orange?style=flat-square\" alt=\"OpenCode\"\u002F>\n  \u003C\u002Fp>\n\n\n  \u003Cstrong>Language\u003C\u002Fstrong>: \u003Ca href=\"README.md\">English\u003C\u002Fa> | \u003Ca href=\"README.zh-CN.md\">中文\u003C\u002Fa> | \u003Ca href=\"README.ja-JP.md\">日本語\u003C\u002Fa>\n\n\u003C\u002Fdiv>\n\n> Semi-automated research assistant for academic research and software development, especially for computer science and AI researchers. Supports [Claude Code](https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-code), [Codex CLI](https:\u002F\u002Fgithub.com\u002Fopenai\u002Fcodex), and [OpenCode](https:\u002F\u002Fgithub.com\u002Fopencode-ai\u002Fopencode) across literature review, coding, experiments, reporting, writing, and project knowledge management.\n\n  \u003Cp>\u003Cem>Branch note\u003C\u002Fem>: the \u003Ccode>main\u003C\u002Fcode> branch is the Claude Code workflow. If you use Codex CLI, please see the \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FGalaxy-Dawn\u002Fclaude-scholar\u002Ftree\u002Fcodex\">\u003Ccode>codex\u003C\u002Fcode> branch\u003C\u002Fa>. If you use OpenCode, please see the \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FGalaxy-Dawn\u002Fclaude-scholar\u002Ftree\u002Fopencode\">\u003Ccode>opencode\u003C\u002Fcode> branch\u003C\u002Fa>.\u003C\u002Fp>\n\n## Recent News\n\n- **2026-03-31**: **Zotero smart-import workflow docs aligned** — updated Claude Scholar's research-facing docs around the latest `zotero-mcp` public surface: `zotero_add_items_by_identifier` is now the default paper-import path, `zotero_reconcile_collection_duplicates` is the standard post-import cleanup step, source-aware PDF cascade behavior is documented more accurately, and public vs internal diagnostics are now clearly separated.\n- **2026-03-31**: **README onboarding refreshed** — clarified that Claude Scholar is especially well-suited to computer science and AI researchers, added practical getting-started scenarios after installation, improved prerequisite and branch guidance, and made the “existing local md files must be manually merged” expectation much more explicit.\n- **2026-03-31**: **Installer and hook behavior tightened** — the installer now preserves existing local `CLAUDE.md` \u002F `CLAUDE.zh-CN.md` files while installing repo-managed sidecar copies, and the default hook summaries were trimmed to reduce noisy temp-file \u002F uncommitted-file output while keeping safer write-guard behavior.\n- **2026-03-31**: **Japanese documentation added** — added Japanese docs for the main README plus `CLAUDE`, `MCP_SETUP`, and `OBSIDIAN_SETUP`, so the repository now has a more complete multilingual documentation surface.\n\n\u003Cdetails>\n\u003Csummary>View older changelog\u003C\u002Fsummary>\n\n- **2026-02-25**: **Codex CLI** support — added `codex` branch supporting [OpenAI Codex CLI](https:\u002F\u002Fgithub.com\u002Fopenai\u002Fcodex) with config.toml, 40 skills, 14 agents, and sandbox security\n- **2026-02-23**: Added `setup.sh` installer — backup-aware incremental updates for existing `~\u002F.claude`, auto-backup `settings.json`, additive hooks\u002FmcpServers\u002Fplugins merge\n- **2026-02-21**: **OpenCode** support — Claude Scholar now supports [OpenCode](https:\u002F\u002Fgithub.com\u002Fopencode-ai\u002Fopencode) as an alternative CLI; switch to the `opencode` branch for OpenCode-compatible configuration\n- **2026-02-20**: Bilingual config — translated `CLAUDE.md` to English for international readability; added `CLAUDE.zh-CN.md` as Chinese backup; Chinese users can switch with `cp CLAUDE.zh-CN.md CLAUDE.md`\n- **2026-02-15**: Zotero MCP integration — added `\u002Fzotero-review` and `\u002Fzotero-notes` commands, updated `research-ideation` skill with Zotero integration guide, enhanced `literature-reviewer` agent with Zotero MCP support for automated paper import, collection management, full-text reading, and citation export\n- **2026-02-14**: Hooks optimization — restructured `security-guard` to two-tier system (Block + Confirm), `skill-forced-eval` now groups skills into 6 categories with silent scan mode, `session-start` limits display to top 5, `session-summary` adds 30-day log auto-cleanup, `stop-summary` shows separate added\u002Fmodified\u002Fdeleted counts; removed deprecated shell scripts (lib\u002Fcommon.sh, lib\u002Fplatform.sh)\n- **2026-02-11**: Major update — added 10 new skills (research-ideation, results-analysis, citation-verification, review-response, paper-self-review, post-acceptance, daily-coding, frontend-design, ui-ux-pro-max, web-design-reviewer), 7 new agents, 8 research workflow commands, 2 new rules (security, experiment-reproducibility); restructured CLAUDE.md; 89 files changed\n- **2026-01-26**: Rewrote all Hooks to cross-platform Node.js; completely rewrote README; expanded ML paper writing knowledge base; merged PR #1 (cross-platform support)\n- **2026-01-25**: Project open-sourced, v1.0.0 released with 25 skills (architecture-design, bug-detective, git-workflow, kaggle-learner, scientific-writing, etc.), 2 agents (paper-miner, kaggle-miner), 30+ commands (including SuperClaude suite), 5 Shell Hooks, and 2 rules (coding-style, agents)\n\n\u003C\u002Fdetails>\n\n## Quick Navigation\n\n| Section | What it helps with |\n|---|---|\n| [Why Claude Scholar](#why-claude-scholar) | Understand the project positioning and target use cases. |\n| [Core Workflow](#core-workflow) | See the end-to-end research pipeline from ideation to publication. |\n| [Quick Start](#quick-start) | Install Claude Scholar in full, minimal, or selective mode. |\n| [Getting Started Scenarios](#getting-started-scenarios) | See a few realistic first-use scenarios after installation. |\n| [Integrations](#integrations) | Learn how Zotero and Obsidian fit into the workflow. |\n| [Primary Workflows](#primary-workflows) | Browse the main research and development workflows. |\n| [Supporting Workflows](#supporting-workflows) | See the background systems that strengthen the main workflow. |\n| [Documentation](#documentation) | Jump to setup docs, configuration, and templates. |\n| [Citation](#citation) | Cite Claude Scholar in papers, reports, or project docs. |\n\n## Why Claude Scholar\n\nClaude Scholar is **not** an end-to-end autonomous research system that tries to replace the researcher.\n\nIts core idea is simple:\n\n> **human decision-making stays at the center; the assistant accelerates the workflow around it.**\n\nThat means Claude Scholar is designed to help with the heavy, repetitive, and structure-sensitive parts of research — literature organization, note-taking, experiment analysis, reporting, and writing support — while still keeping the key judgments in human hands:\n\n- which problem is worth pursuing,\n- which papers actually matter,\n- which hypotheses are worth testing,\n- which results are convincing,\n- and what should be written, submitted, or abandoned.\n\nIn other words, Claude Scholar is a **semi-automated research assistant**, not a “fully automated scientist.”\n\n## Who This Is For\n\nClaude Scholar is especially well-suited to:\n\n- **computer science researchers** who move between literature review, coding, experiments, and paper writing,\n- **AI \u002F ML researchers** who need one assistant workflow spanning ideation, implementation, analysis, reporting, and rebuttal,\n- **research engineers and graduate students** who want stronger workflow structure without giving up human judgment,\n- and **software-heavy academic projects** that benefit from Zotero, Obsidian, CLI automation, and reproducible project memory.\n\nIt can still help in other research settings, but its current workflow design is most aligned with computer science, AI, and adjacent computational research.\n\n## Core Workflow\n\n- **Ideation**: turn a vague topic into concrete questions, research gaps, and an initial plan.\n- **Literature**: search, import, organize, and read papers through Zotero collections.\n- **Paper notes**: convert papers into structured reading notes and reusable claims.\n- **Knowledge base**: route durable knowledge into Obsidian across `Papers \u002F Knowledge \u002F Experiments \u002F Results \u002F Writing`, with round-level experiment reports stored under `Results\u002FReports\u002F`.\n- **Experiments**: track hypotheses, experiment lines, run history, findings, and next actions.\n- **Analysis**: generate strict statistics, real scientific figures, and analysis artifacts with `results-analysis`.\n- **Reporting**: produce a complete post-experiment report with `results-report`, then write it back into Obsidian.\n- **Writing and publication**: carry stable findings into literature reviews, papers, rebuttals, slides, posters, and promotion.\n\n## Quick Start\n\n### Requirements\n\n- [Claude Code](https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-code)\n- Git\n- (Optional) Python + [uv](https:\u002F\u002Fdocs.astral.sh\u002Fuv\u002F) for Python development\n- (Optional) [Zotero](https:\u002F\u002Fwww.zotero.org\u002F) + [Galaxy-Dawn\u002Fzotero-mcp](https:\u002F\u002Fgithub.com\u002FGalaxy-Dawn\u002Fzotero-mcp) for literature workflows\n- (Optional) [Obsidian](https:\u002F\u002Fobsidian.md\u002F) for project knowledge-base workflows\n\n### Option 1: Full Installation (Recommended)\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002FGalaxy-Dawn\u002Fclaude-scholar.git \u002Ftmp\u002Fclaude-scholar\nbash \u002Ftmp\u002Fclaude-scholar\u002Fscripts\u002Fsetup.sh\n```\n\n**Windows**: please use Git Bash or WSL to run the installer.\n\nThe installer is **backup-aware and incremental-update friendly**:\n- updates repo-managed `skills\u002Fcommands\u002Fagents\u002Frules\u002Fhooks\u002Fscripts\u002FCLAUDE*.md`,\n- backs up overwritten files to `~\u002F.claude\u002F.claude-scholar-backups\u002F\u003Ctimestamp>\u002F`,\n- backs up `settings.json` to `settings.json.bak`,\n- preserves an existing `~\u002F.claude\u002FCLAUDE.md` and installs the repo-managed version as `~\u002F.claude\u002FCLAUDE.scholar.md`,\n- preserves an existing `~\u002F.claude\u002FCLAUDE.zh-CN.md` and installs the repo-managed version as `~\u002F.claude\u002FCLAUDE.zh-CN.scholar.md`,\n- preserves your existing `env`, model\u002Fprovider settings, API keys, permissions, and current `mcpServers` values,\n- adds missing hook entries instead of replacing your entire hook set.\n\n**Important CLAUDE note**: if you already maintain your own `~\u002F.claude\u002FCLAUDE.md` or `~\u002F.claude\u002FCLAUDE.zh-CN.md`, review `~\u002F.claude\u002FCLAUDE.scholar.md` and `~\u002F.claude\u002FCLAUDE.zh-CN.scholar.md` after installation and manually merge the Claude Scholar sections you want into your own files. Do not assume the sidecar files are applied automatically.\n\nTo update later:\n\n```bash\ncd \u002Ftmp\u002Fclaude-scholar\ngit pull --ff-only\nbash scripts\u002Fsetup.sh\n```\n\n### Option 2: Minimal Installation\n\nInstall only a small research-focused subset:\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002FGalaxy-Dawn\u002Fclaude-scholar.git \u002Ftmp\u002Fclaude-scholar\nmkdir -p ~\u002F.claude\u002Fhooks ~\u002F.claude\u002Fskills\ncp \u002Ftmp\u002Fclaude-scholar\u002Fhooks\u002F*.js ~\u002F.claude\u002Fhooks\u002F\ncp -r \u002Ftmp\u002Fclaude-scholar\u002Fskills\u002Fml-paper-writing ~\u002F.claude\u002Fskills\u002F\ncp -r \u002Ftmp\u002Fclaude-scholar\u002Fskills\u002Fresearch-ideation ~\u002F.claude\u002Fskills\u002F\ncp -r \u002Ftmp\u002Fclaude-scholar\u002Fskills\u002Fresults-analysis ~\u002F.claude\u002Fskills\u002F\ncp -r \u002Ftmp\u002Fclaude-scholar\u002Fskills\u002Fresults-report ~\u002F.claude\u002Fskills\u002F\ncp -r \u002Ftmp\u002Fclaude-scholar\u002Fskills\u002Freview-response ~\u002F.claude\u002Fskills\u002F\ncp -r \u002Ftmp\u002Fclaude-scholar\u002Fskills\u002Fwriting-anti-ai ~\u002F.claude\u002Fskills\u002F\ncp -r \u002Ftmp\u002Fclaude-scholar\u002Fskills\u002Fgit-workflow ~\u002F.claude\u002Fskills\u002F\ncp -r \u002Ftmp\u002Fclaude-scholar\u002Fskills\u002Fbug-detective ~\u002F.claude\u002Fskills\u002F\n```\n\n**Post-install**: minimal\u002Fmanual install does **not** auto-merge `settings.json`; copy only the hooks or MCP entries you want from `settings.json.template`. If you already have your own `~\u002F.claude\u002FCLAUDE.md` or `~\u002F.claude\u002FCLAUDE.zh-CN.md`, also merge the relevant sections from this repo's Claude files into yours instead of blindly overwriting them.\n\n### Option 3: Selective Installation\n\nCopy only the parts you need:\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002FGalaxy-Dawn\u002Fclaude-scholar.git \u002Ftmp\u002Fclaude-scholar\ncd \u002Ftmp\u002Fclaude-scholar\n\ncp hooks\u002F*.js ~\u002F.claude\u002Fhooks\u002F\ncp -r skills\u002Flatex-conference-template-organizer ~\u002F.claude\u002Fskills\u002F\ncp -r skills\u002Farchitecture-design ~\u002F.claude\u002Fskills\u002F\ncp agents\u002Fpaper-miner.md ~\u002F.claude\u002Fagents\u002F\ncp rules\u002Fcoding-style.md ~\u002F.claude\u002Frules\u002F\ncp rules\u002Fagents.md ~\u002F.claude\u002Frules\u002F\n```\n\n**Post-install**: selective\u002Fmanual install does **not** auto-merge `settings.json`; copy only the hooks or MCP entries you actually want from `settings.json.template`. If you already have your own `~\u002F.claude\u002FCLAUDE.md` or `~\u002F.claude\u002FCLAUDE.zh-CN.md`, merge the relevant sections from this repo's Claude files into yours instead of blindly overwriting them.\n\n### Option 4: Plugin Marketplace Installation\n\n**Step 1: Install the Plugin**\n\n```bash\n\u002Fplugin marketplace add Galaxy-Dawn\u002Fclaude-scholar\n\u002Fplugin install claude-scholar@claude-scholar\n```\n\nThis auto-loads all skills, commands, agents, and hooks. During installation, you can choose the scope: user (all projects) or project (single project).\n\n**Step 2: Install Rules (Required)**\n\nClaude Code plugins cannot distribute rules automatically. Install them manually:\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002FGalaxy-Dawn\u002Fclaude-scholar.git \u002Ftmp\u002Fclaude-scholar\n\n# User-level (all projects)\nmkdir -p ~\u002F.claude\u002Frules\ncp \u002Ftmp\u002Fclaude-scholar\u002Frules\u002F*.md ~\u002F.claude\u002Frules\u002F\n\n# Or project-level (current project only)\nmkdir -p .claude\u002Frules\ncp \u002Ftmp\u002Fclaude-scholar\u002Frules\u002F*.md .claude\u002Frules\u002F\n```\n\n**Post-install**: plugin installation does **not** auto-load `CLAUDE.md` or configure `settings.json`; if you already have your own `~\u002F.claude\u002FCLAUDE.md` or `~\u002F.claude\u002FCLAUDE.zh-CN.md`, merge the relevant Claude Scholar sections into yours instead of assuming the plugin applies them automatically. If you need Zotero MCP or other integrations, see the [Integrations](#integrations) section for manual setup.\n\n## Getting Started Scenarios\n\nAfter installation, the simplest way to begin is to describe your task in natural language. You do not need to memorize the whole system first. Below are a few realistic starting points.\n\n### 1. Start a New Research Topic\n**You can say:**\n> Help me start research on [your topic]. I want a literature-grounded plan, the key open questions, and the next concrete steps.\n\n**What Claude Scholar will typically help with:**\n- clarify the topic and narrow the research question,\n- identify promising literature directions,\n- suggest an initial plan or hypothesis list,\n- optionally route the work into Zotero or Obsidian if you use them.\n\n### 2. Review a Zotero Collection\n**You can say:**\n> Review my Zotero collection on brain foundation models and summarize the main directions, gaps, and promising next steps.\n\n**Typical output:**\n- paper grouping by theme,\n- a short literature synthesis,\n- gap analysis,\n- candidate research directions worth pursuing next.\n\n### 3. Analyze Finished Experiment Results\n**You can say:**\n> Analyze the results in this experiment folder, check what changed across runs, and write a decision-oriented summary.\n\n**Typical output:**\n- metric comparison,\n- ablation or error-analysis suggestions,\n- a result summary that highlights what is solid, what is weak, and what to run next.\n\n### 4. Draft a Paper or Rebuttal Section\n**You can say:**\n> Help me draft the related work section for this project based on the current findings and paper notes.\n\nor:\n\n> Help me write a rebuttal draft for these reviewer comments.\n\n**Typical output:**\n- a structured section draft,\n- improved argument flow,\n- clearer claims and evidence mapping,\n- follow-up points that still need support or verification.\n\n### Practical Notes\n- Start with one concrete task, not a vague request for \"everything.\"\n- If you already maintain your own local `CLAUDE.md` files, merge the Claude Scholar sections you want into them instead of assuming sidecar files apply automatically.\n- Zotero and Obsidian are optional, but they become much more useful when you want durable literature notes or project memory rather than one-off chat output.\n\n## Platform Support\n\nClaude Scholar is maintained for:\n\n- **Claude Code** — the primary installation target.\n- **Codex CLI** — supported workflow and documentation are available in this repo ecosystem.\n- **OpenCode** — supported as an alternative CLI workflow.\n\nThe top-level workflow is the same: research, coding, experiments, reporting, and project knowledge management.\n\n## Integrations\n\n### Zotero\n\nUse Zotero when you want Claude Scholar to help with:\n- paper import via DOI \u002F arXiv \u002F URL,\n- collection-based reading workflows,\n- full-text access through Zotero MCP,\n- detailed paper notes and literature synthesis.\n\nSee [MCP_SETUP.md](.\u002FMCP_SETUP.md).\n\n### Obsidian\n\nUse Obsidian when you want Claude Scholar to maintain a filesystem-first research knowledge base:\n- `Papers\u002F`\n- `Experiments\u002F`\n- `Results\u002F`\n- `Results\u002FReports\u002F`\n- `Writing\u002F`\n- `Daily\u002F`\n\nSee [OBSIDIAN_SETUP.md](.\u002FOBSIDIAN_SETUP.md).\n\n## Primary Workflows\n\nComplete academic research lifecycle — 7 stages from idea to publication.\n\n### 1. Research Ideation (Zotero-Integrated)\n\nEnd-to-end research startup from idea generation to literature management.\n\n| Type | Name | One-line explanation |\n|---|---|---|\n| Skill | `research-ideation` | Turn vague topics into structured questions, gap analysis, and an initial research plan. |\n| Agent | `literature-reviewer` | Search, classify, and synthesize papers into an actionable literature picture. |\n| Command | `\u002Fresearch-init` | Start a new topic from literature search to Zotero organization and proposal drafting. |\n| Command | `\u002Fzotero-review` | Review an existing Zotero collection and generate a structured literature synthesis. |\n| Command | `\u002Fzotero-notes` | Batch-read a Zotero collection and create structured paper reading notes. |\n\n**How it works**\n- **5W1H Brainstorming**: turn a vague topic into structured questions (`What \u002F Why \u002F Who \u002F When \u002F Where \u002F How`).\n- **Literature Search & Import**: search papers, extract DOI\u002FarXiv\u002FURLs, import them into Zotero, and organize them into themed collections.\n- **PDF & Full Text**: attach PDFs when available, read full text when possible, and fall back to abstract-level analysis when necessary.\n- **Gap Analysis**: identify literature, methodological, application, interdisciplinary, or temporal gaps.\n- **Research Question & Planning**: convert the review into concrete questions, initial hypotheses, and next-step planning.\n\n**Typical output**\n- literature review notes\n- structured Zotero collection\n- project proposal \u002F research direction draft\n\n### 2. ML Project Development\n\nMaintainable ML project structure for experiment code and iteration.\n\n| Type | Name | One-line explanation |\n|---|---|---|\n| Skill | `architecture-design` | Define maintainable ML project structure when new registrable components or modules are introduced. |\n| Skill | `git-workflow` | Enforce branch hygiene, commit conventions, and safer collaboration workflows. |\n| Skill | `bug-detective` | Debug stack traces, shell failures, and code-path issues systematically. |\n| Agent | `code-reviewer` | Review modified code for correctness, maintainability, and implementation quality. |\n| Agent | `dev-planner` | Break complex engineering work into concrete implementation steps. |\n| Command | `\u002Fplan` | Create or refine an implementation plan before coding. |\n| Command | `\u002Fcommit` | Prepare a conventional commit for the current changes. |\n| Command | `\u002Fcode-review` | Run a focused review on the current code changes. |\n| Command | `\u002Ftdd` | Drive feature work through small, test-backed implementation steps. |\n\n**How it works**\n- **Structure**: use Factory \u002F Registry patterns for new ML components when appropriate.\n- **Code Quality**: keep files maintainable, typed, and config-driven.\n- **Debugging**: inspect stack traces, shell failures, and code-path issues systematically.\n- **Git Discipline**: use branch hygiene, conventional commits, and safer merge\u002Frebase workflows.\n\n### 3. Experiment Analysis\n\nStrict analysis of experimental results with scientific figures and report-ready artifacts.\n\n| Type | Name | One-line explanation |\n|---|---|---|\n| Skill | `results-analysis` | Produce a strict analysis bundle with rigorous statistics, real scientific figures, and analysis artifacts. |\n| Skill | `results-report` | Turn analysis artifacts into a complete post-experiment report with decisions, limitations, and next actions. |\n| Command | `\u002Fanalyze-results` | Run the full experiment workflow in one shot: strict analysis first, then final report generation. |\n\n**How it works**\n- **Data Processing**: read experiment logs, metrics files, and result directories.\n- **Statistical Testing**: run strict statistical checks such as t-test \u002F ANOVA \u002F Wilcoxon where appropriate.\n- **Visualization**: generate real scientific figures with interpretation guidance, not just vague plotting suggestions.\n- **Ablation & Comparison**: analyze component contribution, performance tradeoffs, and stability.\n- **Post-Experiment Reporting**: turn the analysis bundle into a full retrospective report with conclusions, limitations, and next actions.\n\n**Typical output**\n- `analysis-report.md`\n- `stats-appendix.md`\n- `figure-catalog.md`\n- `figures\u002F`\n- post-experiment summary report in Obsidian `Results\u002FReports\u002F`\n\n### 4. Paper Writing\n\nSystematic academic writing from structure setup to draft refinement.\n\n| Type | Name | One-line explanation |\n|---|---|---|\n| Skill | `ml-paper-writing` | Draft publication-oriented ML\u002FAI papers from repo context, evidence, and literature. |\n| Skill | `citation-verification` | Check references, metadata, and claim-citation alignment to prevent citation mistakes. |\n| Skill | `writing-anti-ai` | Reduce robotic phrasing and improve clarity, rhythm, and human academic tone. |\n| Skill | `latex-conference-template-organizer` | Clean messy conference templates into an Overleaf-ready writing structure. |\n| Agent | `paper-miner` | Mine strong papers for reusable writing patterns, structure, and venue expectations. |\n| Command | `\u002Fmine-writing-patterns` | Read a paper and merge reusable writing knowledge into the global paper-miner writing memory. |\n\n**How it works**\n- **Template Preparation**: clean conference templates into an Overleaf-ready structure.\n- **Citation Verification**: verify references, metadata, and claim-citation alignment.\n- **Systematic Writing**: draft sections from repo context, experiment evidence, and literature notes.\n- **Style Refinement**: reduce robotic phrasing and improve rhythm, clarity, and tone.\n\n### 5. Paper Self-Review\n\nQuality assurance before submission.\n\n| Type | Name | One-line explanation |\n|---|---|---|\n| Skill | `paper-self-review` | Audit structure, logic, citations, figures, and compliance before submission. |\n\n**How it works**\n- **Structure Check**: logical flow, section balance, and narrative coherence.\n- **Logic Validation**: claim-evidence alignment, assumption clarity, and argument consistency.\n- **Citation Audit**: reference correctness and completeness.\n- **Figure Quality**: caption completeness, readability, and accessibility.\n- **Compliance**: page limits, formatting, and disclosure requirements.\n\n### 6. Submission & Rebuttal\n\nSubmission preparation and review response workflow.\n\n| Type | Name | One-line explanation |\n|---|---|---|\n| Skill | `review-response` | Structure reviewer comments into an evidence-based rebuttal workflow. |\n| Agent | `rebuttal-writer` | Draft professional, respectful, and strategically organized rebuttal text. |\n| Command | `\u002Frebuttal` | Generate a complete rebuttal draft from review comments and evidence. |\n\n**How it works**\n- **Pre-submission Checks**: venue-specific formatting, anonymization, and checklist requirements.\n- **Review Analysis**: classify reviewer comments into actionable categories.\n- **Response Strategy**: decide whether to accept, defend, clarify, or propose new experiments.\n- **Rebuttal Writing**: generate structured, evidence-based responses with professional tone.\n\n### 7. Post-Acceptance Processing\n\nConference preparation and research promotion after acceptance.\n\n| Type | Name | One-line explanation |\n|---|---|---|\n| Skill | `post-acceptance` | Support talks, posters, and research promotion after acceptance. |\n| Command | `\u002Fpresentation` | Generate presentation structure and speaking guidance for the accepted work. |\n| Command | `\u002Fposter` | Organize the work into poster-ready content and layout guidance. |\n| Command | `\u002Fpromote` | Draft public-facing promotion content such as summaries, posts, or threads. |\n\n**How it works**\n- **Presentation**: prepare talk structure and slide guidance.\n- **Poster**: organize content into poster-ready layout and hierarchy.\n- **Promotion**: generate social media, blog, or summary material for broader communication.\n\n## Supporting Workflows\n\nThese workflows run in the background to strengthen the primary workflows.\n\n### Obsidian Project Knowledge Base\n\nUse Obsidian as the durable sink for project knowledge, not just as a note dump.\n\n| Type | Name | One-line explanation |\n|---|---|---|\n| Skill | `obsidian-project-memory` | Maintain the project-level Obsidian knowledge base and decide what durable knowledge should be written back. |\n| Skill | `obsidian-project-bootstrap` | Initialize an Obsidian knowledge base for a new or existing research project. |\n| Skill | `obsidian-research-log` | Record daily research progress, plans, ideas, and TODOs into the knowledge base. |\n| Skill | `obsidian-experiment-log` | Capture experiment setup, run history, outcomes, and follow-up actions in Obsidian. |\n| Command | `\u002Fobsidian-ingest` | Ingest a new Markdown file or folder into the correct place in the knowledge base. |\n| Command | `\u002Fobsidian-note` | Manage a single note lifecycle such as lookup, rename, archive, or purge. |\n| Command | `\u002Fobsidian-views` | Generate or refresh optional Obsidian views such as `.base` files. |\n\n**How it works**\n- bind an existing repo to an Obsidian vault,\n- route stable knowledge into `Papers \u002F Knowledge \u002F Experiments \u002F Results \u002F Writing`, with round-level experiment reports stored under `Results\u002FReports\u002F`,\n- keep `Daily\u002F` and project memory updated conservatively,\n- ingest new Markdown files into the correct canonical destination,\n- optionally generate extra views and canvases.\n\n**Note language configuration**\n\nGenerated and synced Obsidian notes resolve their language with this priority:\n1. project config: `.claude\u002Fproject-memory\u002Fregistry.yaml` -> `note_language`\n2. environment variable: `OBSIDIAN_NOTE_LANGUAGE`\n3. default: `en`\n\nNote: the file is currently named `registry.yaml` for historical reasons, but its on-disk format is JSON.\n\nPer-project example:\n\n```json\n{\n  \"projects\": {\n    \"my-project\": {\n      \"project_id\": \"my-project\",\n      \"vault_root\": \"\u002Fpath\u002Fto\u002Fvault\u002FResearch\u002Fmy-project\",\n      \"note_language\": \"zh-CN\"\n    }\n  }\n}\n```\n\nEnglish and Chinese section headings remain mutually compatible during sync, so older notes in either language can still be updated safely after switching configuration.\n\n### Automated Enforcement Workflow\n\nCross-platform hooks automate routine workflow checks and reminders.\n\n**Hooks**\n- `skill-forced-eval.js`\n- `session-start.js`\n- `session-summary.js`\n- `stop-summary.js`\n- `security-guard.js`\n\n**How it works**\n- **Before prompts**: evaluate applicable skills and surface relevant workflow hints.\n- **At session start**: show Git state, available commands, and project-memory context.\n- **At session end\u002Fstop**: summarize work and remind the user about minimum maintenance tasks.\n- **Security**: block catastrophic commands and require confirmation for dangerous but legitimate ones.\n\n### Knowledge Extraction Workflow\n\nSpecialized agents can mine reusable knowledge from papers and competitions.\n\n| Type | Name | One-line explanation |\n|---|---|---|\n| Agent | `paper-miner` | Extract reusable writing knowledge, structure patterns, and venue heuristics from strong papers. |\n| Agent | `kaggle-miner` | Extract engineering practices and solution patterns from strong Kaggle workflows. |\n\n**How it works**\n- extract writing patterns, venue expectations, and rebuttal strategies from papers,\n- extract engineering patterns and solution structure from Kaggle workflows,\n- feed those insights back into skills and reference material.\n\n### Skill Evolution System\n\nClaude Scholar also contains a self-improvement loop for its own skills.\n\n| Type | Name | One-line explanation |\n|---|---|---|\n| Skill | `skill-development` | Create new skills with clear triggers, structure, and progressive disclosure. |\n| Skill | `skill-quality-reviewer` | Review skills across content quality, organization, style, and structural integrity. |\n| Skill | `skill-improver` | Apply structured improvement plans to evolve existing skills. |\n\n**How it works**\n- create new skills with clear trigger descriptions,\n- review them across quality dimensions,\n- apply structured improvements and iterate.\n\n## Documentation\n\n- [MCP_SETUP.md](.\u002FMCP_SETUP.md) — Zotero\u002Fbrowser MCP setup\n- [OBSIDIAN_SETUP.md](.\u002FOBSIDIAN_SETUP.md) — Obsidian knowledge base workflow\n- [CLAUDE.md](.\u002FCLAUDE.md) — full local configuration, skill list, and workflow details\n- [CLAUDE.zh-CN.md](.\u002FCLAUDE.zh-CN.md) — Chinese version of the main configuration doc\n- [settings.json.template](.\u002Fsettings.json.template) — optional settings template for hooks\u002Fplugins\u002FMCP\n\n## Project Rules\n\nClaude Scholar includes project rules for:\n- coding style,\n- agent orchestration,\n- security,\n- experiment reproducibility.\n\nThese are reflected in the shipped rules and in `CLAUDE.md`.\n\n## Contributing\n\nIssues, PRs, and workflow improvements are welcome.\n\nIf you propose changes to installer behavior, Zotero workflows, or Obsidian routing, please include:\n- the user scenario,\n- the current limitation,\n- the expected behavior,\n- and any compatibility concerns.\n\n## License\n\nMIT License.\n\n## Citation\n\nIf Claude Scholar helps your research or engineering workflow, you can cite the repository as:\n\n```bibtex\n@misc{claude_scholar_2026,\n  title        = {Claude Scholar: Semi-automated research assistant for academic research and software development},\n  author       = {Gaorui Zhang},\n  year         = {2026},\n  howpublished = {\\url{https:\u002F\u002Fgithub.com\u002FGalaxy-Dawn\u002Fclaude-scholar}},\n  note         = {GitHub repository}\n}\n```\n\n## Acknowledgments\n\nBuilt with Claude Code CLI and enhanced by the open-source community.\n\n### References\n\nThis project is inspired by and builds upon excellent work from the community:\n\n- **[everything-claude-code](https:\u002F\u002Fgithub.com\u002Fanthropics\u002Feverything-claude-code)** - Comprehensive resource for Claude Code CLI\n- **[AI-research-SKILLs](https:\u002F\u002Fgithub.com\u002FzechenzhangAGI\u002FAI-research-SKILLs)** - Research-focused skills and configurations\n\nThese projects provided valuable insights and foundations for the research-oriented features in Claude Scholar.\n\n---\n\n**For data science, AI research, and academic writing.**\n\nRepository: [https:\u002F\u002Fgithub.com\u002FGalaxy-Dawn\u002Fclaude-scholar](https:\u002F\u002Fgithub.com\u002FGalaxy-Dawn\u002Fclaude-scholar)\n","\u003Cdiv align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FGalaxy-Dawn_claude-scholar_readme_dc32474688cf.png\" alt=\"Claude Scholar Logo\" width=\"100%\"\u002F>\n\n  \u003Cp>\n    \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FGalaxy-Dawn\u002Fclaude-scholar\u002Fstargazers\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FGalaxy-Dawn\u002Fclaude-scholar?style=flat-square&color=yellow\" alt=\"Stars\"\u002F>\u003C\u002Fa>\n    \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FGalaxy-Dawn\u002Fclaude-scholar\u002Fnetwork\u002Fmembers\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fforks\u002FGalaxy-Dawn\u002Fclaude-scholar?style=flat-square\" alt=\"Forks\"\u002F>\u003C\u002Fa>\n    \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Flast-commit\u002FGalaxy-Dawn\u002Fclaude-scholar?style=flat-square\" alt=\"Last Commit\"\u002F>\n    \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-MIT-green?style=flat-square\" alt=\"License\"\u002F>\n    \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FClaude_Code-Compatible-blueviolet?style=flat-square\" alt=\"Claude Code\"\u002F>\n    \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FCodex_CLI-Compatible-blue?style=flat-square\" alt=\"Codex CLI\"\u002F>\n    \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FOpenCode-Compatible-orange?style=flat-square\" alt=\"OpenCode\"\u002F>\n  \u003C\u002Fp>\n\n\n  \u003Cstrong>语言\u003C\u002Fstrong>: \u003Ca href=\"README.md\">English\u003C\u002Fa> | \u003Ca href=\"README.zh-CN.md\">中文\u003C\u002Fa> | \u003Ca href=\"README.ja-JP.md\">日本語\u003C\u002Fa>\n\n\u003C\u002Fdiv>\n\n> 面向学术研究与软件开发的半自动化科研助手，尤其适用于计算机科学和人工智能领域的研究人员。支持 [Claude Code](https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-code)、[Codex CLI](https:\u002F\u002Fgithub.com\u002Fopenai\u002Fcodex) 和 [OpenCode](https:\u002F\u002Fgithub.com\u002Fopencode-ai\u002Fopencode)，覆盖文献综述、编码、实验、报告撰写、写作以及项目知识管理等环节。\n\n  \u003Cp>\u003Cem>分支说明\u003C\u002Fem>: \u003Ccode>main\u003C\u002Fcode> 分支为 Claude Code 工作流。若您使用 Codex CLI，请查看 \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FGalaxy-Dawn\u002Fclaude-scholar\u002Ftree\u002Fcodex\">\u003Ccode>codex\u003C\u002Fcode> 分支\u003C\u002Fa>。若您使用 OpenCode，请查看 \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FGalaxy-Dawn\u002Fclaude-scholar\u002Ftree\u002Fopencode\">\u003Ccode>opencode\u003C\u002Fcode> 分支\u003C\u002Fa>。\u003C\u002Fp>\n\n## 最新动态\n\n- **2026-03-31**: **Zotero 智能导入工作流文档已对齐** — 更新了 Claude Scholar 面向科研的文档，围绕最新的 `zotero-mcp` 公开接口：`zotero_add_items_by_identifier` 现已成为默认的论文导入路径，`zotero_reconcile_collection_duplicates` 则是标准的导入后清理步骤；同时更准确地记录了基于来源的 PDF 层叠行为，并清晰区分了公开与内部诊断信息。\n- **2026-03-31**: **README 入门指南焕新** — 明确指出 Claude Scholar 特别适合计算机科学和人工智能领域的研究人员，新增了安装后的实用入门场景，优化了前置条件与分支指引，并将“现有本地 md 文件需手动合并”的要求表述得更加明确。\n- **2026-03-31**: **安装程序与钩子行为进一步收紧** — 安装程序在安装仓库管理的辅助副本时，会保留现有的本地 `CLAUDE.md` \u002F `CLAUDE.zh-CN.md` 文件；同时，默认钩子摘要经过精简，以减少临时文件及未提交文件带来的冗余输出，同时保持更安全的写入保护机制。\n- **2026-03-31**: **新增日文文档** — 为主 README 以及 `CLAUDE`、`MCP_SETUP` 和 `OBSIDIAN_SETUP` 添加了日文文档，使该仓库拥有了更为完整的多语言文档界面。\n\n\u003Cdetails>\n\u003Csummary>查看历史变更日志\u003C\u002Fsummary>\n\n- **2026-02-25**: **支持 Codex CLI** — 新增 `codex` 分支，支持 [OpenAI Codex CLI](https:\u002F\u002Fgithub.com\u002Fopenai\u002Fcodex)，配备 config.toml 配置文件、40 种技能、14 个代理及沙箱安全机制。\n- **2026-02-23**: 增加 `setup.sh` 安装脚本 — 提供备份感知的增量更新功能，针对现有 `~\u002F.claude` 目录进行操作；自动备份 `settings.json` 文件，并以追加方式合并钩子、mcpServers 和插件配置。\n- **2026-02-21**: **支持 OpenCode** — Claude Scholar 现在支持 [OpenCode](https:\u002F\u002Fgithub.com\u002Fopencode-ai\u002Fopencode) 作为替代 CLI；请切换至 `opencode` 分支以获得 OpenCode 兼容的配置。\n- **2026-02-20**: 双语配置 — 将 `CLAUDE.md` 翻译成英文以提升国际可读性；同时添加了 `CLAUDE.zh-CN.md` 作为中文备份；中文用户可通过 `cp CLAUDE.zh-CN.md CLAUDE.md` 进行切换。\n- **2026-02-15**: Zotero MCP 集成 — 新增 `\u002Fzotero-review` 和 `\u002Fzotero-notes` 命令，更新了 `research-ideation` 技能中的 Zotero 集成指南；增强了 `literature-reviewer` 代理的功能，使其能够通过 Zotero MCP 实现自动化论文导入、馆藏管理、全文阅读及引文导出。\n- **2026-02-14**: 钩子优化 — 重新构建了 `security-guard` 为两级系统（Block + Confirm）；`skill-forced-eval` 现在将技能分为 6 类，并新增静默扫描模式；`session-start` 限制显示前 5 项内容；`session-summary` 增加了 30 天日志自动清理功能；`stop-summary` 则分别显示新增、修改和删除的数量；同时移除了已废弃的 Shell 脚本（lib\u002Fcommon.sh、lib\u002Fplatform.sh）。\n- **2026-02-11**: 重大更新 — 新增 10 种技能（research-ideation、results-analysis、citation-verification、review-response、paper-self-review、post-acceptance、daily-coding、frontend-design、ui-ux-pro-max、web-design-reviewer）、7 个代理、8 条科研工作流命令以及 2 条新规则（安全、实验可重复性）；重构了 CLAUDE.md 文件；共修改了 89 个文件。\n- **2026-01-26**: 将所有钩子重写为跨平台 Node.js 脚本；全面重写了 README 文档；扩充了机器学习论文写作知识库；合并了 PR #1（跨平台支持）。\n- **2026-01-25**: 项目开源，发布 v1.0.0 版本，包含 25 种技能（architecture-design、bug-detective、git-workflow、kaggle-learner、scientific-writing 等）、2 个代理（paper-miner、kaggle-miner）、30 多条命令（包括 SuperClaude 套件）、5 个 Shell 钩子以及 2 条规则（coding-style、agents）。\n\n\u003C\u002Fdetails>\n\n## 快速导航\n\n| 片段 | 帮助解决的问题 |\n|---|---|\n| [为何选择 Claude Scholar](#why-claude-scholar) | 了解项目定位及目标使用场景。 |\n| [核心工作流程](#core-workflow) | 查看从创意构思到成果发表的端到端科研流程。 |\n| [快速入门](#quick-start) | 以完整、精简或选择性模式安装 Claude Scholar。 |\n| [入门场景](#getting-started-scenarios) | 浏览安装后的几个实际使用场景。 |\n| [集成](#integrations) | 了解 Zotero 和 Obsidian 如何融入工作流程。 |\n| [主要工作流程](#primary-workflows) | 浏览主要的科研与开发工作流程。 |\n| [支持性工作流程](#supporting-workflows) | 了解强化主工作流程的后台系统。 |\n| [文档](#documentation) | 跳转至设置文档、配置及模板页面。 |\n| [引用](#citation) | 在论文、报告或项目文档中引用 Claude Scholar。 |\n\n## 为什么选择 Claude Scholar\n\nClaude Scholar **并不是**一个试图取代研究人员的端到端自主研究系统。\n\n它的核心理念很简单：\n\n> **以人类决策为核心；助手则围绕这一核心加速工作流程。**\n\n这意味着，Claude Scholar 的设计目的是帮助处理研究中繁重、重复且对结构敏感的部分——文献整理、笔记记录、实验分析、报告撰写以及写作支持——同时仍将关键判断权掌握在人类手中：\n\n- 哪个问题值得深入研究，\n- 哪些论文真正重要，\n- 哪些假设值得检验，\n- 哪些结果具有说服力，\n- 以及哪些内容应该被撰写、提交或放弃。\n\n换句话说，Claude Scholar 是一款 **半自动化研究助手**，而非“完全自动化的科学家”。\n\n## 适合人群\n\nClaude Scholar 尤其适用于：\n\n- **计算机科学研究人员**，他们需要在文献综述、代码编写、实验和论文写作之间来回切换；\n- **人工智能\u002F机器学习研究人员**，他们需要一个贯穿构思、实现、分析、报告和答辩的综合助手工作流；\n- **研究工程师和研究生**，他们希望拥有更强大的工作流结构，同时不放弃人类的判断力；\n- 以及那些受益于 Zotero、Obsidian、CLI 自动化和可复现项目记忆的 **软件密集型学术项目**。\n\n它同样可以在其他研究领域发挥作用，但其当前的工作流设计与计算机科学、人工智能及相关的计算研究最为契合。\n\n## 核心工作流\n\n- **构思阶段**：将模糊的主题转化为具体的研究问题、研究空白以及初步计划。\n- **文献管理**：通过 Zotero 文献库进行文献检索、导入、整理和阅读。\n- **论文笔记**：将论文转化为结构化的阅读笔记和可重复使用的论点。\n- **知识库**：将持久的知识归档至 Obsidian 中的 `Papers \u002F Knowledge \u002F Experiments \u002F Results \u002F Writing` 模块，并将每一轮实验报告存储在 `Results\u002FReports\u002F` 下。\n- **实验阶段**：跟踪假设、实验方案、运行历史、发现结果及后续行动。\n- **分析阶段**：使用 `results-analysis` 工具生成严谨的统计分析、科学图表及分析成果。\n- **报告阶段**：利用 `results-report` 工具生成完整的实验后报告，并将其回写到 Obsidian 中。\n- **写作与发表阶段**：将稳定的实验结果融入文献综述、论文、答辩材料、幻灯片、海报及职称晋升材料中。\n\n## 快速入门\n\n### 系统要求\n\n- [Claude Code](https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-code)\n- Git\n- （可选）Python + [uv](https:\u002F\u002Fdocs.astral.sh\u002Fuv\u002F) 用于 Python 开发\n- （可选）[Zotero](https:\u002F\u002Fwww.zotero.org\u002F) + [Galaxy-Dawn\u002Fzotero-mcp](https:\u002F\u002Fgithub.com\u002FGalaxy-Dawn\u002Fzotero-mcp) 用于文献管理工作流\n- （可选）[Obsidian](https:\u002F\u002Fobsidian.md\u002F) 用于项目知识库工作流\n\n### 选项 1：完整安装（推荐）\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002FGalaxy-Dawn\u002Fclaude-scholar.git \u002Ftmp\u002Fclaude-scholar\nbash \u002Ftmp\u002Fclaude-scholar\u002Fscripts\u002Fsetup.sh\n```\n\n**Windows 用户**：请使用 Git Bash 或 WSL 来运行安装程序。\n\n该安装程序具备 **备份功能且支持增量更新**：\n\n- 更新由仓库管理的 `skills\u002Fcommands\u002Fagents\u002Frules\u002Fhooks\u002Fscripts\u002FCLAUDE*.md` 文件；\n- 将被覆盖的文件备份至 `~\u002F.claude\u002F.claude-scholar-backups\u002F\u003Ctimestamp>\u002F` 目录下；\n- 备份 `settings.json` 文件为 `settings.json.bak`；\n- 保留现有的 `~\u002F.claude\u002FCLAUDE.md` 文件，并将仓库管理的版本安装为 `~\u002F.claude\u002FCLAUDE.scholar.md`；\n- 保留现有的 `~\u002F.claude\u002FCLAUDE.zh-CN.md` 文件，并将仓库管理的版本安装为 `~\u002F.claude\u002FCLAUDE.zh-CN.scholar.md`；\n- 保留您现有的 `env`、模型\u002F提供商设置、API 密钥、权限及当前的 `mcpServers` 值；\n- 添加缺失的钩子条目，而不是替换您的整个钩子集。\n\n**关于 CLAUDE 的重要提示**：如果您已经维护了自己的 `~\u002F.claude\u002FCLAUDE.md` 或 `~\u002F.claude\u002FCLAUDE.zh-CN.md` 文件，请在安装后仔细审阅 `~\u002F.claude\u002FCLAUDE.scholar.md` 和 `~\u002F.claude\u002FCLAUDE.zh-CN.scholar.md` 文件，并手动将您希望保留的 Claude Scholar 部分合并到您自己的文件中。请勿假定这些辅助文件会自动应用。\n\n后续更新方法如下：\n\n```bash\ncd \u002Ftmp\u002Fclaude-scholar\ngit pull --ff-only\nbash scripts\u002Fsetup.sh\n```\n\n### 选项 2：最小化安装\n\n仅安装少量专注于研究的功能模块：\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002FGalaxy-Dawn\u002Fclaude-scholar.git \u002Ftmp\u002Fclaude-scholar\nmkdir -p ~\u002F.claude\u002Fhooks ~\u002F.claude\u002Fskills\ncp \u002Ftmp\u002Fclaude-scholar\u002Fhooks\u002F*.js ~\u002F.claude\u002Fhooks\u002F\ncp -r \u002Ftmp\u002Fclaude-scholar\u002Fskills\u002Fml-paper-writing ~\u002F.claude\u002Fskills\u002F\ncp -r \u002Ftmp\u002Fclaude-scholar\u002Fskills\u002Fresearch-ideation ~\u002F.claude\u002Fskills\u002F\ncp -r \u002Ftmp\u002Fclaude-scholar\u002Fskills\u002Fresults-analysis ~\u002F.claude\u002Fskills\u002F\ncp -r \u002Ftmp\u002Fclaude-scholar\u002Fskills\u002Fresults-report ~\u002F.claude\u002Fskills\u002F\ncp -r \u002Ftmp\u002Fclaude-scholar\u002Fskills\u002Freview-response ~\u002F.claude\u002Fskills\u002F\ncp -r \u002Ftmp\u002Fclaude-scholar\u002Fskills\u002Fwriting-anti-ai ~\u002F.claude\u002Fskills\u002F\ncp -r \u002Ftmp\u002Fclaude-scholar\u002Fskills\u002Fgit-workflow ~\u002F.claude\u002Fskills\u002F\ncp -r \u002Ftmp\u002Fclaude-scholar\u002Fskills\u002Fbug-detective ~\u002F.claude\u002Fskills\u002F\n```\n\n**安装后注意事项**：最小化\u002F手动安装不会自动合并 `settings.json` 文件；请仅从 `settings.json.template` 中复制您所需的钩子或 MCP 条目。如果您已经拥有自己的 `~\u002F.claude\u002FCLAUDE.md` 或 `~\u002F.claude\u002FCLAUDE.zh-CN.md` 文件，则应将本仓库中的相关部分手动合并到您自己的文件中，而不是盲目覆盖。\n\n### 选项 3：选择性安装\n\n仅复制您需要的部分：\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002FGalaxy-Dawn\u002Fclaude-scholar.git \u002Ftmp\u002Fclaude-scholar\ncd \u002Ftmp\u002Fclaude-scholar\n\ncp hooks\u002F*.js ~\u002F.claude\u002Fhooks\u002F\ncp -r skills\u002Flatex-conference-template-organizer ~\u002F.claude\u002Fskills\u002F\ncp -r skills\u002Farchitecture-design ~\u002F.claude\u002Fskills\u002F\ncp agents\u002Fpaper-miner.md ~\u002F.claude\u002Fagents\u002F\ncp rules\u002Fcoding-style.md ~\u002F.claude\u002Frules\u002F\ncp rules\u002Fagents.md ~\u002F.claude\u002Frules\u002F\n```\n\n**安装后注意事项**：选择性\u002F手动安装不会自动合并 `settings.json` 文件；请仅从 `settings.json.template` 中复制您实际需要的钩子或 MCP 条目。如果您已经拥有自己的 `~\u002F.claude\u002FCLAUDE.md` 或 `~\u002F.claude\u002FCLAUDE.zh-CN.md` 文件，则应将本仓库中的相关部分手动合并到您自己的文件中而不是盲目覆盖。\n\n### 选项 4：插件市场安装\n\n**步骤 1：安装插件**\n\n```bash\n\u002Fplugin marketplace add Galaxy-Dawn\u002Fclaude-scholar\n\u002Fplugin install claude-scholar@claude-scholar\n```\n\n这将自动加载所有技能、命令、代理和钩子。在安装过程中，您可以选择作用范围：用户级（所有项目）或项目级（单个项目）。\n\n**步骤 2：安装规则（必需）**\n\nClaude Code 插件无法自动分发规则，因此需手动安装：\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002FGalaxy-Dawn\u002Fclaude-scholar.git \u002Ftmp\u002Fclaude-scholar\n\n# 用户级（所有项目）\nmkdir -p ~\u002F.claude\u002Frules\ncp \u002Ftmp\u002Fclaude-scholar\u002Frules\u002F*.md ~\u002F.claude\u002Frules\u002F\n\n# 或者项目级（仅当前项目）\nmkdir -p .claude\u002Frules\ncp \u002Ftmp\u002Fclaude-scholar\u002Frules\u002F*.md .claude\u002Frules\u002F\n```\n\n**安装后操作**: 插件安装并不会自动加载 `CLAUDE.md` 或配置 `settings.json`；如果您已经有自己的 `~\u002F.claude\u002FCLAUDE.md` 或 `~\u002F.claude\u002FCLAUDE.zh-CN.md`，请将 Claude Scholar 的相关部分合并到您的文件中，而不要假设插件会自动应用这些内容。如果需要 Zotero MCP 或其他集成，请参阅 [集成](#integrations) 部分进行手动设置。\n\n## 入门场景\n\n安装完成后，最简单的开始方式是用自然语言描述您的任务。您无需先记住整个系统。以下是一些实际的起点。\n\n### 1. 开始一个新的研究课题\n**您可以这样说：**\n> 帮我开始关于[您的课题]的研究。我想要一个基于文献的计划、关键的开放性问题以及下一步的具体行动。\n\n**Claude Scholar 通常会帮助您：**\n- 澄清课题并缩小研究范围，\n- 确定有前景的文献方向，\n- 提出初步的计划或假设清单，\n- 如果您使用 Zotero 或 Obsidian，还可以选择将工作流程导入其中。\n\n### 2. 审查 Zotero 文献库\n**您可以这样说：**\n> 审查我在 Zotero 中关于大脑基础模型的文献库，并总结主要方向、研究空白以及有前景的下一步行动。\n\n**典型输出：**\n- 按主题对论文进行分组，\n- 简短的文献综述，\n- 研究空白分析，\n- 下一步值得探索的研究方向建议。\n\n### 3. 分析已完成的实验结果\n**您可以这样说：**\n> 分析这个实验文件夹中的结果，检查各次运行之间的变化，并撰写一份以决策为导向的总结。\n\n**典型输出：**\n- 指标对比，\n- 割除实验或错误分析建议，\n- 结果总结，突出哪些结论可靠、哪些较弱，以及接下来应该进行哪些实验。\n\n### 4. 撰写论文或反驳意见的部分\n**您可以这样说：**\n> 根据目前的研究成果和论文笔记，帮我起草该项目的相关工作部分。\n\n或者：\n\n> 帮我为这些审稿人的意见撰写一份反驳草稿。\n\n**典型输出：**\n- 结构化的段落草稿，\n- 更流畅的论证逻辑，\n- 更清晰的观点与证据对应关系，\n- 需要进一步支持或验证的后续要点。\n\n### 实用提示\n- 从一个具体的任务开始，而不是笼统地要求“全部”。\n- 如果您已经维护了自己的本地 `CLAUDE.md` 文件，可以将您需要的 Claude Scholar 部分合并进去，而不应假设附带文件会自动生效。\n- Zotero 和 Obsidian 是可选的，但当您希望获得持久的文献笔记或项目记忆，而非一次性的对话输出时，它们会变得更加有用。\n\n## 平台支持\n\nClaude Scholar 目前支持以下平台：\n\n- **Claude Code** — 主要的安装目标。\n- **Codex CLI** — 本仓库生态系统中提供了支持的工作流和文档。\n- **OpenCode** — 作为替代的 CLI 工作流也受到支持。\n\n顶层的工作流程保持一致：研究、编码、实验、报告以及项目知识管理。\n\n## 集成\n\n### Zotero\n\n当您希望 Claude Scholar 协助以下工作时，可以使用 Zotero：\n- 通过 DOI \u002F arXiv \u002F URL 导入论文，\n- 基于文献库的阅读工作流，\n- 通过 Zotero MCP 获取全文，\n- 创建详细的论文笔记和文献综述。\n\n详情请参阅 [MCP_SETUP.md](.\u002FMCP_SETUP.md)。\n\n### Obsidian\n\n当您希望 Claude Scholar 维护一个以文件系统为核心的科研知识库时，可以使用 Obsidian：\n- `Papers\u002F`\n- `Experiments\u002F`\n- `Results\u002F`\n- `Results\u002FReports\u002F`\n- `Writing\u002F`\n- `Daily\u002F`\n\n详情请参阅 [OBSIDIAN_SETUP.md](.\u002FOBSIDIAN_SETUP.md)。\n\n## 主要工作流\n\n完整的学术研究生命周期——从想法到发表的7个阶段。\n\n### 1. 研究构思（集成 Zotero）\n\n从产生想法到文献管理的端到端研究启动流程。\n\n| 类型 | 名称 | 一句话说明 |\n|---|---|---|\n| 技能 | `research-ideation` | 将模糊的主题转化为结构化的研究问题、空白分析及初步的研究计划。 |\n| 代理 | `literature-reviewer` | 搜索、分类并综合论文，形成可操作的文献图景。 |\n| 命令 | `\u002Fresearch-init` | 从文献搜索开始，组织 Zotero 文献库并起草研究提案，开启新课题。 |\n| 命令 | `\u002Fzotero-review` | 审查现有的 Zotero 文献库，生成结构化的文献综述。 |\n| 命令 | `\u002Fzotero-notes` | 批量阅读 Zotero 文献库，创建结构化的论文阅读笔记。 |\n\n**工作原理**\n- **5W1H 头脑风暴**：将模糊的主题转化为结构化的研究问题（What \u002F Why \u002F Who \u002F When \u002F Where \u002F How）。\n- **文献搜索与导入**：搜索论文，提取 DOI\u002FarXiv\u002FURL，将其导入 Zotero，并按主题整理成不同的文献库。\n- **PDF 与全文**：在可用时附加 PDF，尽可能阅读全文，必要时退回到摘要层面的分析。\n- **空白分析**：识别文献、方法论、应用、跨学科或时间上的空白。\n- **研究问题与规划**：将综述内容转化为具体的科学问题、初始假设以及下一步的行动计划。\n\n**典型输出**\n- 文献综述笔记，\n- 结构化的 Zotero 文献库，\n- 项目提案 \u002F 研究方向草案。\n\n### 2. 机器学习项目开发\n\n用于实验代码和迭代的可维护的机器学习项目结构。\n\n| 类型 | 名称 | 一句话说明 |\n|---|---|---|\n| 技能 | `architecture-design` | 在引入新的可注册组件或模块时，定义可维护的机器学习项目结构。 |\n| 技能 | `git-workflow` | 强制执行分支规范、提交约定以及更安全的合作工作流。 |\n| 技能 | `bug-detective` | 系统地调试堆栈跟踪、Shell 错误以及代码路径问题。 |\n| 代理 | `code-reviewer` | 审查修改后的代码，确保其正确性、可维护性和实现质量。 |\n| 代理 | `dev-planner` | 将复杂的工程任务分解为具体的实施步骤。 |\n| 命令 | `\u002Fplan` | 在编码之前制定或完善实施计划。 |\n| 命令 | `\u002Fcommit` | 为当前的更改准备符合规范的提交信息。 |\n| 命令 | `\u002Fcode-review` | 对当前的代码变更进行重点审查。 |\n| 命令 | `\u002Ftdd` | 通过小规模、测试驱动的实现步骤推进功能开发。 |\n\n**工作原理**\n- **结构设计**：在适当的情况下，使用工厂模式或注册表模式来组织新的机器学习组件。\n- **代码质量**：保持代码文件的可维护性、类型安全以及配置驱动。\n- **调试**：系统地检查堆栈跟踪、Shell 错误以及代码路径问题。\n- **Git 规范**：遵循分支卫生、规范化的提交以及更安全的合并\u002F变基工作流。\n\n### 3. 实验分析\n\n对实验结果进行严格分析，生成科学图表和可供报告使用的成果。\n\n| 类型 | 名称 | 一句话说明 |\n|---|---|---|\n| 技能 | `results-analysis` | 产出包含严谨统计、真实科学图表及分析成果的严格分析包。 |\n| 技能 | `results-report` | 将分析成果转化为包含决策、局限性及后续行动的完整实验后报告。 |\n| 命令 | `\u002Fanalyze-results` | 一次性执行完整的实验工作流：先进行严格分析，再生成最终报告。 |\n\n**工作原理**\n- **数据处理**：读取实验日志、指标文件及结果目录。\n- **统计检验**：在适当情况下运行严格的统计检验，如t检验、方差分析或威尔科克森检验。\n- **可视化**：生成带有解读指导的真实科学图表，而非仅提供模糊的绘图建议。\n- **消融与对比**：分析各组件的贡献、性能权衡及稳定性。\n- **实验后报告**：将分析包转化为包含结论、局限性和下一步行动的完整回顾报告。\n\n**典型输出**\n- `analysis-report.md`\n- `stats-appendix.md`\n- `figure-catalog.md`\n- `figures\u002F`\n- Obsidian 格式的实验后总结报告，位于 `Results\u002FReports\u002F` 目录下。\n\n### 4. 论文写作\n\n从结构搭建到初稿润色的系统性学术写作流程。\n\n| 类型 | 名称 | 一句话说明 |\n|---|---|---|\n| 技能 | `ml-paper-writing` | 根据代码库上下文、实验证据及文献资料，撰写面向发表的机器学习\u002F人工智能论文。 |\n| 技能 | `citation-verification` | 检查参考文献、元数据及论点与引用的一致性，防止引用错误。 |\n| 技能 | `writing-anti-ai` | 减少机械化措辞，提升清晰度、节奏感及人文学术语气。 |\n| 技能 | `latex-conference-template-organizer` | 将杂乱无章的会议模板整理成可直接用于 Overleaf 的写作框架。 |\n| 代理 | `paper-miner` | 从优秀论文中挖掘可复用的写作模式、结构及目标会议的要求。 |\n| 命令 | `\u002Fmine-writing-patterns` | 阅读一篇论文，并将其中可复用的写作知识整合到全局的论文挖掘器记忆中。 |\n\n**工作原理**\n- **模板准备**：将会议模板清理为适合 Overleaf 使用的结构。\n- **引文验证**：核对参考文献、元数据以及论点与引用是否一致。\n- **系统化写作**：基于代码库上下文、实验证据及文献笔记，逐部分撰写论文。\n- **风格优化**：减少机械化表达，提升节奏感、清晰度和语气。\n\n### 5. 论文自审\n\n投稿前的质量保证。\n\n| 类型 | 名称 | 一句话说明 |\n|---|---|---|\n| 技能 | `paper-self-review` | 在投稿前对论文的结构、逻辑、引用、图表及合规性进行审核。 |\n\n**工作原理**\n- **结构检查**：逻辑流程、各部分平衡及叙述连贯性。\n- **逻辑验证**：论点与证据是否匹配、假设是否清晰以及论证是否一致。\n- **引用审计**：检查参考文献的正确性和完整性。\n- **图表质量**：检查标题的完整性、可读性及可访问性。\n- **合规性**：页数限制、格式要求及披露义务。\n\n### 6. 投稿与回复审稿意见\n\n投稿准备及审稿意见回复流程。\n\n| 类型 | 名称 | 一句话说明 |\n|---|---|---|\n| 技能 | `review-response` | 将审稿意见分类整理为基于证据的回复流程。 |\n| 代理 | `rebuttal-writer` | 撰写专业、尊重且策略性强的回复文本。 |\n| 命令 | `\u002Frebuttal` | 根据审稿意见和证据生成完整的回复草稿。 |\n\n**工作原理**\n- **投稿前检查**：针对特定会议的格式要求、匿名化措施及清单式检查项。\n- **审稿意见分析**：将审稿意见归类为可操作的子类别。\n- **回复策略**：决定是接受、辩护、澄清，还是提出新的实验方案。\n- **回复撰写**：以专业语气生成结构化、基于证据的回复内容。\n\n### 7. 接受后的处理\n\n论文被接收后，用于会议准备及研究推广的工作流程。\n\n| 类型 | 名称 | 一句话说明 |\n|---|---|---|\n| 技能 | `post-acceptance` | 支持论文被接收后的报告演讲、海报展示及研究推广。 |\n| 命令 | `\u002Fpresentation` | 为已接收的论文生成演示文稿结构及演讲指导。 |\n| 命令 | `\u002Fposter` | 将研究成果整理为适合海报展示的内容及布局指导。 |\n| 命令 | `\u002Fpromote` | 撰写面向公众的宣传材料，如摘要、博文或系列推文。 |\n\n**工作原理**\n- **演示文稿**：准备演讲结构及幻灯片使用指南。\n- **海报**：将内容组织成适合海报展示的布局与层次结构。\n- **推广**：生成社交媒体、博客或摘要等材料，以便更广泛地传播研究成果。\n\n## 支撑性工作流程\n\n这些工作流程在后台运行，以强化主要工作流程。\n\n### Obsidian 项目知识库\n\n将 Obsidian 用作项目知识的持久存储，而不仅仅是笔记的堆栈。\n\n| 类型 | 名称 | 一句话说明 |\n|---|---|---|\n| 技能 | `obsidian-project-memory` | 维护项目级别的 Obsidian 知识库，并决定哪些持久性知识应当被记录回其中。 |\n| 技能 | `obsidian-project-bootstrap` | 为新或现有的研究项目初始化一个 Obsidian 知识库。 |\n| 技能 | `obsidian-research-log` | 将每日的研究进展、计划、想法和待办事项记录到知识库中。 |\n| 技能 | `obsidian-experiment-log` | 在 Obsidian 中记录实验设置、运行历史、结果及后续行动。 |\n| 命令 | `\u002Fobsidian-ingest` | 将新的 Markdown 文件或文件夹摄入知识库中的正确位置。 |\n| 命令 | `\u002Fobsidian-note` | 管理单个笔记的生命周期，例如查找、重命名、归档或清除。 |\n| 命令 | `\u002Fobsidian-views` | 生成或刷新可选的 Obsidian 视图，例如 `.base` 文件。 |\n\n**工作原理**\n- 将现有仓库绑定到 Obsidian 保险库，\n- 将稳定的知识路由至 `Papers \u002F Knowledge \u002F Experiments \u002F Results \u002F Writing`，其中每一轮实验报告存储在 `Results\u002FReports\u002F` 下，\n- 保守地保持 `Daily\u002F` 和项目记忆的更新，\n- 将新的 Markdown 文件摄入正确的规范目标位置，\n- 可选地生成额外的视图和画布。\n\n**笔记语言配置**\n\n生成并同步的 Obsidian 笔记会按照以下优先级解析其语言：\n1. 项目配置：`.claude\u002Fproject-memory\u002Fregistry.yaml` -> `note_language`\n2. 环境变量：`OBSIDIAN_NOTE_LANGUAGE`\n3. 默认：`en`\n\n注意：该文件目前仍名为 `registry.yaml` 是出于历史原因，但其磁盘格式实际上是 JSON。\n\n项目示例：\n\n```json\n{\n  \"projects\": {\n    \"my-project\": {\n      \"project_id\": \"my-project\",\n      \"vault_root\": \"\u002Fpath\u002Fto\u002Fvault\u002FResearch\u002Fmy-project\",\n      \"note_language\": \"zh-CN\"\n    }\n  }\n}\n```\n\n英文和中文的小节标题在同步过程中仍然相互兼容，因此即使切换配置后，旧的笔记也可以安全地继续更新。\n\n### 自动化执行工作流\n\n跨平台钩子自动化了常规的工作流检查和提醒。\n\n**钩子**\n- `skill-forced-eval.js`\n- `session-start.js`\n- `session-summary.js`\n- `stop-summary.js`\n- `security-guard.js`\n\n**工作原理**\n- **提示之前**：评估适用技能并提示相关工作流建议。\n- **会话开始时**：显示 Git 状态、可用命令以及项目记忆上下文。\n- **会话结束\u002F停止时**：总结工作并提醒用户完成最低限度的维护任务。\n- **安全性**：阻止灾难性的命令，并要求对危险但合法的命令进行确认。\n\n### 知识提取工作流\n\n专业代理可以从论文和竞赛中挖掘可复用的知识。\n\n| 类型 | 名称 | 一句话说明 |\n|---|---|---|\n| 代理 | `paper-miner` | 从优秀的论文中提取可复用的写作知识、结构模式和投稿经验法则。 |\n| 代理 | `kaggle-miner` | 从优秀的 Kaggle 工作流程中提取工程实践和解决方案模式。 |\n\n**工作原理**\n- 从论文中提取写作模式、投稿期望和反驳策略，\n- 从 Kaggle 流程中提取工程模式和解决方案结构，\n- 将这些见解反馈到技能和参考资料中。\n\n### 技能进化系统\n\nClaude Scholar 还包含一个用于自我改进的技能循环。\n\n| 类型 | 名称 | 一句话说明 |\n|---|---|---|\n| 技能 | `skill-development` | 创建具有明确触发条件、结构和渐进式披露的新技能。 |\n| 技能 | `skill-quality-reviewer` | 从内容质量、组织、风格和结构完整性等方面审查技能。 |\n| 技能 | `skill-improver` | 应用结构化的改进计划来提升现有技能。\n\n**工作原理**\n- 创建带有清晰触发描述的新技能，\n- 从多个质量维度对其进行审查，\n- 应用结构化的改进并迭代。\n\n## 文档\n\n- [MCP_SETUP.md](.\u002FMCP_SETUP.md) — Zotero\u002F浏览器 MCP 设置\n- [OBSIDIAN_SETUP.md](.\u002FOBSIDIAN_SETUP.md) — Obsidian 知识库工作流\n- [CLAUDE.md](.\u002FCLAUDE.md) — 完整的本地配置、技能列表和工作流细节\n- [CLAUDE.zh-CN.md](.\u002FCLAUDE.zh-CN.md) — 主配置文档的中文版\n- [settings.json.template](.\u002Fsettings.json.template) — 钩子\u002F插件\u002FMCP 的可选设置模板\n\n## 项目规则\n\nClaude Scholar 包含以下方面的项目规则：\n- 编码风格，\n- 代理编排，\n- 安全，\n- 实验可重复性。\n\n这些规则体现在随附的规则文件以及 `CLAUDE.md` 中。\n\n## 贡献\n\n欢迎提交问题、拉取请求和工作流改进。\n\n如果您提议更改安装程序行为、Zotero 工作流或 Obsidian 路由，请包括：\n- 用户场景，\n- 当前限制，\n- 预期行为，\n- 以及任何兼容性顾虑。\n\n## 许可证\n\nMIT 许可证。\n\n## 引用\n\n如果 Claude Scholar 对您的研究或工程工作有所帮助，您可以这样引用该仓库：\n\n```bibtex\n@misc{claude_scholar_2026,\n  title        = {Claude Scholar: 半自动化的学术研究与软件开发助手},\n  author       = {Gaorui Zhang},\n  year         = {2026},\n  howpublished = {\\url{https:\u002F\u002Fgithub.com\u002FGalaxy-Dawn\u002Fclaude-scholar}},\n  note         = {GitHub 仓库}\n}\n```\n\n## 致谢\n\n基于 Claude Code CLI 构建，并由开源社区进一步完善。\n\n### 参考文献\n\n本项目受到社区优秀工作的启发，并在此基础上构建：\n\n- **[everything-claude-code](https:\u002F\u002Fgithub.com\u002Fanthropics\u002Feverything-claude-code)** - 关于 Claude Code CLI 的全面资源\n- **[AI-research-SKILLs](https:\u002F\u002Fgithub.com\u002FzechenzhangAGI\u002FAI-research-SKILLs)** - 以研究为导向的技能和配置\n\n这些项目为 Claude Scholar 中面向研究的功能提供了宝贵的见解和基础。\n\n---\n\n**适用于数据科学、人工智能研究和学术写作。**\n\n仓库：[https:\u002F\u002Fgithub.com\u002FGalaxy-Dawn\u002Fclaude-scholar](https:\u002F\u002Fgithub.com\u002FGalaxy-Dawn\u002Fclaude-scholar)","# Claude Scholar 快速上手指南\n\nClaude Scholar 是一款专为计算机科学和 AI 研究人员设计的**半自动化研究助手**。它并非要取代研究者，而是将人类决策置于核心地位，加速文献综述、代码实验、报告撰写及项目知识管理等繁琐流程。支持 Claude Code、Codex CLI 和 OpenCode。\n\n## 环境准备\n\n在开始之前，请确保您的系统满足以下要求：\n\n*   **操作系统**: Linux, macOS 或 Windows (Windows 用户需使用 **Git Bash** 或 **WSL**)\n*   **核心依赖**:\n    *   [Claude Code](https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-code) (主分支默认支持)\n    *   Git\n*   **可选依赖** (根据需求安装):\n    *   **Python 开发**: Python + [uv](https:\u002F\u002Fdocs.astral.sh\u002Fuv\u002F)\n    *   **文献管理工作流**: [Zotero](https:\u002F\u002Fwww.zotero.org\u002F) + [zotero-mcp](https:\u002F\u002Fgithub.com\u002FGalaxy-Dawn\u002Fzotero-mcp)\n    *   **知识库管理工作流**: [Obsidian](https:\u002F\u002Fobsidian.md\u002F)\n\n> **注意分支选择**：\n> *   使用 **Claude Code**：保持在 `main` 分支（默认）。\n> *   使用 **Codex CLI**：请切换至 `codex` 分支。\n> *   使用 **OpenCode**：请切换至 `opencode` 分支。\n\n## 安装步骤\n\n推荐采用**全量安装**模式，安装脚本具备备份感知和增量更新功能，能安全地合并配置。\n\n### 1. 克隆仓库并运行安装脚本\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002FGalaxy-Dawn\u002Fclaude-scholar.git \u002Ftmp\u002Fclaude-scholar\nbash \u002Ftmp\u002Fclaude-scholar\u002Fscripts\u002Fsetup.sh\n```\n\n### 2. 安装说明与配置合并\n\n安装脚本会自动执行以下操作：\n*   备份被覆盖的文件至 `~\u002F.claude\u002F.claude-scholar-backups\u002F\u003Ctimestamp>\u002F`。\n*   保留您现有的 `env`、模型设置、API Key 及权限配置。\n*   若您已有 `~\u002F.claude\u002FCLAUDE.md` 或 `~\u002F.claude\u002FCLAUDE.zh-CN.md`，脚本**不会**直接覆盖它们，而是将新版安装为侧车文件（如 `CLAUDE.scholar.md`）。\n\n**重要操作**：\n安装完成后，请检查 `~\u002F.claude\u002FCLAUDE.scholar.md` (或 `.zh-CN.scholar.md`)，将其中的核心配置段落**手动合并**到您现有的 `CLAUDE.md` 文件中，以确保所有技能（Skills）和钩子（Hooks）生效。\n\n### 3. 更新方法\n\n后续如需更新，只需运行：\n\n```bash\ncd \u002Ftmp\u002Fclaude-scholar\ngit pull --ff-only\nbash scripts\u002Fsetup.sh\n```\n\n*(注：若仅需最小化安装，可参考原文手动复制 `hooks` 和特定 `skills` 目录，但需手动配置 `settings.json`)*\n\n## 基本使用\n\n安装完成后，Claude Scholar 将通过自然语言指令或特定命令辅助您的研究工作。以下是几个典型的使用场景：\n\n### 场景一：研究构思与文献梳理\n在对话中直接描述您的研究兴趣，助手将帮助细化问题并规划文献检索策略。\n```text\n我想研究大语言模型在低资源场景下的微调效率，请帮我制定一个初步的研究计划，并列出关键的研究缺口。\n```\n若已配置 Zotero MCP，可使用专用命令导入文献：\n```text\n\u002Fzotero-review\n```\n\n### 场景二：实验分析与报告生成\n完成代码实验后，利用内置技能进行数据分析和报告撰写。\n```text\n请使用 results-analysis 技能分析当前的实验数据，生成统计图表，并解释结果显著性。\n```\n生成完整的实验报告并归档到 Obsidian：\n```text\n请使用 results-report 技能生成本次实验的完整报告，并写入知识库。\n```\n\n### 场景三：论文写作与润色\n在撰写论文草稿时，调用写作辅助技能。\n```text\n请帮我审查这段相关工作的描述，确保逻辑连贯且引用准确，并优化学术表达风格。\n```\n\n### 核心工作流概览\n整个工具链围绕以下闭环设计：\n1.  **Ideation**: 模糊想法 -> 具体问题与计划\n2.  **Literature**: 检索、导入 (Zotero)、阅读与笔记\n3.  **Knowledge**: 知识持久化至 Obsidian (`Papers`, `Experiments` 等目录)\n4.  **Experiments**: 假设追踪、运行记录与发现总结\n5.  **Analysis**: 严格统计分析与科学绘图\n6.  **Writing**: 论文撰写、回复审稿意见及幻灯片制作","一位计算机科学的博士生正试图复现一篇最新的顶会论文，并在此基础上改进算法以撰写自己的研究文章。\n\n### 没有 claude-scholar 时\n- **文献管理混乱**：手动从 PDF 提取代码片段和公式极易出错，Zotero 中的文献元数据与本地代码库完全割裂，查找特定实现细节如同大海捞针。\n- **环境配置耗时**：复现论文所需的依赖复杂，开发者需花费数天时间反复调试虚拟环境、解决版本冲突，严重挤占核心算法研究时间。\n- **实验记录碎片化**：实验参数、运行日志和初步结论散落在不同的终端窗口和临时笔记中，难以形成系统的知识脉络，写作时需重新梳理。\n- **多工具切换繁琐**：需要在文献阅读器、IDE、终端和写作软件间频繁切换，上下文不断中断，导致思维连贯性受损，研发效率低下。\n\n### 使用 claude-scholar 后\n- **智能文献导入**：通过集成的 Zotero 工作流，一键将论文元数据和 PDF 内容同步至项目知识库，自动关联相关代码片段，实现“所读即所得”。\n- **自动化环境搭建**：利用内置的 Codex CLI 或 OpenCode 技能，自动分析论文依赖并生成配置文件，分钟级完成沙箱环境构建，让研究者即刻聚焦算法本身。\n- **全链路实验追踪**：在 ideation 到实验的全过程中，自动记录参数调整与运行结果，并结构化存储至 Obsidian 或本地 Markdown，为论文写作提供现成的素材库。\n- **无缝工作流整合**：统一调度文献综述、代码编写、实验运行及报告生成环节，研究者只需关注核心逻辑，claude-scholar 自动处理跨工具的上下文衔接。\n\nclaude-scholar 将原本割裂的科研环节编织成自动化闭环，让研究者从繁琐的工程杂务中解放，真正回归创新本质。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FGalaxy-Dawn_claude-scholar_dc324746.png","Galaxy-Dawn","Gaorui Zhang","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002FGalaxy-Dawn_af978a70.jpg","Hello world.","Zhejiang University",null,"gaoruizhang@zju.edu.cn","https:\u002F\u002Fgithub.com\u002FGalaxy-Dawn",[84,88,92,96],{"name":85,"color":86,"percentage":87},"Python","#3572A5",63.7,{"name":89,"color":90,"percentage":91},"JavaScript","#f1e05a",19.5,{"name":93,"color":94,"percentage":95},"Shell","#89e051",16.5,{"name":97,"color":98,"percentage":99},"Makefile","#427819",0.3,3099,263,"2026-04-04T10:45:49","Linux, macOS, Windows","未说明",{"notes":106,"python":107,"dependencies":108},"该工具是基于 Claude Code、Codex CLI 或 OpenCode 的半自动化研究助手工作流，而非本地运行的深度学习模型，因此无特定 GPU 或显存需求。Windows 用户需使用 Git Bash 或 WSL 运行安装脚本。核心依赖是 Anthropic 的 Claude Code 客户端。可选依赖包括用于文献管理的 Zotero 及其 MCP 插件，以及用于知识库管理的 Obsidian。","未说明 (可选，配合 uv 使用)",[109,110,111,112,113],"Claude Code","Git","uv (可选)","Zotero + zotero-mcp (可选)","Obsidian (可选)",[45,46,15],[116,117,118,119,120,121,122,123,124,125,126],"academic-research","ai-agents","claude","claude-code","developer-tools","mcp","paper-writing","zotero","codex-cli","openai-codex","opencode","2026-03-27T02:49:30.150509","2026-04-06T05:44:22.828276",[],[131],{"id":132,"version":133,"summary_zh":134,"released_at":135},81897,"v1.0.0","# Claude Scholar v1.0.0\n\n一套面向学术研究与软件开发的全面 Claude 代码配置——覆盖从选题构思到论文发表的完整科研生命周期。\n\n## 亮点\n\n- **32项技能**，涵盖科研、写作、开发、插件开发及网页设计\n- **50+条命令**，用于科研工作流和开发自动化\n- **14个专业化智能体**，用于文献综述、代码审查、调试等任务\n- **5个跨平台钩子**（Node.js），实现工作流的自动化执行\n- **4项项目规范**，分别针对代码风格、安全性、智能体编排及实验可重复性\n- **Zotero MCP 集成**，支持文献管理、全文阅读与引用导出\n\n## 科研生命周期（7个阶段）\n\n1. **研究选题构思** — 5W1H头脑风暴、Zotero集成的文献检索、研究空白分析\n2. **机器学习项目开发** — 工厂\u002F注册表模式、配置驱动模型、Git工作流\n3. **实验结果分析** — 统计检验、可视化、消融实验\n4. **论文撰写** — 支持 NeurIPS\u002FICML\u002FICLR\u002FACL\u002FAAAI\u002FCOLM 等会议，以及 Nature\u002FScience\u002FCell\u002FPNAS 等期刊\n5. **论文自审** — 6项质量检查清单\n6. **投稿与回复审稿意见** — 带语气控制的系统性回复流程\n7. **论文接收后** — 学术报告、海报展示、成果推广内容制作\n\n## 更改记录\n\n### 2026年2月23日\n- 新增 `setup.sh` 安装脚本，支持安全合并、自动备份及智能钩子\u002FMCP服务器\u002F插件的合并\n\n### 2026年2月21日\n- 通过 `opencode` 分支支持 OpenCode 开源社区\n\n### 2026年2月20日\n- 双语配置（英文 + 中文）\n\n### 2026年2月15日\n- Zotero MCP 集成：新增 `\u002Fzotero-review`、`\u002Fzotero-notes` 命令及 `literature-reviewer` 智能体\n\n### 2026年2月14日\n- 钩子优化：两层安全机制、按6大类划分技能、30天日志自动清理\n\n### 2026年2月11日\n- 重大更新：新增10项技能、7个智能体、8条命令、2项规范（共修改89个文件）\n\n### 2026年1月26日\n- 跨平台 Node.js 钩子，扩展机器学习论文写作知识库\n\n### 2026年1月25日\n- 初始发布：包含25项技能、2个智能体、30+条命令、5个钩子及2项规范\n\n## 安装步骤\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002FGalaxy-Dawn\u002Fclaude-scholar.git \u002Ftmp\u002Fclaude-scholar\nbash \u002Ftmp\u002Fclaude-scholar\u002Fscripts\u002Fsetup.sh\n```\n\n更多详细安装选项，请参阅 [README](https:\u002F\u002Fgithub.com\u002FGalaxy-Dawn\u002Fclaude-scholar#readme)。","2026-02-25T03:15:09"]