[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-carlrannaberg--claudekit":3,"tool-carlrannaberg--claudekit":64},[4,17,27,35,43,56],{"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",140436,2,"2026-04-05T23:32:43",[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},3704,"NextChat","ChatGPTNextWeb\u002FNextChat","NextChat 是一款轻量且极速的 AI 助手，旨在为用户提供流畅、跨平台的大模型交互体验。它完美解决了用户在多设备间切换时难以保持对话连续性，以及面对众多 AI 模型不知如何统一管理的痛点。无论是日常办公、学习辅助还是创意激发，NextChat 都能让用户随时随地通过网页、iOS、Android、Windows、MacOS 或 Linux 端无缝接入智能服务。\n\n这款工具非常适合普通用户、学生、职场人士以及需要私有化部署的企业团队使用。对于开发者而言，它也提供了便捷的自托管方案，支持一键部署到 Vercel 或 Zeabur 等平台。\n\nNextChat 的核心亮点在于其广泛的模型兼容性，原生支持 Claude、DeepSeek、GPT-4 及 Gemini Pro 等主流大模型，让用户在一个界面即可自由切换不同 AI 能力。此外，它还率先支持 MCP（Model Context Protocol）协议，增强了上下文处理能力。针对企业用户，NextChat 提供专业版解决方案，具备品牌定制、细粒度权限控制、内部知识库整合及安全审计等功能，满足公司对数据隐私和个性化管理的高标准要求。",87618,"2026-04-05T07:20:52",[13,26],{"id":44,"name":45,"github_repo":46,"description_zh":47,"stars":48,"difficulty_score":23,"last_commit_at":49,"category_tags":50,"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",[14,51,52,53,15,54,26,13,55],"数据工具","视频","插件","其他","音频",{"id":57,"name":58,"github_repo":59,"description_zh":60,"stars":61,"difficulty_score":10,"last_commit_at":62,"category_tags":63,"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",[15,14,13,26,54],{"id":65,"github_repo":66,"name":67,"description_en":68,"description_zh":69,"ai_summary_zh":70,"readme_en":71,"readme_zh":72,"quickstart_zh":73,"use_case_zh":74,"hero_image_url":75,"owner_login":76,"owner_name":77,"owner_avatar_url":78,"owner_bio":79,"owner_company":80,"owner_location":81,"owner_email":82,"owner_twitter":76,"owner_website":83,"owner_url":84,"languages":85,"stars":98,"forks":99,"last_commit_at":100,"license":101,"difficulty_score":23,"env_os":102,"env_gpu":103,"env_ram":103,"env_deps":104,"category_tags":109,"github_topics":80,"view_count":23,"oss_zip_url":80,"oss_zip_packed_at":80,"status":16,"created_at":110,"updated_at":111,"faqs":112,"releases":153},2261,"carlrannaberg\u002Fclaudekit","claudekit","A toolkit of custom commands, hooks, and utilities for Claude Code","claudekit 是一套专为 Claude Code 打造的增强工具包，旨在通过自定义命令、钩子和实用程序，为 AI 编程过程提供智能护栏与工作流自动化。它主要解决了开发者在使用 AI 辅助编码时常见的痛点：如类型安全缺失难以即时发现、高风险重构失败后恢复困难、测试破坏滞后察觉，以及代码审查流于表面等问题。\n\n借助 claudekit，开发者能在 AI 生成代码的瞬间拦截潜在错误（例如禁止随意使用 'any' 类型），并通过一键命令恢复项目快照，大幅降低试错成本。其核心亮点在于引入了“多代理并行协作”机制，例如执行代码审查时，可同时调动六个专注于架构、安全、性能等不同领域的专家代理进行深度分析；此外，它还具备自动构建项目地图的能力，让 AI 无需反复搜索即可精准定位文件与依赖关系，显著提升上下文理解效率。\n\n这款工具特别适合希望将 Claude Code 深度集成到日常开发流程中的软件工程师和技术团队。无论是需要严格质量管控的企业级项目，还是追求高效迭代的个人开发者，claudekit 都能通过标准化的六阶段规范实施流程和实时错误预防机制，帮助你将 AI 编程从“实验性尝试”转变为“","claudekit 是一套专为 Claude Code 打造的增强工具包，旨在通过自定义命令、钩子和实用程序，为 AI 编程过程提供智能护栏与工作流自动化。它主要解决了开发者在使用 AI 辅助编码时常见的痛点：如类型安全缺失难以即时发现、高风险重构失败后恢复困难、测试破坏滞后察觉，以及代码审查流于表面等问题。\n\n借助 claudekit，开发者能在 AI 生成代码的瞬间拦截潜在错误（例如禁止随意使用 'any' 类型），并通过一键命令恢复项目快照，大幅降低试错成本。其核心亮点在于引入了“多代理并行协作”机制，例如执行代码审查时，可同时调动六个专注于架构、安全、性能等不同领域的专家代理进行深度分析；此外，它还具备自动构建项目地图的能力，让 AI 无需反复搜索即可精准定位文件与依赖关系，显著提升上下文理解效率。\n\n这款工具特别适合希望将 Claude Code 深度集成到日常开发流程中的软件工程师和技术团队。无论是需要严格质量管控的企业级项目，还是追求高效迭代的个人开发者，claudekit 都能通过标准化的六阶段规范实施流程和实时错误预防机制，帮助你将 AI 编程从“实验性尝试”转变为“可靠的生产力工具”，让代码交付更安全、更规范。","![claudekit banner](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fcarlrannaberg_claudekit_readme_9029c0dfd7d3.png)\n\n# claudekit\n\n> Smart guardrails and workflow automation for Claude Code - catch errors in real-time, save checkpoints, and enhance AI coding with expert subagents\n\n[![npm version](https:\u002F\u002Fimg.shields.io\u002Fnpm\u002Fv\u002Fclaudekit.svg)](https:\u002F\u002Fwww.npmjs.com\u002Fpackage\u002Fclaudekit)\n[![npm downloads](https:\u002F\u002Fimg.shields.io\u002Fnpm\u002Fdt\u002Fclaudekit.svg)](https:\u002F\u002Fwww.npmjs.com\u002Fpackage\u002Fclaudekit)\n[![License: MIT](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-MIT-yellow.svg)](https:\u002F\u002Fopensource.org\u002Flicenses\u002FMIT)\n[![Node.js Version](https:\u002F\u002Fimg.shields.io\u002Fnode\u002Fv\u002Fclaudekit.svg)](https:\u002F\u002Fnodejs.org)\n[![Mentioned in Awesome Claude Code](https:\u002F\u002Fawesome.re\u002Fmentioned-badge-flat.svg)](https:\u002F\u002Fgithub.com\u002Fhesreallyhim\u002Fawesome-claude-code)\n[![Buy Me a Coffee](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F☕-Buy%20me%20a%20coffee-ffdd00?style=flat&logo=buy-me-a-coffee)](https:\u002F\u002Fbuymeacoffee.com\u002Fcarlrannaberg)\n\n## 🚀 Installation\n\n> **⚠️ Requires:** Claude Code **Max plan** (for optimal token usage) • Node.js 20+\n\n```bash\nnpm install -g claudekit\n# or: yarn global add claudekit\n# or: pnpm add -g claudekit\n```\n\n## ⚡ Quick Start\n\n```bash\n# Initialize in your project\nclaudekit setup\n\n# In Claude Code, try these commands:\n\u002Fgit:status                      # Groups changes by type & suggests commit strategy\n\u002Fspec:create \"your next feature\" # Researches codebase & writes full spec\n\u002Fcode-review                     # 6 specialized agents analyze code in parallel\n```\n\n## 🎯 What It Does\n\nClaudekit acts as your safety net while coding with Claude:\n\n```\nBefore: Claude adds 'any' type → ❌ Lost type safety discovered in code review\nAfter:  Claude adds 'any' type → ✅ Instant block: \"Use specific type: User | null\"\n\nBefore: Risky refactor fails → ❌ Git archaeology to find working version  \nAfter:  Risky refactor fails → ✅ One command: \u002Fcheckpoint:restore\n\nBefore: Claude breaks tests → ❌ You discover it after Claude finishes\nAfter:  Claude breaks tests → ✅ Claude sees error immediately and fixes it\n\nBefore: Ask Claude to review → ❌ Shallow analysis, general feedback, sequential execution\nAfter:  Use \u002Fcode-review → ✅ 6 specialized agents analyze in parallel, dynamically pull technology-specific expertise\n```\n\n## Key Features\n\n### 🗺️ Instant Codebase Navigation\n**Claude sees your entire project structure automatically - skips most discovery searches.**\n- **Navigate directly to code** - Jump straight to files, functions, and classes\n- **No trial-and-error** - Avoid sifting through test files, docs, and irrelevant matches  \n- **Confident code access** - Claude knows exactly what exists and where\n- **See relationships** - Claude understands dependencies and architecture instantly\n- **Automatic setup** - Runs invisibly on first prompt, updates as you code\n\n[Codebase map guide →](docs\u002Fguides\u002Fcodebase-map.md)\n\n### 🔍 Comprehensive Code Review\n- **Multi-aspect analysis**: 6 parallel agents for architecture, security, performance, testing, quality, and documentation\n- **Smart targeting**: Automatically selects relevant review agents based on file types and scope\n- **Structured reports**: Prioritized issues (Critical\u002FHigh\u002FMedium) with type icons and actionable solutions\n- **Quality metrics**: Scoring system and issue distribution tracking for measurable improvements\n\n\n### 🚀 Iterative Spec Implementation\n- **6-phase workflow**: Implementation → Test Writing → Code Review → Iterative Improvement → Commit → Progress Tracking\n- **Dynamic agent selection**: Real-time detection of available specialized agents with automatic fallbacks\n- **Quality gates**: Each phase includes validation before proceeding to ensure high-quality deliverables\n- **Atomic commits**: Standardized commit messages with comprehensive change documentation\n\n[Spec workflow guide →](docs\u002Fguides\u002Fspec-workflow.md)\n\n### 🛡️ Real-time Error Prevention\n- **🔒 Advanced Bash Security Analysis**: Comprehensive protection against sensitive file access via shell commands\n  - **Intelligent Command Parsing**: Analyzes complex bash pipelines, variable assignments, and shell constructs\n  - **Pattern Detection**: Identifies sensitive file access via `xargs`, `find`, pipes, and command substitution\n  - **Path Extraction**: Safely processes globs, variables, and @file references in curl\u002Fhttpie uploads\n  - **Security Heuristics**: Detects risky patterns like `find -name '*.env' | xargs cat` before execution\n- **🔒 Enhanced Sensitive File Protection**: Comprehensive security with 195+ patterns across 12 categories including cloud credentials, cryptocurrencies, databases, tokens, and production data\n- **Multi-category Protection**: Environment files, SSH keys, cloud provider credentials (AWS\u002FAzure\u002FGCP), package manager auth, crypto wallets, production databases\n- **TypeScript Guard**: Blocks `any` types and type errors as Claude edits\n- **Linting**: Catches style issues immediately  \n- **Anti-patterns**: Prevents code replacement with comments\n- **Test Runner**: Runs relevant tests on file changes\n- **Multi-tool ignore file support**: Respects `.agentignore`, `.aiignore`, `.aiexclude`, `.geminiignore`, `.codeiumignore`, `.cursorignore`\n\n### 💾 Git Checkpoint System\n- **Auto-save**: Creates checkpoints when Claude stops\n- **Easy restore**: `\u002Fcheckpoint:restore` to undo changes\n- **Checkpoint management**: List, restore, or clean up checkpoints\n\n[Checkpoint guide →](docs\u002Fguides\u002Fcheckpoint.md)\n\n### 🤖 AI Subagents\n- **Research Expert**: `research-expert` for comprehensive parallel research with 90% time reduction, automatic source analysis, and structured reporting\n- **Code Search**: `code-search` for fast, parallel codebase searches with 3-10x speed improvements\n- **Triage Expert**: `triage-expert` for initial problem diagnosis and routing to specialists\n- **Domain experts**: TypeScript, React, database, testing, infrastructure, documentation specialists\n- **Code review**: Dedicated `code-review-expert` agent with 6 focused analysis aspects\n- **Framework specialists**: Next.js, Nest.js, AI SDK, and build tool experts\n- **Performance optimization**: Intelligent hook management with `disableHooks` configuration to skip unnecessary validations during specialized operations\n- **Mandatory delegation**: All technical issues MUST be handled by specialized subagents\n- **Proactive help**: Run `\u002Fagents-md:init` to update AGENTS.md\u002FCLAUDE.md for automatic subagent delegation\n\n### 📊 Hook Performance Profiling\n- **Performance monitoring**: Analyze hook execution time and output size with `claudekit-hooks profile`\n- **Batch analysis**: Automatically profiles all configured hooks from `.claude\u002Fsettings.json`\n- **Statistical analysis**: Support for multiple iterations with averaging for accurate measurements\n- **Smart warnings**: Color-coded alerts for slow hooks (>5s) and output size limits\n- **Claude Code integration**: Specialized limits for UserPromptSubmit hooks that inject context\n- **Token estimation**: Calculate estimated token usage for hook outputs\n\n[Hook profiling guide →](docs\u002Fguides\u002Fhook-profiling.md)\n\n### 🧠 Enhanced AI Reasoning\n- **Automatic enhancement**: The `thinking-level` hook invisibly enhances Claude's reasoning without manual prompt engineering\n- **4 configurable levels**: From disabled (0) to maximum reasoning (3) with smart defaults\n- **Invisible operation**: Keywords are injected silently without user awareness\n- **Lightweight**: \u003C 5ms overhead with minimal token usage\n- **Session-aware**: Activates on first user prompt to enhance AI reasoning throughout the session\n\n[Thinking level guide →](docs\u002Fguides\u002Fthinking-level.md)\n\n### 🎛️ Session-Based Hook Control\n- **Temporary control**: Disable\u002Fenable hooks for individual Claude Code sessions\n- **Session isolation**: Changes only affect your current session, not other developers\n- **Smart state management**: 4-state system (enabled\u002Fdisabled\u002Fnot-configured\u002Fnot-found)\n- **Fuzzy matching**: Partial hook names with intelligent suggestions\n- **Zero permanent changes**: No configuration files modified\n\n[Session hook control guide →](docs\u002Fguides\u002Fsession-hook-control.md)\n\n### 🔬 Multi-Agent Research System\n- **Parallel Information Gathering**: Launch 5-10 research agents simultaneously for 90% faster investigation\n- **Intelligent Query Classification**: Automatically determines research strategy (breadth-first, depth-first, or simple factual)\n- **Comprehensive Source Analysis**: Cross-references findings across multiple domains with automatic citation management\n- **Structured Research Reports**: Executive summaries, key findings, detailed analysis, and methodology documentation\n- **Filesystem Artifacts**: Subagents write full reports to `\u002Ftmp\u002Fresearch_*.md` for efficient synthesis and token optimization\n\n### 📝 Smart Commands\n- `\u002Fresearch [query]` - Deep parallel research with specialized subagents and structured reporting\n- `\u002Fcode-review` - Multi-aspect code reviews with 6 parallel agents\n- `\u002Fgit:commit` - Creates commits following your project's conventions\n- `\u002Fgit:checkout` - Smart branch creation and switching with conventional naming\n- `\u002Fgit:ignore-init` - Initialize AI-safe .gitignore patterns for sensitive file protection\n- `\u002Fvalidate-and-fix` - Runs all quality checks and fixes issues\n- `\u002Fspec:create` - Generates comprehensive specifications\n- `\u002Fspec:execute` - Implements specifications with iterative quality assurance workflow\n- `\u002Fcreate-subagent` - Build custom AI assistants\n- `\u002Fhook:disable [hook-name]` - Temporarily disable hooks for current session\n- `\u002Fhook:enable [hook-name]` - Re-enable disabled hooks for current session\n- `\u002Fhook:status [hook-name]` - Show hook status (enabled\u002Fdisabled\u002Fnot-configured)\n\n## CLI Commands\n\n```bash\nclaudekit setup              # Interactive setup wizard\nclaudekit setup --yes        # Quick setup with defaults\nclaudekit setup --all        # Install everything (all agents)\nclaudekit setup --user       # Install in user directory (~\u002F.claude) instead of project\nclaudekit list               # Show all components (hooks, commands, agents, config)\nclaudekit list agents        # List available agents with token counts\nclaudekit list commands      # List available commands with token counts\nclaudekit show agent \u003Cid>    # Display agent prompt for external use\nclaudekit show command \u003Cid>  # Display command prompt for external use\nclaudekit show agent \u003Cid> -f json    # Output agent metadata and content as JSON\nclaudekit show command \u003Cid> -f json  # Output command metadata and content as JSON\nclaudekit doctor             # Check your installation and validate documentation links\n\n# Tool status checking\nclaudekit status stm         # Check Simple Task Master (STM) installation and initialization status\n\n# Linting and validation\nclaudekit lint-commands      # Lint slash command markdown files\nclaudekit lint-agents        # Lint agent markdown files (formerly lint-subagents)\n\n# Hook management and profiling\nclaudekit-hooks run \u003Chook>   # Execute a specific hook manually\nclaudekit-hooks profile      # Profile all configured hooks performance\nclaudekit-hooks profile \u003Chook> --iterations 5  # Profile specific hook with averaging\n\n# Session-based hook control (within Claude Code sessions)\nclaudekit-hooks disable \u003Chook>  # Disable hook for current session\nclaudekit-hooks enable \u003Chook>   # Re-enable hook for current session  \nclaudekit-hooks status \u003Chook>   # Show hook status (enabled\u002Fdisabled)\n```\n\n### Using Prompts with External LLMs\n\nThe `show` command extracts agent and command prompts for use with Claude Code's non-interactive mode and other AI coding tools:\n\n```bash\n# Extract agent prompt for use with Claude Code CLI\nEXPERT=$(claudekit show agent typescript-expert)\ncat src\u002Fapp.ts | claude -p --append-system-prompt \"$EXPERT\" \"Review this code\"\n\n# Get structured metadata for integration with tools\nclaudekit show agent typescript-expert --format json | jq '.description'\n\n# Works with any AI coding CLI tool\nclaudekit show agent react-expert | amp -x \"Fix React hooks violations\"\n```\n\nSee the [external LLMs guide](docs\u002Fguides\u002Fusing-prompts-with-external-llms.md) for examples with Claude Code, Amp, Gemini CLI, and others.\n\n## Slash Commands\n\nUse these commands directly in Claude Code:\n\n### Git & Checkpoints\n- `\u002Fcheckpoint:create [msg]` - Save current state\n- `\u002Fcheckpoint:restore [n]` - Restore to checkpoint\n- `\u002Fcheckpoint:list` - View all checkpoints\n- `\u002Fgit:commit` - Smart commit with conventions\n- `\u002Fgit:status` - Intelligent git analysis\n- `\u002Fgit:checkout [branch]` - Smart branch creation and switching with conventional naming\n- `\u002Fgit:ignore-init` - Initialize AI-safe .gitignore patterns\n\n### Code Quality\n- `\u002Fcode-review [target]` - Multi-aspect code review with parallel agents\n- `\u002Fvalidate-and-fix` - Run all quality checks\n\n### Development Tools\n- `\u002Fresearch [query]` - Deep parallel research with specialized subagents\n- `\u002Fspec:create [feature]` - Generate specifications\n- `\u002Fspec:execute [file]` - Implement specifications with 6-phase iterative workflow\n- `\u002Fagents-md:init` - Configure AI assistants\n- `\u002Fcreate-subagent` - Build custom agents\n- `\u002Fcreate-command` - Create custom commands\n\n[View all commands →](docs\u002Freference\u002Fcommands.md)\n\n## Hooks\n\nHooks automatically enforce quality as Claude works:\n\n### Available Hooks\n\n**File Security Hooks** (PreToolUse - run before file access)\n- `file-guard` - Block AI access to sensitive files (`.env`, keys, credentials) using ignore file patterns\n\n**File Change Hooks** (PostToolUse - run on edit)\n- `typecheck-changed` - Run TypeScript type checking on file changes\n- `lint-changed` - Run linting validation on changed files (Biome, ESLint, etc)\n- `test-changed` - Run tests for changed files\n- `check-any-changed` - Forbid any types in changed TypeScript files\n- `check-comment-replacement` - Detect when code is replaced with comments\n- `check-unused-parameters` - Detect lazy refactoring where parameters are prefixed with _ instead of being removed\n\n**UserPromptSubmit Hooks** (run on first user prompt of each session)\n- `codebase-map` - Provide invisible codebase context once per session using codebase-map CLI\n- `thinking-level` - Inject thinking level keywords to enhance Claude's reasoning capabilities\n\n**PostToolUse Hooks** (also includes file monitoring)\n- `codebase-map-update` - Incrementally update codebase map when files change\n\n**Project-Wide Hooks** (typically for Stop\u002FSubagentStop events)\n- `typecheck-project` - TypeScript validation on entire project\n- `lint-project` - Linting validation on entire project (Biome, ESLint, etc)\n- `test-project` - Run full test suite\n- `create-checkpoint` - Git auto-checkpoint on stop\n- `check-todos` - Validate todo completions\n- `self-review` - Enhanced targeted self-review with implementation completeness detection, prioritizing mock implementations, placeholder code, and incomplete features over code quality concerns\n\n[Self-review guide →](docs\u002Fguides\u002Fself-review.md)\n\n### Hook Events\n\n- **PreToolUse** - Triggered before file access (Read, Write, Edit, MultiEdit) for permission control\n- **PostToolUse** - Triggered after file modifications (Write, Edit, MultiEdit)\n- **Stop** - Triggered when Claude Code stops or conversation ends\n- **SubagentStop** - Triggered when subagents complete their tasks\n- **UserPromptSubmit** - Triggered when users submit their first prompt in a session\n\n### Hook Commands\n\n```bash\n# See all available hooks with descriptions\nclaudekit-hooks list\n\n# Profile hook performance and output size\nclaudekit-hooks profile                   # Profile all hooks\nclaudekit-hooks profile typecheck-changed  # Profile specific hook\nclaudekit-hooks profile --iterations 5    # Multiple runs for averages\n\n# Session-based hook control (requires active Claude Code session)\nclaudekit-hooks disable typecheck-changed  # Disable specific hook for current session\nclaudekit-hooks enable typecheck          # Re-enable hook (supports partial matching)\nclaudekit-hooks status                     # Show status of all hooks\nclaudekit-hooks status lint                # Show status of specific hook\n```\n\nThe profile command helps identify:\n- Slow hooks (>5s) that delay your workflow\n- High-output hooks that consume context window\n- UserPromptSubmit hooks approaching the 10k character limit\n\nSession-based hook control allows you to temporarily disable hooks within specific Claude Code sessions without modifying configuration files. Features include:\n- Fuzzy matching for hook names (partial matches work)\n- Visual status indicators (🔒 disabled, ✅ enabled)\n- Session-isolated state (changes only affect current session)\n- Automatic session detection via transcript UUID\n\n[Hook configuration →](docs\u002Freference\u002Fhooks.md) | [Profiling guide →](docs\u002Fguides\u002Fhook-profiling.md)\n\n## Subagents\n\nSpecialized AI assistants for different domains:\n\n**Popular Agents**\n- `code-review-expert` - 6-agent parallel code review (architecture, security, performance, testing, quality, docs)\n- `triage-expert` - Context gathering and problem diagnosis routing to specialized experts\n- `documentation-expert` - Documentation structure, cohesion, flow, and information architecture specialist\n- `oracle` - Deep debugging and complex problem analysis ([requires setup](docs\u002Fintegrations\u002Foracle.md))\n- `refactoring-expert` - Code smell detection and comprehensive refactoring guidance\n- `typescript-expert` - TypeScript\u002FJavaScript development and type system\n- `react-expert` - React components, hooks, and performance\n- `testing-expert` - Test strategy, coverage, and best practices\n- `database-expert` - Query optimization and schema design\n- `ai-sdk-expert` - Vercel AI SDK v5 streaming and model integration\n- `nestjs-expert` - Nest.js architecture and dependency injection\n- `kafka-expert` - Apache Kafka producer\u002Fconsumer configuration, topic management, cluster operations, and performance tuning\n- `loopback-expert` - LoopBack 4 framework development, model relationships, authentication, and dependency injection\n\n**Usage**\n```bash\n# List all available agents with token counts\nclaudekit list agents\n\n# Install specific agents\nclaudekit setup --agents typescript-expert,react-expert\n```\n\nOnce installed, agents work automatically or can be invoked directly:\n```\n\"Use the oracle agent to debug this issue\"\n\"Have the typescript-expert review this type definition\"\n```\n\n[View all agents →](docs\u002Freference\u002Fsubagents.md)\n\n## Security\n\n### Protecting Sensitive Files\n\nThe `file-guard` hook automatically protects sensitive files from AI access. Quick setup:\n\n```bash\n# New projects: Install claudekit with file-guard protection\nnpm install -g claudekit && claudekit setup --yes --force --hooks file-guard\n\n# Existing projects: Add file-guard to current setup\nclaudekit setup --yes --force --hooks file-guard\n```\n\nOr manually create a `.agentignore` file:\n\n```gitignore\n# .agentignore\n.env\n.env.*\n*.key\n*.pem\n.aws\u002Fcredentials\n.ssh\u002F*\n\n# Allow example files\n!.env.example\n```\n\nThe `file-guard` hook supports multiple ignore file formats (`.agentignore`, `.aiignore`, `.cursorignore`, `.geminiignore`, `.codeiumignore`, `.aiexclude`) and merges patterns from all of them - a unique feature that provides comprehensive protection regardless of which AI tools your team uses.\n\n**Key features:**\n- **Advanced Bash Command Analysis**: Parses complex shell commands to detect sensitive file access patterns\n- **Pipeline Security**: Detects risky constructs like `find -name '*.env' | xargs cat` and `echo '.env' | xargs cat`\n- **Variable Expansion**: Safely handles bash variables, command substitution, and parameter expansion\n- **Upload Protection**: Blocks file uploads via curl, httpie, scp, rsync, and cloud CLI tools (aws, gcloud, az)\n- Merges patterns from ALL ignore files (unlike other tools)\n- Supports gitignore-style syntax including negation (!)\n- Resolves symlinks to prevent bypasses\n- Blocks path traversal attempts outside project\n\nSee the [File Guard Guide](docs\u002Fguides\u002Ffile-guard.md) for detailed configuration and best practices.\n\n## Configuration\n\nClaudekit uses two configuration files:\n\n**`.claude\u002Fsettings.json`** - Project settings (hooks, commands)\n```json\n{\n  \"hooks\": {\n    \"PreToolUse\": [{\n      \"matcher\": \"Read|Edit|MultiEdit|Write\",\n      \"hooks\": [{\"type\": \"command\", \"command\": \"claudekit-hooks run file-guard\"}]\n    }],\n    \"PostToolUse\": [{\n      \"matcher\": \"Write|Edit|MultiEdit\",\n      \"hooks\": [{\"type\": \"command\", \"command\": \"claudekit-hooks run typecheck-changed\"}]\n    }],\n    \"Stop\": [{\n      \"matcher\": \"*\",\n      \"hooks\": [{\"type\": \"command\", \"command\": \"claudekit-hooks run create-checkpoint\"}]\n    }],\n    \"SubagentStop\": [{\n      \"matcher\": \"*\",\n      \"hooks\": [{\"type\": \"command\", \"command\": \"claudekit-hooks run create-checkpoint\"}]\n    }],\n    \"UserPromptSubmit\": [{\n      \"matcher\": \"*\",\n      \"hooks\": [{\"type\": \"command\", \"command\": \"claudekit-hooks run codebase-map\"}]\n    }]\n  }\n}\n```\n\n**`.claudekit\u002Fconfig.json`** - Hook configuration\n```json\n{\n  \"hooks\": {\n    \"typecheck-changed\": {\n      \"command\": \"npm run typecheck\"\n    },\n    \"thinking-level\": {\n      \"level\": 2\n    },\n    \"codebase-map\": {\n      \"include\": [\"src\u002F**\", \"lib\u002F**\"],\n      \"exclude\": [\"**\u002F*.test.ts\", \"**\u002F*.spec.ts\"],\n      \"format\": \"dsl\"\n    }\n  }\n}\n```\n\n[Configuration guide →](docs\u002Fgetting-started\u002Fconfiguration.md)\n\n## Getting Help\n\n- **Documentation**: [Full docs](docs\u002F)\n- **Issues**: [GitHub Issues](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fissues)\n- **Quick test**: `claudekit doctor` to check setup\n\n## Common Issues\n\n**Hooks not triggering?**\n```bash\nclaudekit doctor                # Check configuration\nnpm list -g claudekit           # Verify installation\n```\n\n**Tests timing out?**\nUse the ultra-fast test configuration for hook execution:\n```json\n{\n  \"hooks\": {\n    \"test-project\": {\n      \"command\": \"npm run test:fast\",\n      \"timeout\": 30000\n    }\n  }\n}\n```\n\nOr configure a custom faster test command:\n```json\n{\n  \"hooks\": {\n    \"test-project\": {\n      \"command\": \"npm run test:unit\",\n      \"timeout\": 50000\n    }\n  }\n}\n```\n\n**Codebase map not generating?**\nInstall the codebase-map CLI tool:\n```bash\nnpm install -g codebase-map\n# Check installation:\ncodebase-map --version\n```\n\n[Full troubleshooting →](docs\u002Fgetting-started\u002Ftroubleshooting.md)\n\n## MCP Integration\n\nSome commands support optional MCP server integration:\n\n### Context7 (for documentation access)\n```bash\n# Install globally\nnpm install -g @upstash\u002Fcontext7-mcp\n\n# Add to Claude Code\nclaude mcp add context7 context7-mcp\n```\n\nEnables `\u002Fspec:create` to fetch up-to-date library documentation.\n\n## Development\n\n```bash\n# Clone and install\ngit clone https:\u002F\u002Fgithub.com\u002Fclaudekit\u002Fclaudekit.git\ncd claudekit\nnpm install\n\n# Build\nnpm run build\n\n# Test\nnpm test\n\n# Validate production build\nnpm run validate:build         # Check dependencies and build artifacts\nnpm run validate:install       # Test clean installation\n\n# Create symlinks for development\nnpm run symlinks\n```\n\n## Contributing\n\nWe welcome contributions! See [CONTRIBUTING.md](CONTRIBUTING.md) for guidelines.\n\n## License\n\nMIT © [claudekit contributors](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fgraphs\u002Fcontributors)","![claudekit 横幅](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fcarlrannaberg_claudekit_readme_9029c0dfd7d3.png)\n\n# claudekit\n\n> 为 Claude Code 提供智能护栏和工作流自动化——实时捕获错误、保存检查点，并通过专家子代理增强 AI 编码能力\n\n[![npm 版本](https:\u002F\u002Fimg.shields.io\u002Fnpm\u002Fv\u002Fclaudekit.svg)](https:\u002F\u002Fwww.npmjs.com\u002Fpackage\u002Fclaudekit)\n[![npm 下载量](https:\u002F\u002Fimg.shields.io\u002Fnpm\u002Fdt\u002Fclaudekit.svg)](https:\u002F\u002Fwww.npmjs.com\u002Fpackage\u002Fclaudekit)\n[![许可证：MIT](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-MIT-yellow.svg)](https:\u002F\u002Fopensource.org\u002Flicenses\u002FMIT)\n[![Node.js 版本](https:\u002F\u002Fimg.shields.io\u002Fnode\u002Fv\u002Fclaudekit.svg)](https:\u002F\u002Fnodejs.org)\n[![在 Awesome Claude Code 中提及](https:\u002F\u002Fawesome.re\u002Fmentioned-badge-flat.svg)](https:\u002F\u002Fgithub.com\u002Fhesreallyhim\u002Fawesome-claude-code)\n[![请我喝杯咖啡](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F☕-请%20我%20喝%20杯%20咖啡-ffdd00?style=flat&logo=buy-me-a-coffee)](https:\u002F\u002Fbuymeacoffee.com\u002Fcarlrannaberg)\n\n## 🚀 安装\n\n> **⚠️ 需要：** Claude Code **Max 方案**（以实现最佳 token 使用）• Node.js 20+\n\n```bash\nnpm install -g claudekit\n# 或：yarn global add claudekit\n# 或：pnpm add -g claudekit\n```\n\n## ⚡ 快速入门\n\n```bash\n# 在您的项目中初始化\nclaudekit setup\n\n# 在 Claude Code 中，尝试以下命令：\n\u002Fgit:status                      # 按类型分组更改并建议提交策略\n\u002Fspec:create \"您的下一个功能\" # 研究代码库并编写完整规范\n\u002Fcode-review                     # 6 个专业代理并行分析代码\n```\n\n## 🎯 它的作用\n\nClaudekit 在您使用 Claude 进行编码时充当您的安全网：\n\n```\n之前：Claude 添加 'any' 类型 → ❌ 在代码审查中发现丢失了类型安全性  \n之后：Claude 添加 'any' 类型 → ✅ 即时阻止：“请使用具体类型：User | null”\n\n之前：危险的重构失败 → ❌ 需要进行 Git 考古来寻找可用版本  \n之后：危险的重构失败 → ✅ 一条命令：\u002Fcheckpoint:restore\n\n之前：Claude 导致测试失败 → ❌ 您是在 Claude 完成后才发现的  \n之后：Claude 导致测试失败 → ✅ Claude 立即看到错误并修复它\n\n之前：让 Claude 进行代码审查 → ❌ 浅层分析、通用反馈、顺序执行  \n之后：使用 \u002Fcode-review → ✅ 6 个专业代理并行分析，动态调用特定于技术领域的专业知识\n```\n\n## 核心功能\n\n### 🗺️ 即时代码库导航\n**Claude 会自动查看您的整个项目结构——省去了大部分查找步骤。**\n- **直接跳转到代码**：直接定位到文件、函数和类\n- **无需反复试错**：避免翻阅测试文件、文档和其他无关内容\n- **自信地访问代码**：Claude 清楚地知道哪些内容存在以及它们的位置\n- **查看关系**：Claude 瞬间理解依赖关系和架构\n- **自动设置**：首次提示时隐形运行，并随您的编码更新\n\n[代码库地图指南 →](docs\u002Fguides\u002Fcodebase-map.md)\n\n### 🔍 全面代码审查\n- **多方面分析**：6 个并行代理分别负责架构、安全、性能、测试、质量和文档\n- **智能定位**：根据文件类型和范围自动选择相关审查代理\n- **结构化报告**：按优先级列出问题（严重\u002F高\u002F中），附带类型图标和可操作的解决方案\n- **质量指标**：评分系统及问题分布跟踪，便于衡量改进效果\n\n### 🚀 迭代式规范实施\n- **6 阶段工作流程**：实现 → 编写测试 → 代码审查 → 迭代改进 → 提交 → 进度跟踪\n- **动态代理选择**：实时检测可用的专业代理，并自动回退至备用方案\n- **质量关卡**：每个阶段都会进行验证，确保高质量交付\n- **原子提交**：标准化提交信息，包含全面的变更说明\n\n[规范工作流程指南 →](docs\u002Fguides\u002Fspec-workflow.md)\n\n### 🛡️ 实时错误预防\n- **🔒 高级 Bash 安全分析**：全面防护敏感文件通过 Shell 命令被访问\n  - **智能命令解析**：分析复杂的 Bash 管道、变量赋值和 Shell 结构\n  - **模式检测**：识别通过 `xargs`、`find`、管道和命令替换等方式访问敏感文件的行为\n  - **路径提取**：安全处理 `curl` 和 `httpie` 上传中的通配符、变量和 `@file` 引用\n  - **安全启发式**：在执行前检测到类似 `find -name '*.env' | xargs cat` 的危险模式\n- **🔒 增强的敏感文件保护**：涵盖云凭证、加密货币、数据库、令牌和生产数据等 12 个类别，共 195+ 种模式\n- **多类别保护**：环境文件、SSH 密钥、云服务商凭证（AWS\u002FAzure\u002FGCP）、包管理器认证、加密钱包、生产数据库\n- **TypeScript 防护**：在 Claude 编辑时阻止 `any` 类型和类型错误\n- **代码风格检查**：立即捕捉样式问题\n- **反模式检测**：防止代码被注释覆盖\n- **测试运行器**：在文件更改时运行相关测试\n- **多工具忽略文件支持**：尊重 `.agentignore`、`.aiignore`、`.aiexclude`、`.geminiignore`、`.codeiumignore` 和 `.cursorignore`\n\n### 💾 Git 检查点系统\n- **自动保存**：当 Claude 停止时创建检查点\n- **轻松恢复**：使用 `\u002Fcheckpoint:restore` 撤销更改\n- **检查点管理**：列出、恢复或清理检查点\n\n[检查点指南 →](docs\u002Fguides\u002Fcheckpoint.md)\n\n### 🤖 AI 子代理\n- **研究专家**：`research-expert` 可进行综合并行研究，时间缩短 90%，自动分析来源并生成结构化报告\n- **代码搜索**：`code-search` 可快速并行搜索代码库，速度提升 3–10 倍\n- **分诊专家**：`triage-expert` 用于初步诊断问题并将其分配给专业人员\n- **领域专家**：TypeScript、React、数据库、测试、基础设施和文档方面的专家\n- **代码审查**：专门的 `code-review-expert` 代理，从 6 个方面进行专注分析\n- **框架专家**：Next.js、Nest.js、AI SDK 和构建工具方面的专家\n- **性能优化**：通过 `disableHooks` 配置智能管理钩子，在专业操作期间跳过不必要的验证\n- **强制委托**：所有技术问题必须由专业子代理处理\n- **主动帮助**：运行 `\u002Fagents-md:init` 更新 AGENTS.md\u002FCLAUDE.md，以实现子代理的自动委派\n\n### 📊 挂钩性能剖析\n- **性能监控**：使用 `claudekit-hooks profile` 分析挂钩执行时间和输出大小\n- **批量分析**：自动剖析 `.claude\u002Fsettings.json` 中配置的所有挂钩\n- **统计分析**：支持多次迭代并取平均值以获得准确的测量结果\n- **智能警告**：对耗时超过 5 秒的慢速挂钩和超出输出大小限制的情况进行颜色编码提醒\n- **Claude Code 集成**：为注入上下文的 UserPromptSubmit 挂钩设置专用限制\n- **Token 估算**：计算挂钩输出的预估 Token 使用量\n\n[挂钩剖析指南 →](docs\u002Fguides\u002Fhook-profiling.md)\n\n### 🧠 增强型 AI 推理\n- **自动增强**：`thinking-level` 挂钩可在无需手动提示工程的情况下，隐形地提升 Claude 的推理能力\n- **4 种可配置级别**：从禁用（0）到最大推理能力（3），并提供智能默认值\n- **隐形运行**：关键词会在用户不知情的情况下悄然注入\n- **轻量级**：开销小于 5 毫秒，且 Token 使用量极低\n- **会话感知**：在首次用户提示时激活，从而在整个会话中持续增强 AI 推理能力\n\n[思考层级指南 →](docs\u002Fguides\u002Fthinking-level.md)\n\n### 🎛️ 基于会话的挂钩控制\n- **临时控制**：可为单个 Claude Code 会话禁用或启用挂钩\n- **会话隔离**：更改仅影响当前会话，不会波及其他开发者\n- **智能状态管理**：采用 4 状态系统（已启用\u002F已禁用\u002F未配置\u002F未找到）\n- **模糊匹配**：支持部分挂钩名称，并提供智能建议\n- **无永久性更改**：不会修改任何配置文件\n\n[会话挂钩控制指南 →](docs\u002Fguides\u002Fsession-hook-control.md)\n\n### 🔬 多智能体研究系统\n- **并行信息收集**：同时启动 5–10 个研究智能体，使调查速度提升 90%\n- **智能查询分类**：自动确定研究策略（广度优先、深度优先或简单事实查询）\n- **全面的来源分析**：跨多个领域交叉核对研究结果，并自动管理引用\n- **结构化研究报告**：包括执行摘要、关键发现、详细分析及方法论文档\n- **文件系统产物**：子智能体将完整报告写入 `\u002Ftmp\u002Fresearch_*.md`，以便高效整合和优化 Token 使用\n\n### 📝 智能命令\n- `\u002Fresearch [query]` - 使用专业子智能体进行深度并行研究，并生成结构化报告\n- `\u002Fcode-review` - 使用 6 个并行智能体进行多维度代码审查\n- `\u002Fgit:commit` - 按照项目规范创建提交\n- `\u002Fgit:checkout` - 智能创建分支并切换，命名符合常规\n- `\u002Fgit:ignore-init` - 初始化 AI 安全的 .gitignore 模板，保护敏感文件\n- `\u002Fvalidate-and-fix` - 运行所有质量检查并修复问题\n- `\u002Fspec:create` - 生成全面的规格说明书\n- `\u002Fspec:execute` - 按照迭代式质量保证流程实施规格说明书\n- `\u002Fcreate-subagent` - 构建自定义 AI 助手\n- `\u002Fhook:disable [hook-name]` - 临时禁用当前会话中的挂钩\n- `\u002Fhook:enable [hook-name]` - 重新启用当前会话中已禁用的挂钩\n- `\u002Fhook:status [hook-name]` - 显示挂钩状态（已启用\u002F已禁用\u002F未配置）\n\n## CLI 命令\n\n```bash\nclaudekit setup              # 交互式设置向导\nclaudekit setup --yes        # 使用默认设置快速安装\nclaudekit setup --all        # 安装所有组件（所有智能体）\nclaudekit setup --user       # 安装到用户目录 (~\u002F.claude)，而非项目目录\nclaudekit list               # 显示所有组件（挂钩、命令、智能体、配置）\nclaudekit list agents        # 列出可用智能体及其 Token 数量\nclaudekit list commands      # 列出可用命令及其 Token 数量\nclaudekit show agent \u003Cid>    # 展示智能体提示，供外部使用\nclaudekit show command \u003Cid>  # 展示命令提示，供外部使用\nclaudekit show agent \u003Cid> -f json    # 将智能体元数据和内容以 JSON 格式输出\nclaudekit show command \u003Cid> -f json  # 将命令元数据和内容以 JSON 格式输出\nclaudekit doctor             # 检查您的安装情况，并验证文档链接是否有效\n\n# 工具状态检查\nclaudekit status stm         # 检查 Simple Task Master (STM) 的安装和初始化状态\n\n# 代码检查与验证\nclaudekit lint-commands      # 检查斜杠命令的 Markdown 文件\nclaudekit lint-agents        # 检查智能体 Markdown 文件（原名 lint-subagents）\n\n# 挂钩管理和剖析\nclaudekit-hooks run \u003Chook>   # 手动执行特定挂钩\nclaudekit-hooks profile      # 剖析所有已配置挂钩的性能\nclaudekit-hooks profile \u003Chook> --iterations 5  # 对特定挂钩进行剖析，并取平均值\n\n# 基于会话的挂钩控制（在 Claude Code 会话中）\nclaudekit-hooks disable \u003Chook>  # 禁用当前会话中的挂钩\nclaudekit-hooks enable \u003Chook>   # 重新启用当前会话中的挂钩  \nclaudekit-hooks status \u003Chook>   # 显示挂钩状态（已启用\u002F已禁用）\n```\n\n### 与外部 LLM 结合使用提示\n`show` 命令可提取智能体和命令的提示，用于 Claude Code 的非交互模式以及其他 AI 编程工具：\n\n```bash\n# 提取智能体提示，用于 Claude Code CLI\nEXPERT=$(claudekit show agent typescript-expert)\ncat src\u002Fapp.ts | claude -p --append-system-prompt \"$EXPERT\" \"Review this code\"\n\n# 获取结构化元数据，便于与其他工具集成\nclaudekit show agent typescript-expert --format json | jq '.description'\n\n# 可与任何 AI 编程 CLI 工具配合使用\nclaudekit show agent react-expert | amp -x \"Fix React hooks violations\"\n```\n\n更多关于 Claude Code、Amp、Gemini CLI 等工具的示例，请参阅 [与外部 LLM 结合使用提示指南](docs\u002Fguides\u002Fusing-prompts-with-external-llms.md)。\n\n## 斜杠命令\n您可以在 Claude Code 中直接使用以下命令：\n\n### Git 与检查点\n- `\u002Fcheckpoint:create [msg]` - 保存当前状态\n- `\u002Fcheckpoint:restore [n]` - 恢复到指定检查点\n- `\u002Fcheckpoint:list` - 查看所有检查点\n- `\u002Fgit:commit` - 智能提交，遵循规范\n- `\u002Fgit:status` - 智能 Git 分析\n- `\u002Fgit:checkout [branch]` - 智能创建分支并切换，命名符合常规\n- `\u002Fgit:ignore-init` - 初始化 AI 安全的 .gitignore 模板\n\n### 代码质量\n- `\u002Fcode-review [target]` - 使用并行智能体进行多维度代码审查\n- `\u002Fvalidate-and-fix` - 运行所有质量检查\n\n### 开发工具\n- `\u002Fresearch [query]` - 使用专业子智能体进行深度并行研究\n- `\u002Fspec:create [feature]` - 生成功能规格说明书\n- `\u002Fspec:execute [file]` - 按照 6 阶段迭代流程实施规格说明书\n- `\u002Fagents-md:init` - 配置 AI 助手\n- `\u002Fcreate-subagent` - 构建自定义智能体\n- `\u002Fcreate-command` - 创建自定义命令\n\n[查看所有命令 →](docs\u002Freference\u002Fcommands.md)\n\n## 挂钩\n挂钩会在 Claude 工作时自动强制执行质量标准：\n\n### 可用钩子\n\n**文件安全钩子**（PreToolUse - 在文件访问前执行）\n- `file-guard` - 使用忽略文件模式阻止 AI 访问敏感文件（`.env`、密钥、凭据）\n\n**文件变更钩子**（PostToolUse - 在编辑后执行）\n- `typecheck-changed` - 对文件变更运行 TypeScript 类型检查\n- `lint-changed` - 对已更改文件运行代码风格检查（Biome、ESLint 等）\n- `test-changed` - 为已更改文件运行测试\n- `check-any-changed` - 禁止已更改的 TypeScript 文件中出现任何类型\n- `check-comment-replacement` - 检测代码是否被注释替换\n- `check-unused-parameters` - 检测懒惰重构：参数被添加下划线前缀而非直接移除\n\n**用户提示提交钩子**（在每个会话的首次用户提示时执行）\n- `codebase-map` - 使用 codebase-map CLI 每会话提供一次不可见的代码库上下文\n- `thinking-level` - 注入思考层级关键词以增强 Claude 的推理能力\n\n**PostToolUse 钩子**（也包括文件监控）\n- `codebase-map-update` - 当文件发生变更时增量更新代码库地图\n\n**项目级钩子**（通常用于 Stop\u002FSubagentStop 事件）\n- `typecheck-project` - 对整个项目进行 TypeScript 验证\n- `lint-project` - 对整个项目进行代码风格检查（Biome、ESLint 等）\n- `test-project` - 运行完整的测试套件\n- `create-checkpoint` - 在停止时自动创建 Git 检查点\n- `check-todos` - 验证待办事项是否已完成\n- `self-review` - 增强的目标性自审，可检测实现完整性，优先关注模拟实现、占位符代码和未完成的功能，而非代码质量问题\n\n[自审指南 →](docs\u002Fguides\u002Fself-review.md)\n\n### 钩子事件\n\n- **PreToolUse** - 在文件访问（读取、写入、编辑、多文件编辑）之前触发，用于权限控制\n- **PostToolUse** - 在文件修改（写入、编辑、多文件编辑）之后触发\n- **Stop** - 当 Claude Code 停止或对话结束时触发\n- **SubagentStop** - 当子代理完成任务时触发\n- **UserPromptSubmit** - 当用户在会话中提交第一条提示时触发\n\n### 钩子命令\n\n```bash\n# 查看所有可用钩子及其描述\nclaudekit-hooks list\n\n# 分析钩子性能和输出大小\nclaudekit-hooks profile                   # 分析所有钩子\nclaudekit-hooks profile typecheck-changed  # 分析特定钩子\nclaudekit-hooks profile --iterations 5    # 多次运行以获取平均值\n\n# 会话级别的钩子控制（需要活跃的 Claude Code 会话）\nclaudekit-hooks disable typecheck-changed  # 禁用当前会话中的特定钩子\nclaudekit-hooks enable typecheck          # 重新启用钩子（支持部分匹配）\nclaudekit-hooks status                     # 显示所有钩子的状态\nclaudekit-hooks status lint                # 显示特定钩子的状态\n```\n\n`profile` 命令有助于识别：\n- 执行缓慢的钩子（>5秒），会拖慢工作流程\n- 输出量大的钩子，会占用大量上下文窗口\n- 接近 10k 字符限制的 UserPromptSubmit 钩子\n\n通过会话级别的钩子控制，您可以在不修改配置文件的情况下，临时禁用特定 Claude Code 会话中的钩子。功能包括：\n- 钩子名称的模糊匹配（支持部分匹配）\n- 可视化状态指示器（🔒 已禁用，✅ 已启用）\n- 会话隔离状态（更改仅影响当前会话）\n- 自动通过会话记录 UUID 检测会话\n\n[钩子配置 →](docs\u002Freference\u002Fhooks.md) | [性能分析指南 →](docs\u002Fguides\u002Fhook-profiling.md)\n\n## 子代理\n\n针对不同领域的专业 AI 助手：\n\n**热门代理**\n- `code-review-expert` - 6 个并行代理进行代码评审（架构、安全、性能、测试、质量、文档）\n- `triage-expert` - 收集上下文并诊断问题，然后将其分配给专业专家\n- `documentation-expert` - 专注于文档结构、连贯性、流畅性和信息架构\n- `oracle` - 深度调试和复杂问题分析（[需单独设置](docs\u002Fintegrations\u002Foracle.md)）\n- `refactoring-expert` - 检测代码异味并提供全面的重构指导\n- `typescript-expert` - 专注于 TypeScript\u002FJavaScript 开发和类型系统\n- `react-expert` - 专注于 React 组件、Hooks 和性能优化\n- `testing-expert` - 提供测试策略、覆盖率和最佳实践建议\n- `database-expert` - 专注于查询优化和模式设计\n- `ai-sdk-expert` - 专注于 Vercel AI SDK v5 流式传输和模型集成\n- `nestjs-expert` - 专注于 Nest.js 架构和依赖注入\n- `kafka-expert` - 专注于 Apache Kafka 生产者\u002F消费者配置、主题管理、集群操作和性能调优\n- `loopback-expert` - 专注于 LoopBack 4 框架开发、模型关系、认证和依赖注入\n\n**使用方法**\n```bash\n# 列出所有可用代理及其令牌数量\nclaudekit list agents\n\n# 安装特定代理\nclaudekit setup --agents typescript-expert,react-expert\n```\n\n安装完成后，代理会自动工作，也可以直接调用：\n```\n“请使用 oracle 代理来调试这个问题”\n“让 typescript-expert 审核这个类型定义”\n```\n\n[查看所有代理 →](docs\u002Freference\u002Fsubagents.md)\n\n## 安全\n\n### 保护敏感文件\n\n`file-guard` 钩子会自动保护敏感文件免受 AI 访问。快速设置方法如下：\n\n```bash\n# 新项目：安装带有 file-guard 保护的 claudekit\nnpm install -g claudekit && claudekit setup --yes --force --hooks file-guard\n\n# 现有项目：将 file-guard 添加到现有设置中\nclaudekit setup --yes --force --hooks file-guard\n```\n\n或者手动创建 `.agentignore` 文件：\n\n```gitignore\n# .agentignore\n.env\n.env.*\n*.key\n*.pem\n.aws\u002Fcredentials\n.ssh\u002F*\n\n# 允许示例文件\n!.env.example\n```\n\n`file-guard` 钩子支持多种忽略文件格式（`.agentignore`、`.aiignore`、`.cursorignore`、`.geminiignore`、`.codeiumignore`、`.aiexclude`），并将所有这些格式中的模式合并在一起——这一独特功能可在团队使用任何 AI 工具时提供全面保护。\n\n**关键特性：**\n- **高级 Bash 命令分析**：解析复杂的 Shell 命令以检测敏感文件访问模式\n- **管道安全**：检测如 `find -name '*.env' | xargs cat` 和 `echo '.env' | xargs cat` 等危险构造\n- **变量扩展**：安全处理 Bash 变量、命令替换和参数扩展\n- **上传保护**：阻止通过 curl、httpie、scp、rsync 以及云 CLI 工具（aws、gcloud、az）上传文件\n- 合并所有忽略文件中的模式（不同于其他工具）\n- 支持 gitignore 样式的语法，包括否定规则（!）\n- 解析符号链接以防止绕过\n- 阻止项目之外的路径遍历尝试\n\n有关详细配置和最佳实践，请参阅 [File Guard 指南](docs\u002Fguides\u002Ffile-guard.md)。\n\n## 配置\n\nClaudekit 使用两个配置文件：\n\n**`.claude\u002Fsettings.json`** - 项目设置（钩子、命令）\n```json\n{\n  \"hooks\": {\n    \"PreToolUse\": [{\n      \"matcher\": \"Read|Edit|MultiEdit|Write\",\n      \"hooks\": [{\"type\": \"command\", \"command\": \"claudekit-hooks run file-guard\"}]\n    }],\n    \"PostToolUse\": [{\n      \"matcher\": \"Write|Edit|MultiEdit\",\n      \"hooks\": [{\"type\": \"command\", \"command\": \"claudekit-hooks run typecheck-changed\"}]\n    }],\n    \"Stop\": [{\n      \"matcher\": \"*\",\n      \"hooks\": [{\"type\": \"command\", \"command\": \"claudekit-hooks run create-checkpoint\"}]\n    }],\n    \"SubagentStop\": [{\n      \"matcher\": \"*\",\n      \"hooks\": [{\"type\": \"command\", \"command\": \"claudekit-hooks run create-checkpoint\"}]\n    }],\n    \"UserPromptSubmit\": [{\n      \"matcher\": \"*\",\n      \"hooks\": [{\"type\": \"command\", \"command\": \"claudekit-hooks run codebase-map\"}]\n    }]\n  }\n}\n```\n\n**`.claudekit\u002Fconfig.json`** - 钩子配置\n```json\n{\n  \"hooks\": {\n    \"typecheck-changed\": {\n      \"command\": \"npm run typecheck\"\n    },\n    \"thinking-level\": {\n      \"level\": 2\n    },\n    \"codebase-map\": {\n      \"include\": [\"src\u002F**\", \"lib\u002F**\"],\n      \"exclude\": [\"**\u002F*.test.ts\", \"**\u002F*.spec.ts\"],\n      \"format\": \"dsl\"\n    }\n  }\n}\n```\n\n[配置指南 →](docs\u002Fgetting-started\u002Fconfiguration.md)\n\n## 获取帮助\n\n- **文档**: [完整文档](docs\u002F)\n- **问题**: [GitHub Issues](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fissues)\n- **快速测试**: 运行 `claudekit doctor` 检查设置\n\n## 常见问题\n\n**钩子未触发？**\n```bash\nclaudekit doctor                # 检查配置\nnpm list -g claudekit           # 确认安装\n```\n\n**测试超时？**\n使用超快的测试配置来执行钩子：\n```json\n{\n  \"hooks\": {\n    \"test-project\": {\n      \"command\": \"npm run test:fast\",\n      \"timeout\": 30000\n    }\n  }\n}\n```\n\n或者配置一个自定义的更快测试命令：\n```json\n{\n  \"hooks\": {\n    \"test-project\": {\n      \"command\": \"npm run test:unit\",\n      \"timeout\": 50000\n    }\n  }\n}\n```\n\n**代码库地图未生成？**\n安装代码库地图 CLI 工具：\n```bash\nnpm install -g codebase-map\n# 检查安装：\ncodebase-map --version\n```\n\n[完整故障排除 →](docs\u002Fgetting-started\u002Ftroubleshooting.md)\n\n## MCP 集成\n\n部分命令支持可选的 MCP 服务器集成：\n\n### Context7（用于访问文档）\n```bash\n# 全局安装\nnpm install -g @upstash\u002Fcontext7-mcp\n\n# 添加到 Claude Code\nclaude mcp add context7 context7-mcp\n```\n\n启用 `\u002Fspec:create` 功能，以获取最新的库文档。\n\n## 开发\n\n```bash\n# 克隆并安装\ngit clone https:\u002F\u002Fgithub.com\u002Fclaudekit\u002Fclaudekit.git\ncd claudekit\nnpm install\n\n# 构建\nnpm run build\n\n# 测试\nnpm test\n\n# 验证生产构建\nnpm run validate:build         # 检查依赖和构建产物\nnpm run validate:install       # 测试干净安装\n\n# 创建开发用符号链接\nnpm run symlinks\n```\n\n## 贡献\n\n我们欢迎贡献！请参阅 [CONTRIBUTING.md](CONTRIBUTING.md) 查看指南。\n\n## 许可证\n\nMIT © [claudekit 贡献者](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fgraphs\u002Fcontributors)","# claudekit 快速上手指南\n\nclaudekit 是为 Claude Code 打造的智能护栏与工作流自动化工具。它能实时拦截错误、保存代码检查点，并通过专家级子代理（Subagents）增强 AI 编程能力。\n\n## 环境准备\n\n在开始之前，请确保满足以下系统要求：\n\n*   **Claude Code 订阅**：需要 **Max plan**（以优化 Token 使用并获得最佳体验）。\n*   **Node.js 版本**：需安装 Node.js 20 或更高版本。\n    *   检查版本命令：`node -v`\n*   **包管理器**：推荐使用 `npm`、`yarn` 或 `pnpm`。\n    *   *国内加速建议*：如下载缓慢，可配置淘宝镜像源：\n        ```bash\n        npm config set registry https:\u002F\u002Fregistry.npmmirror.com\n        ```\n\n## 安装步骤\n\n使用以下任一命令全局安装 claudekit：\n\n```bash\n# 使用 npm (推荐)\nnpm install -g claudekit\n\n# 或使用 yarn\nyarn global add claudekit\n\n# 或使用 pnpm\npnpm add -g claudekit\n```\n\n安装完成后，在项目根目录运行初始化向导：\n\n```bash\nclaudekit setup\n```\n> 提示：若希望使用默认配置快速完成设置，可运行 `claudekit setup --yes`。\n\n## 基本使用\n\n初始化后，直接在 **Claude Code** 会话中使用以下增强命令即可：\n\n### 1. 智能代码审查\n并行调用 6 个专用代理对代码进行多维度分析（架构、安全、性能等）：\n```bash\n\u002Fcode-review\n```\n\n### 2. 需求规格生成\n让 AI 研究现有代码库并撰写完整的功能规格说明书：\n```bash\n\u002Fspec:create \"你的下一个功能描述\"\n```\n\n### 3. Git 工作流增强\n按类型分组变更并建议提交策略：\n```bash\n\u002Fgit:status\n```\n\n### 4. 错误恢复与检查点\n如果重构失败或出现意外，一键恢复到之前的安全状态：\n```bash\n\u002Fcheckpoint:restore\n```\n\n### 核心特性速览\n*   **实时防护**：自动阻止 TypeScript 中的 `any` 类型滥用，即时发现测试失败。\n*   **安全护栏**：内置高级 Bash 安全分析，防止敏感文件（如 `.env`、SSH 密钥）被意外读取或泄露。\n*   **极速导航**：自动构建代码库地图，让 Claude 直接定位文件与依赖，减少无效搜索。\n*   **多代理协作**：内置研究专家、代码搜索专家及各类领域专家（React, DB, Next.js 等），自动分配复杂任务。","某全栈开发者正利用 Claude Code 重构一个遗留的电商订单模块，需要在保证类型安全的前提下快速迭代新功能。\n\n### 没有 claudekit 时\n- **类型安全缺失**：Claude 为图省事频繁使用 `any` 类型，导致潜在的运行时错误直到代码审查阶段才被发现。\n- **试错成本高昂**：一旦激进的重构导致项目崩溃，开发者不得不手动翻阅 Git 历史，耗时费力地找回可用版本。\n- **反馈滞后严重**：测试用例被破坏后，往往要等 Claude 完成所有任务并运行 CI 时才报错，修复链条过长。\n- **审查流于表面**：人工发起的代码审查通常只能获得通用的建议，缺乏针对架构、安全和性能的深度并行分析。\n\n### 使用 claudekit 后\n- **实时类型拦截**：当 Claude 试图写入 `any` 时，claudekit 立即阻断并提示“请使用具体类型：User | null\"，从源头杜绝隐患。\n- **一键状态回滚**：若重构失败，只需执行 `\u002Fcheckpoint:restore` 命令，即可瞬间将代码库恢复到上一个稳定快照。\n- **即时错误自愈**：测试一旦报错，claudekit 会让 Claude 立刻感知并自动修复，无需等待后续流程，实现“发现即解决”。\n- **专家级并行审查**：运行 `\u002Fcode-review` 命令，6 个专用智能体同时从架构、安全等维度深度扫描，输出带优先级排序的专业报告。\n\nclaudekit 通过实时防护网和自动化工作流，将原本滞后的“编写 - 出错 - 修复”循环转变为高效、安全的“智能协作”闭环。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fcarlrannaberg_claudekit_9029c0df.png","carlrannaberg","Carl Rannaberg","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002Fcarlrannaberg_a34ccc5d.jpg","Experienced SaaS-builder, ex-Pipedrive",null,"Tallinn","carlrannaberg@gmail.com","http:\u002F\u002Fcarlrannaberg.com","https:\u002F\u002Fgithub.com\u002Fcarlrannaberg",[86,90,94],{"name":87,"color":88,"percentage":89},"TypeScript","#3178c6",84.3,{"name":91,"color":92,"percentage":93},"Shell","#89e051",10.9,{"name":95,"color":96,"percentage":97},"JavaScript","#f1e05a",4.8,649,104,"2026-04-05T10:56:42","MIT","Linux, macOS, Windows","未说明",{"notes":105,"python":103,"dependencies":106},"该工具是基于 Node.js 的命令行工具，需通过 npm、yarn 或 pnpm 全局安装。核心依赖是 Claude Code 服务且必须订阅 Max 计划以获得最佳效果。无需 Python 环境或 GPU 支持。",[107,108],"Node.js 20+","Claude Code (Max plan)",[26,15,53],"2026-03-27T02:49:30.150509","2026-04-06T08:48:28.147210",[113,118,123,128,133,138,143,148],{"id":114,"question_zh":115,"answer_zh":116,"source_url":117},10385,"如何在单个会话中临时禁用或启用某个 Hook？","在 v9.0.0 版本中已支持按会话禁用和启用 Hook。你可以使用以下斜杠命令（Slash Commands）：\n- \u002Fhook:disable [hook-name]：禁用指定 Hook\n- \u002Fhook:enable [hook-name]：启用指定 Hook\n- \u002Fhook:status [hook-name]：查看 Hook 状态\n注意：Hook 名称不需要完全匹配，底层支持模糊匹配。","https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fissues\u002F10",{"id":119,"question_zh":120,"answer_zh":121,"source_url":122},10386,"Codebase Map 功能与 Serena MCP 有什么区别？","Codebase Map 功能是向上下文中注入代码库的地图结构，不涉及检索（retrieval）或编辑操作，也没有工具可调用的过程，它会在每个会话中自动添加。相比之下，Serena MCP 是基于 LSP 的检索和编辑工具集合。\n建议：Codebase Map 更适合小型代码库；对于大型复杂代码库，Serena MCP 是更好的选择。\n示例输出格式：\ncli\u002Fcli.ts > cli\u002Futils\u002Flogger\n  fn runCli():Promise\u003Cvoid> async","https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fissues\u002F7",{"id":124,"question_zh":125,"answer_zh":126,"source_url":127},10387,"如何配置 Codebase Map 以排除测试文件和依赖目录，从而减少 Token 消耗？","在 .claudekit\u002Fconfig.json 中配置 hooks.codebase-map 或 codebaseMap 部分，使用 exclude 数组排除不需要的文件。例如：\n{\n  \"hooks\": {\n    \"codebase-map\": {\n      \"format\": \"dsl\",\n      \"exclude\": [\n        \"node_modules\u002F**\",\n        \"**\u002F*.test.ts\",\n        \"**\u002F*.spec.js\",\n        \"**\u002Fdist\u002F**\",\n        \"**\u002Fbuild\u002F**\"\n      ]\n    }\n  }\n}\n提示：如果代码库较大导致 Token 消耗过高，建议仅包含核心部分，排除测试和非核心文件。此外，支持将 format 设置为 'graph'（介于 dsl 和 tree 之间）或 'tree' 以节省字符。还可以将其用于 SessionStart Hook，该 Hook 没有字符限制，但生成的地图会显示在对话中。","https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fissues\u002F5",{"id":129,"question_zh":130,"answer_zh":131,"source_url":132},10388,"安装脚本会对我的 .claude 目录做哪些修改？","安装步骤会根据你的选择修改用户级或项目级的 .claude\u002F 文件夹内容：\n1. 安装 Hooks 时：会修改 settings.json 文件。\n2. 安装斜杠命令（Slash Commands）或子代理（Subagents）时：会将相关文件复制到 .claude\u002F 下的对应文件夹。\n3. 备份机制：如果目标位置已存在文件，系统会自动创建原始文件的备份。\n注意：安装过程不会修改 CLAUDE.md 文件。","https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fissues\u002F9",{"id":134,"question_zh":135,"answer_zh":136,"source_url":137},10389,"是否支持使用 BiomeJS 替代 ESLint 和 Prettier？","是的，从 v0.8.4 版本开始，如果项目中存在 Biome 配置文件，lint-changed 和 lint-project Hook 会自动运行 BiomeJS。\n此外，你也可以在 .claudekit\u002Fconfig.json 中为 lint Hook 配置自定义命令来手动指定使用 Biome。","https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fissues\u002F11",{"id":139,"question_zh":140,"answer_zh":141,"source_url":142},10390,"运行命令时遇到 'Bash command permission check failed' 错误怎么办？","该问题通常出现在未使用 --dangerously-skip-permissions 标志运行 Claude 时，导致某些包含 shell 操作符的命令无法触发审批选项而直接失败。此问题已在最新版本中修复。请确保更新到最新版 claudekit，或者在运行命令时检查权限设置。","https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fissues\u002F6",{"id":144,"question_zh":145,"answer_zh":146,"source_url":147},10391,"Research 命令报错 'model: opus' 找不到模型怎么办？","这通常是由于子代理（Subagent）的前端元数据中使用了模型别名（如 'opus'），而 API 调用时未能正确解析所致。虽然文档说明支持模型别名，但这可能是 Claude Code 或 API 的临时 Bug。\n解决方法：将配置文件中的 model 值改为完整的模型 ID，例如将 'opus' 改为 'claude-opus-4-1-20250805'，即可恢复命令功能。","https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fissues\u002F18",{"id":149,"question_zh":150,"answer_zh":151,"source_url":152},10392,"Fileguard 阻止 Research 子代理将结果写入 tmp 目录怎么办？","这是一个已知问题，Fileguard 的安全机制曾错误地拦截了 Research 子代理向临时目录写入文件的操作。该问题已在 v0.9.3 版本中修复，请升级到此版本或更高版本即可解决。","https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fissues\u002F16",[154,159,164,169,174,179,184,189,194,199,204,209,214,219,224,229,234,239,244,249],{"id":155,"version":156,"summary_zh":157,"released_at":158},107614,"v0.9.5","## Installation\n\n```bash\nnpm install -g claudekit@0.9.5\n```\n\n## What's Changed\n\n\n### Fixed\n- **Shell Argument Escaping**: Fixed command execution to properly handle special characters in file paths\n  - Added shell quoting for arguments containing special characters like parentheses in Next.js route groups\n  - Prevents shell interpretation errors when running commands on files with complex names\n  - Resolves issue #28 where hooks would fail on modern frontend project structures\n- **Test Command Execution**: Improved compatibility with different test runners and package managers\n  - Fixed test command argument passing to work correctly with both npm\u002Fyarn\u002Fpnpm and direct test commands\n  - Only uses '--' separator when running through package managers, avoiding errors with direct test executables\n  - Ensures consistent test execution across different project configurations\n\n### Improved\n- **Test File Discovery**: Enhanced test-changed hook with more robust test file detection\n  - Added two-phase test file search: first co-located tests, then project-wide test directories\n  - Integrated fast-glob for efficient searching in tests\u002F, test\u002F, and __tests__ directories\n  - Added comprehensive JSDoc documentation for test discovery logic\n  - Better error handling for glob operations with graceful fallbacks\n\n## Links\n\n- 📖 [Full Changelog](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fblob\u002Fmain\u002FCHANGELOG.md)\n- 🔗 [NPM Package](https:\u002F\u002Fwww.npmjs.com\u002Fpackage\u002Fclaudekit\u002Fv\u002F0.9.5)\n- 📦 [All Releases](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Freleases)\n\n---\n\n*For the complete list of changes across all versions, see the [CHANGELOG.md](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fblob\u002Fmain\u002FCHANGELOG.md) file.*\n","2026-03-31T20:45:18",{"id":160,"version":161,"summary_zh":162,"released_at":163},107615,"v0.9.4","## Installation\n\n```bash\nnpm install -g claudekit@0.9.4\n```\n\n## What's Changed\n\n\n### Fixed\n- **Documentation URLs**: Updated broken Claude Code documentation URLs to use current domain structure\n  - Fixed SDK documentation URLs to include `\u002Fsdk\u002Fsdk-overview` path for accurate redirects\n  - Updated base domain from `docs.anthropic.com` to `docs.claude.com` across documentation files\n  - Ensures all Claude Code documentation links resolve correctly for users\n\n### Documentation\n- **Agent Formatting**: Improved formatting and structure of Kafka and LoopBack expert subagents\n  - Cleaned up verbose YAML frontmatter comments in LoopBack expert agent\n  - Enhanced code formatting consistency and documentation structure for better readability\n  - Streamlined agent definitions to focus on practical troubleshooting guidance\n\n## Links\n\n- 📖 [Full Changelog](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fblob\u002Fmain\u002FCHANGELOG.md)\n- 🔗 [NPM Package](https:\u002F\u002Fwww.npmjs.com\u002Fpackage\u002Fclaudekit\u002Fv\u002F0.9.4)\n- 📦 [All Releases](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Freleases)\n\n---\n\n*For the complete list of changes across all versions, see the [CHANGELOG.md](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fblob\u002Fmain\u002FCHANGELOG.md) file.*\n","2025-09-28T22:17:04",{"id":165,"version":166,"summary_zh":167,"released_at":168},107616,"v0.9.3","## Installation\n\n```bash\nnpm install -g claudekit@0.9.3\n```\n\n## What's Changed\n\n\n### Added\n- **Kafka Expert Subagent**: Added specialized subagent for Apache Kafka development and operations\n  - Added `kafka-expert.md` subagent with expertise in producer\u002Fconsumer configuration, topic management, cluster operations, serialization, performance tuning, and operational troubleshooting\n  - Added symlink at `.claude\u002Fagents\u002Fkafka` for easy access\n  - Covers Kafka errors, performance issues, consumer lag, rebalancing problems, and configuration challenges\n- **LoopBack Expert Subagent**: Added specialized subagent for LoopBack 4 framework development\n  - Added `loopback-expert.md` subagent with expertise in model relationships, authentication, dependency injection, API architecture, database performance, and production deployment\n  - Added symlink at `.claude\u002Fagents\u002Floopback` for easy access\n  - Handles JWT authentication errors, repository patterns, DI binding problems, and migration from LoopBack 3\n\n### Fixed\n- **File Guard Security**: Enhanced file protection service to allow absolute paths for research subagents\n  - Fixed issue where absolute paths (like `\u002Ftmp\u002Fresearch_results.txt`) were incorrectly blocked by file-guard hook\n  - Updated path protection logic in `FileProtectionService` to skip protection for absolute paths outside project root\n  - Allows research subagents to write to temporary directories and system paths as needed\n  - Updated tests to verify absolute paths are now allowed while maintaining protection for project files\n- **Link Validation**: Fixed markdown link checker configuration to handle npmjs.com URLs\n  - Added npmjs.com URL pattern to ignore list in `.markdown-link-check.json` to prevent false positives\n  - Added specific headers for npmjs.com requests to improve reliability when validation does occur\n  - Resolves issues with package documentation links being flagged as broken\n\n### Documentation\n- **Technical Overview**: Added comprehensive technical overview documentation with table of contents and architecture details\n- **Subagent Creation**: Streamlined subagent creation guide and updated model configurations\n- **Integration Guides**: Updated various documentation URLs to reflect current Claude Code documentation structure\n\n## Links\n\n- 📖 [Full Changelog](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fblob\u002Fmain\u002FCHANGELOG.md)\n- 🔗 [NPM Package](https:\u002F\u002Fwww.npmjs.com\u002Fpackage\u002Fclaudekit\u002Fv\u002F0.9.3)\n- 📦 [All Releases](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Freleases)\n\n---\n\n*For the complete list of changes across all versions, see the [CHANGELOG.md](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fblob\u002Fmain\u002FCHANGELOG.md) file.*\n","2025-09-28T19:09:25",{"id":170,"version":171,"summary_zh":172,"released_at":173},107617,"v0.9.2","## Installation\n\n```bash\nnpm install -g claudekit@0.9.2\n```\n\n## What's Changed\n\n\n### Fixed\n- **User-Scope Hook Configuration**: Fixed user-level installations to properly configure hooks in settings.json\n  - Fixed issue where user-scope installations (`claudekit setup --user`) were not creating hook configurations in `~\u002F.claude\u002Fsettings.json`\n  - Unified hook configuration logic by renaming `createProjectSettings` to `createSettings` and making it work for both user and project installations\n  - Ensured both user-scope and project-scope installations now properly generate embedded hook commands (`claudekit-hooks run \u003Chook-name>`)\n  - Updated user installation workflow to create `~\u002F.claude\u002Fsettings.json` with proper hook trigger configurations\n  - Enhanced installation consistency so hooks work correctly regardless of installation scope\n\n## Links\n\n- 📖 [Full Changelog](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fblob\u002Fmain\u002FCHANGELOG.md)\n- 🔗 [NPM Package](https:\u002F\u002Fwww.npmjs.com\u002Fpackage\u002Fclaudekit\u002Fv\u002F0.9.2)\n- 📦 [All Releases](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Freleases)\n\n---\n\n*For the complete list of changes across all versions, see the [CHANGELOG.md](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fblob\u002Fmain\u002FCHANGELOG.md) file.*\n","2025-09-17T11:27:36",{"id":175,"version":176,"summary_zh":177,"released_at":178},107618,"v0.9.1","## Installation\n\n```bash\nnpm install -g claudekit@0.9.1\n```\n\n## What's Changed\n\n\n### Changed\n- **Hook Extension Configuration**: Refactored hook utilities to eliminate code duplication and provide consistent file extension handling\n  - Extracted shared extension configuration utilities into `cli\u002Fhooks\u002Futils.ts` with `ExtensionConfigurable` interface\n  - Added `shouldProcessFileByExtension()` function for standardized file filtering across hooks\n  - Added `createExtensionPattern()` function for robust regex-based extension matching with proper escaping\n  - Updated `LintChangedHook` and `TestChangedHook` to use shared extension utilities instead of duplicated logic\n  - Enhanced type safety with guaranteed non-undefined file paths after extension validation\n\n## Links\n\n- 📖 [Full Changelog](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fblob\u002Fmain\u002FCHANGELOG.md)\n- 🔗 [NPM Package](https:\u002F\u002Fwww.npmjs.com\u002Fpackage\u002Fclaudekit\u002Fv\u002F0.9.1)\n- 📦 [All Releases](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Freleases)\n\n---\n\n*For the complete list of changes across all versions, see the [CHANGELOG.md](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fblob\u002Fmain\u002FCHANGELOG.md) file.*\n","2025-09-05T14:22:25",{"id":180,"version":181,"summary_zh":182,"released_at":183},107619,"v0.9.0","## Installation\n\n```bash\nnpm install -g claudekit@0.9.0\n```\n\n## What's Changed\n\n\n### Added\n- **Session-Based Hook Control**: Added comprehensive hook management system for granular control within Claude Code sessions\n  - Added `claudekit-hooks disable [hook-name]` CLI command to disable hooks for current session\n  - Added `claudekit-hooks enable [hook-name]` CLI command to re-enable disabled hooks for current session  \n  - Added `claudekit-hooks status [hook-name]` CLI command to show hook status with visual indicators\n  - Added `\u002Fhook:disable`, `\u002Fhook:enable`, and `\u002Fhook:status` slash commands for use within Claude Code\n  - Added `SessionTracker` and `SessionHookManager` classes for reliable session state management\n  - Added fuzzy matching for hook names with partial matching and helpful suggestions\n  - Added session state persistence in `~\u002F.claudekit\u002Fsessions\u002F` with atomic file operations\n  - Added automatic transcript UUID extraction from Claude Code environment for session identification\n\n### Changed\n- **Hook Architecture**: Enhanced base hook system to support session-based disable\u002Fenable functionality\n  - Updated `BaseHook` class to automatically check session state before execution\n  - Enhanced hook execution flow to skip disabled hooks silently without user notification\n  - Improved session identification logic with fallback mechanisms for development scenarios\n  - Standardized hook status reporting with emoji indicators (🔒 disabled, ✅ enabled)\n\n## Links\n\n- 📖 [Full Changelog](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fblob\u002Fmain\u002FCHANGELOG.md)\n- 🔗 [NPM Package](https:\u002F\u002Fwww.npmjs.com\u002Fpackage\u002Fclaudekit\u002Fv\u002F0.9.0)\n- 📦 [All Releases](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Freleases)\n\n---\n\n*For the complete list of changes across all versions, see the [CHANGELOG.md](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fblob\u002Fmain\u002FCHANGELOG.md) file.*\n","2025-09-05T14:02:42",{"id":185,"version":186,"summary_zh":187,"released_at":188},107620,"v0.8.14","## Installation\n\n```bash\nnpm install -g claudekit@0.8.14\n```\n\n## What's Changed\n\n\n### Added\n- **Biome Linting Support**: Added comprehensive Biome integration for modern JavaScript\u002FTypeScript linting\n  - Added `detectBiome()` function to detect Biome configuration files (`biome.json`, `biome.jsonc`) and dependencies (`@biomejs\u002Fbiome`, `rome`)\n  - Added `formatBiomeErrors()` utility function for consistent error formatting and actionable fix instructions\n  - Added Biome configuration file with selective linting rules focusing on correctness and security issues\n  - Added `lint:biome` npm script for project-wide Biome linting\n  - Added `@biomejs\u002Fbiome` as development dependency for linting capabilities\n\n### Changed\n- **Tool-Agnostic Linting Architecture**: Refactored linting hooks to support multiple linting tools simultaneously\n  - Updated `LintChangedHook` and `LintProjectHook` to run both Biome and ESLint when configured\n  - Changed hook display names from \"ESLint\" to generic \"Lint\" to reflect multi-tool support\n  - Updated hook dependencies from specific \"eslint\" to generic \"linter\" category\n  - Enhanced project detection system to identify multiple linting tool configurations\n  - Improved error reporting to show results from all configured linters with clear separation\n\n### Fixed\n- **Biome Configuration**: Refined Biome setup to prevent conflicts with existing tooling\n  - Fixed Biome formatter conflicts by disabling formatter and enabling only linter functionality  \n  - Fixed import organization conflicts by disabling Biome's import sorting to avoid ESLint\u002FPrettier conflicts\n  - Fixed test environment configuration with proper rule overrides for test files\n  - Updated test expectations to reflect new hook naming and multi-tool architecture\n\n## Links\n\n- 📖 [Full Changelog](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fblob\u002Fmain\u002FCHANGELOG.md)\n- 🔗 [NPM Package](https:\u002F\u002Fwww.npmjs.com\u002Fpackage\u002Fclaudekit\u002Fv\u002F0.8.14)\n- 📦 [All Releases](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Freleases)\n\n---\n\n*For the complete list of changes across all versions, see the [CHANGELOG.md](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fblob\u002Fmain\u002FCHANGELOG.md) file.*\n","2025-09-04T19:15:16",{"id":190,"version":191,"summary_zh":192,"released_at":193},107621,"v0.8.13","## Installation\n\n```bash\nnpm install -g claudekit@0.8.13\n```\n\n## What's Changed\n\n\n### Fixed\n- **File-Guard Security**: Enhanced .env template file handling to allow common template patterns\n  - Fixed issue where `.env.example`, `.env.template`, and `.env.sample` files were incorrectly blocked\n  - Updated default sensitive patterns to include negation rules for template files (`!.env.example`, `!.env.template`, `!.env.sample`)\n  - Improved ignore engine logic to properly handle negation patterns from custom ignore files\n  - Enhanced file protection service to use current working directory for ignore file detection\n  - Added comprehensive test coverage for ignore file parsing and negation pattern processing\n\n## Links\n\n- 📖 [Full Changelog](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fblob\u002Fmain\u002FCHANGELOG.md)\n- 🔗 [NPM Package](https:\u002F\u002Fwww.npmjs.com\u002Fpackage\u002Fclaudekit\u002Fv\u002F0.8.13)\n- 📦 [All Releases](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Freleases)\n\n---\n\n*For the complete list of changes across all versions, see the [CHANGELOG.md](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fblob\u002Fmain\u002FCHANGELOG.md) file.*\n","2025-09-03T08:52:22",{"id":195,"version":196,"summary_zh":197,"released_at":198},107622,"v0.8.12","## Installation\n\n```bash\nnpm install -g claudekit@0.8.12\n```\n\n## What's Changed\n\n\n### Added\n- **Status Command**: Added new CLI status checking functionality for tool integrations\n  - Added `claudekit status` command with subcommand support for checking tool availability\n  - Added `claudekit status stm` subcommand to check Simple Task Master (STM) installation and initialization status\n  - Added status command with `--verbose` and `--quiet` options for flexible output control\n  - Returns structured status messages: \"Not installed\", \"Available but not initialized\", or \"Available and initialized\"\n\n### Fixed\n- **Specification Commands**: Resolved Claude Code permission errors for STM status checks\n  - Fixed permission issues in spec decompose and execute commands when checking STM availability\n  - Updated allowed-tools configuration to include necessary Bash commands for STM integration\n  - Enhanced specification workflow reliability when STM tool is not available or not initialized\n\n## Links\n\n- 📖 [Full Changelog](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fblob\u002Fmain\u002FCHANGELOG.md)\n- 🔗 [NPM Package](https:\u002F\u002Fwww.npmjs.com\u002Fpackage\u002Fclaudekit\u002Fv\u002F0.8.12)\n- 📦 [All Releases](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Freleases)\n\n---\n\n*For the complete list of changes across all versions, see the [CHANGELOG.md](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fblob\u002Fmain\u002FCHANGELOG.md) file.*\n","2025-09-01T06:47:26",{"id":200,"version":201,"summary_zh":202,"released_at":203},107623,"v0.8.11","## Installation\n\n```bash\nnpm install -g claudekit@0.8.11\n```\n\n## What's Changed\n\n\n### Fixed\n- **Codebase Map Hook**: Resolved session handling and debugging issues\n  - Fixed profile test session detection to prevent caching during performance testing\n  - Enhanced manual run handling to always generate fresh output when session_id is 'unknown'\n  - Added debug logging capabilities via `DEBUG=true` environment variable for troubleshooting\n  - Improved session tracking logic to skip persistent flags for manual and profile test runs\n  - Updated unit tests to reflect corrected behavior for manual execution scenarios\n\n## Links\n\n- 📖 [Full Changelog](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fblob\u002Fmain\u002FCHANGELOG.md)\n- 🔗 [NPM Package](https:\u002F\u002Fwww.npmjs.com\u002Fpackage\u002Fclaudekit\u002Fv\u002F0.8.11)\n- 📦 [All Releases](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Freleases)\n\n---\n\n*For the complete list of changes across all versions, see the [CHANGELOG.md](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fblob\u002Fmain\u002FCHANGELOG.md) file.*\n","2025-08-31T19:57:09",{"id":205,"version":206,"summary_zh":207,"released_at":208},107624,"v0.8.10","## Installation\n\n```bash\nnpm install -g claudekit@0.8.10\n```\n\n## What's Changed\n\n\n### Fixed\n- **File-Guard Security**: Improved git command recognition in bash command parser\n  - Fixed false positives where legitimate git revision syntax (e.g., `@{u}..HEAD`, `origin\u002Fmain..HEAD`, `HEAD~5..HEAD`) was incorrectly flagged as file access\n  - Added `isGitRevisionSyntax()` method to properly detect and skip git revision patterns in command analysis\n  - Enhanced pattern matching to distinguish between git refs and actual file paths\n  - Improved security analysis accuracy while maintaining protection against actual file access attempts\n\n## Links\n\n- 📖 [Full Changelog](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fblob\u002Fmain\u002FCHANGELOG.md)\n- 🔗 [NPM Package](https:\u002F\u002Fwww.npmjs.com\u002Fpackage\u002Fclaudekit\u002Fv\u002F0.8.10)\n- 📦 [All Releases](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Freleases)\n\n---\n\n*For the complete list of changes across all versions, see the [CHANGELOG.md](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fblob\u002Fmain\u002FCHANGELOG.md) file.*\n","2025-08-31T16:44:42",{"id":210,"version":211,"summary_zh":212,"released_at":213},107625,"v0.8.9","## Installation\n\n```bash\nnpm install -g claudekit@0.8.9\n```\n\n## What's Changed\n\n\n### Changed\n- **Dependency Management**: Standardized file globbing library for consistency\n  - Migrated from `glob` to `fast-glob` across all file operations for better performance\n  - Updated `subagent-detector.ts` and `migrate-node-imports.js` to use consistent globbing\n  - Improved build tool compatibility and faster file discovery operations\n- **Build Validation**: Enhanced dependency validation with source-based analysis\n  - Completely rewrote `validate-dependencies.js` to analyze actual TypeScript imports\n  - Improved accuracy in detecting missing and unused dependencies\n  - Better handling of Node.js built-in modules with `node:` prefix\n  - Validates imports directly from source files instead of built artifacts\n- **Test Infrastructure**: Enhanced test runner with silent mode for cleaner CI output\n  - Added `--silent` flag to bash test runner for streamlined release preparation\n  - Updated CI workflows to use npm scripts instead of direct bash test calls\n  - Enhanced test framework and reporter with silent mode support\n  - Improved release validation workflow with cleaner test output\n\n### Fixed\n- **CI\u002FCD Pipeline**: Resolved test execution issues in continuous integration\n  - Fixed GitHub Actions workflows to properly execute tests via npm scripts\n  - Improved test-hooks workflow reliability and consistency\n  - Resolved bash test integration issues in automated builds\n- **Code Organization**: Eliminated code duplication in file-guard service classes\n  - Extracted shared utility functions (`globToRegExp`, `untildify`) into dedicated module\n  - Updated file-guard services to use shared utilities instead of duplicate implementations\n  - Improved maintainability and consistency across file protection components\n\n## Links\n\n- 📖 [Full Changelog](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fblob\u002Fmain\u002FCHANGELOG.md)\n- 🔗 [NPM Package](https:\u002F\u002Fwww.npmjs.com\u002Fpackage\u002Fclaudekit\u002Fv\u002F0.8.9)\n- 📦 [All Releases](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Freleases)\n\n---\n\n*For the complete list of changes across all versions, see the [CHANGELOG.md](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fblob\u002Fmain\u002FCHANGELOG.md) file.*\n","2025-08-31T12:26:50",{"id":215,"version":216,"summary_zh":217,"released_at":218},107626,"v0.8.8","## Installation\n\n```bash\nnpm install -g claudekit@0.8.8\n```\n\n## What's Changed\n\n\n### Added\n- **Comprehensive Bash Security Analysis**: Enhanced file-guard hook with advanced bash command parsing\n  - Added `BashCommandParser` class with comprehensive command analysis and path extraction\n  - Added `SecurityHeuristicsEngine` for detecting sensitive file access patterns in pipelines\n  - Added `FileProtectionService` with support for multiple AI ignore file formats\n  - Added extensive bash command validation with 590+ lines of test coverage\n  - Supports detection of sensitive file access via complex shell constructs (pipes, variables, xargs, etc.)\n- **Enhanced Test Infrastructure**: Expanded testing capabilities\n  - Added comprehensive bash-specific test suite with 47 test cases\n  - Added parallel test execution with both Vitest and bash test framework\n  - Enhanced test scripts configuration for better CI\u002FCD integration\n\n### Changed  \n- **File-guard Architecture**: Refactored file-guard hook into service-oriented architecture\n  - Restructured monolithic file-guard hook into specialized service classes\n  - Improved separation of concerns with dedicated parsing, security, and protection services\n  - Enhanced maintainability and testability of security-critical code\n- **Bash Tool Integration**: Extended file-guard protection to cover Bash tool usage\n  - Updated hook matcher pattern to include `Bash` tool in addition to file operations\n  - Enhanced security coverage for AI assistant bash command execution\n- **Dependency Management**: Added new production dependencies for enhanced functionality\n  - Added `fast-glob` for high-performance file pattern matching\n  - Added `ignore` for gitignore-style pattern processing\n  - Added `is-path-inside` for secure path boundary validation\n  - Added `untildify` for safe home directory path expansion\n  - Replaced `glob` with `fast-glob` for better performance\n\n### Fixed\n- **Build System**: Fixed broken symlink causing CI failures\n  - Removed broken `.claude\u002Fcommands\u002Ftmp` symlink that was pointing to deleted test directory\n  - Resolves build and deployment pipeline failures\n- **Documentation**: Fixed thinking-level hook configuration examples\n  - Removed non-existent `enabled` flag from all configuration examples in documentation\n  - Changed \"superthink\" to \"ultrathink\" to match actual implementation\n  - Added level 0 documentation for properly disabling the hook\n  - Updated all configuration examples to reflect actual schema and available options\n\n## Links\n\n- 📖 [Full Changelog](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fblob\u002Fmain\u002FCHANGELOG.md)\n- 🔗 [NPM Package](https:\u002F\u002Fwww.npmjs.com\u002Fpackage\u002Fclaudekit\u002Fv\u002F0.8.8)\n- 📦 [All Releases](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Freleases)\n\n---\n\n*For the complete list of changes across all versions, see the [CHANGELOG.md](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fblob\u002Fmain\u002FCHANGELOG.md) file.*\n","2025-08-31T06:47:42",{"id":220,"version":221,"summary_zh":222,"released_at":223},107627,"v0.8.7","## Installation\n\n```bash\nnpm install -g claudekit@0.8.7\n```\n\n## What's Changed\n\n\n### Fixed\n- **NPM Package**: Fixed test files being included in published npm package\n  - Removed explicit `files` array from package.json that was including test directories\n  - Added comprehensive .npmignore file with 110+ exclusion patterns\n  - Ensures only production code (dist\u002F, bin\u002F, src\u002F, README.md) is published to npm\n  - Significantly reduces package size by excluding development and test artifacts\n\n## Links\n\n- 📖 [Full Changelog](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fblob\u002Fmain\u002FCHANGELOG.md)\n- 🔗 [NPM Package](https:\u002F\u002Fwww.npmjs.com\u002Fpackage\u002Fclaudekit\u002Fv\u002F0.8.7)\n- 📦 [All Releases](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Freleases)\n\n---\n\n*For the complete list of changes across all versions, see the [CHANGELOG.md](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fblob\u002Fmain\u002FCHANGELOG.md) file.*\n","2025-08-29T07:40:56",{"id":225,"version":226,"summary_zh":227,"released_at":228},107628,"v0.8.6","## Installation\n\n```bash\nnpm install -g claudekit@0.8.6\n```\n\n## What's Changed\n\n\n### Added\n- **Git Workflow Commands**: Enhanced git functionality with new commands\n  - Added `git:checkout` command for smart branch switching with auto-completion\n  - Added `git:ignore-init` command for initializing AI-safe .gitignore patterns\n- **Research Workflow System**: Multi-agent research command for parallel information gathering\n  - Added `research` command with concurrent agent execution for comprehensive analysis\n  - Added `research-expert` subagent with specialized research capabilities\n  - Supports parallel research across multiple domains with structured output\n- **Enhanced File Security**: Expanded sensitive file protection patterns\n  - Added comprehensive sensitive patterns module (`cli\u002Fhooks\u002Fsensitive-patterns.ts`) with 12 security categories\n  - Added 195+ sensitive file patterns covering cloud credentials, cryptocurrencies, databases, tokens, and production data\n  - Enhanced file-guard hook with categorized pattern management and validation utilities\n  - Added extensive test coverage (330+ lines) for all sensitive file pattern categories\n\n### Changed\n- **Command Categories**: Renamed command category system for better clarity\n  - Changed `ai-assistant` category to `claude-setup` throughout the codebase\n  - Updated setup wizard command grouping: \"🤖 Claude Setup & Configuration\" \n  - Updated recommendation engine and component discovery to use new category names\n- **Release Process Integration**: Enhanced release preparation with automated quality checks\n  - Integrated claudekit linting (`@carlrannaberg\u002Fcclint`) into release preparation workflow\n  - Added JSON-formatted linting output parsing and error reporting\n  - Release preparation now validates all command and agent files before publishing\n- **Setup Command Organization**: Improved command organization in setup wizard\n  - Added new git workflow commands to recommended workflow group\n  - Updated command descriptions for better clarity and consistency\n  - Enhanced command grouping logic with better category inference\n\n### Fixed\n- **Documentation Links**: Resolved broken documentation link patterns\n  - Fixed markdown link validation patterns in configuration\n  - Updated documentation structure and internal linking consistency\n- **Build Tool Recognition**: Enhanced ESBuild external dependency detection\n  - Improved validation to properly recognize Node.js built-in modules with `node:` prefix\n  - Fixed build tool configuration for better dependency bundling\n\n## Links\n\n- 📖 [Full Changelog](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fblob\u002Fmain\u002FCHANGELOG.md)\n- 🔗 [NPM Package](https:\u002F\u002Fwww.npmjs.com\u002Fpackage\u002Fclaudekit\u002Fv\u002F0.8.6)\n- 📦 [All Releases](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Freleases)\n\n---\n\n*For the complete list of changes across all versions, see the [CHANGELOG.md](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fblob\u002Fmain\u002FCHANGELOG.md) file.*\n","2025-08-28T21:16:09",{"id":230,"version":231,"summary_zh":232,"released_at":233},107629,"v0.8.5","## Installation\n\n```bash\nnpm install -g claudekit@0.8.5\n```\n\n## What's Changed\n\n\n### Changed\n- **Linting Architecture**: Refactored internal linting system to use external SDK for better maintainability\n  - Replaced internal command linting implementation with `@carlrannaberg\u002Fcclint` SDK (v0.2.7)\n  - Replaced internal agent linting implementation with `@carlrannaberg\u002Fcclint` SDK\n  - Removed 720+ lines of internal linting code (`cli\u002Flib\u002Flinters\u002Fcommands.ts`, `cli\u002Flib\u002Flinters\u002Fsubagents.ts`)\n  - Updated `lint-commands` and `lint-subagents` (now `lint-agents`) commands to use SDK-based approach\n  - Enhanced linting with `followSymlinks: true` option for better symlink handling\n  - Improved error counting and statistics reporting through SDK integration\n- **Command Naming**: Renamed `lint-subagents` command to `lint-agents` for consistency with current terminology\n- **Node.js Import Standards**: Fixed validation utilities to use `node:` prefix for built-in modules (child_process)\n\n### Removed\n- **Internal Linting Code**: Removed internal command and agent linting implementations in favor of external SDK\n  - Deleted `cli\u002Flib\u002Flinters\u002Fcommands.ts` (266 lines) - functionality moved to `@carlrannaberg\u002Fcclint`\n  - Deleted `cli\u002Flib\u002Flinters\u002Fsubagents.ts` (454 lines) - functionality moved to `@carlrannaberg\u002Fcclint`\n  - Deleted associated test files (`tests\u002Flib\u002Flinters\u002Fsubagents.test.ts`)\n  - Reduced codebase complexity while maintaining all linting functionality\n\n### Fixed\n- **Shebang Standardization**: Updated `scripts\u002Fprepare-release.sh` to use standard `#!\u002Fusr\u002Fbin\u002Fenv bash` shebang\n\n## Links\n\n- 📖 [Full Changelog](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fblob\u002Fmain\u002FCHANGELOG.md)\n- 🔗 [NPM Package](https:\u002F\u002Fwww.npmjs.com\u002Fpackage\u002Fclaudekit\u002Fv\u002F0.8.5)\n- 📦 [All Releases](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Freleases)\n\n---\n\n*For the complete list of changes across all versions, see the [CHANGELOG.md](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fblob\u002Fmain\u002FCHANGELOG.md) file.*\n","2025-08-28T15:20:22",{"id":235,"version":236,"summary_zh":237,"released_at":238},107630,"v0.8.4","## Installation\n\n```bash\nnpm install -g claudekit@0.8.4\n```\n\n## What's Changed\n\n\n### Fixed\n- **Hook Installation**: Fixed duplicate hook installation for multi-trigger hooks\n  - Hooks that support multiple trigger events (e.g., both `Stop` and `SubagentStop`) are now installed only once per selection\n  - Added `hookTriggerEvents` Map to track specific trigger event assignments during setup\n  - For individually selected hooks, only the primary (first) trigger event is used to prevent duplicates\n  - For group-selected hooks, the hook is assigned to the specific trigger event of that group\n  - Eliminates cluttered settings.json files with duplicate hook entries and unnecessary multiple hook executions\n- **Documentation Link Validation**: Fixed broken link patterns in markdown link checker configuration\n  - Added specific replacement pattern for self-review hook documentation links\n  - Added status code 525 to accepted alive status codes for better link validation coverage\n\n## Links\n\n- 📖 [Full Changelog](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fblob\u002Fmain\u002FCHANGELOG.md)\n- 🔗 [NPM Package](https:\u002F\u002Fwww.npmjs.com\u002Fpackage\u002Fclaudekit\u002Fv\u002F0.8.4)\n- 📦 [All Releases](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Freleases)\n\n---\n\n*For the complete list of changes across all versions, see the [CHANGELOG.md](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fblob\u002Fmain\u002FCHANGELOG.md) file.*\n","2025-08-27T20:34:00",{"id":240,"version":241,"summary_zh":242,"released_at":243},107631,"v0.8.3","## Installation\n\n```bash\nnpm install -g claudekit@0.8.3\n```\n\n## What's Changed\n\n\n### Changed\n- **Thinking Level Hook**: Simplified configuration and improved defaults\n  - Simplified to 4 levels (0-3) instead of 5, as \"think harder\" and \"ultrathink\" both trigger 31,999 tokens\n  - Changed keywords to use special Claude Code keywords: \"think\" (level 1), \"megathink\" (level 2), \"ultrathink\" (level 3)\n  - Changed default thinking level from 0 to 2 (\"megathink\") for better reasoning out of the box\n  - Removed unnecessary `enabled` flag - use `level: 0` to disable the hook\n\n### Added\n- **Test Coverage**: Added comprehensive unit tests for thinking-level hook\n  - 241 lines of test coverage with 493 test cases validating all hook behavior\n  - Tests cover default behavior, level configuration, invalid input handling, and edge cases\n  - Ensures robust validation of thinking level keyword injection\n\n### Fixed\n- **Dead Code Removal**: Cleaned up unused configuration schemas\n  - Removed `ThinkingBudgetConfigSchema` which was added but never implemented\n  - Cleaned up references to thinking-budget throughout the codebase\n\n## Links\n\n- 📖 [Full Changelog](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fblob\u002Fmain\u002FCHANGELOG.md)\n- 🔗 [NPM Package](https:\u002F\u002Fwww.npmjs.com\u002Fpackage\u002Fclaudekit\u002Fv\u002F0.8.3)\n- 📦 [All Releases](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Freleases)\n\n---\n\n*For the complete list of changes across all versions, see the [CHANGELOG.md](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fblob\u002Fmain\u002FCHANGELOG.md) file.*\n","2025-08-26T06:59:09",{"id":245,"version":246,"summary_zh":247,"released_at":248},107632,"v0.8.2","## Installation\n\n```bash\nnpm install -g claudekit@0.8.2\n```\n\n## What's Changed\n\n\n### Added\n- **Hook Performance Profiling System**: Comprehensive performance analysis for hooks\n  - Added `claudekit-hooks profile` command for measuring hook execution time and output size\n  - Performance analysis includes execution time, character count, and estimated token usage\n  - Support for single runs or averaged measurements across multiple iterations\n  - Automatic detection of configured hooks from `.claude\u002Fsettings.json` for batch profiling\n  - Color-coded warnings for slow hooks (>5 seconds) and output size limits\n  - Specialized UserPromptSubmit limit checking for hooks that inject context into Claude prompts\n  - Comprehensive test coverage with 493 test cases for profiling system validation\n- **Enhanced Hook Testing Infrastructure**: Improved hook validation and testing capabilities\n  - Added intelligent test payload generation for different hook types (PostToolUse, Stop, UserPromptSubmit)\n  - Enhanced file-guard hook testing with support for all AI ignore file formats (`.agentignore`, `.aiignore`, `.aiexclude`, `.geminiignore`, `.codeiumignore`, `.cursorignore`)\n  - Improved pattern matching for synthetic test file generation from glob patterns\n  - Added comprehensive test transcript creation for hooks requiring conversation context\n- **Hook Performance Constants**: Standardized performance thresholds and limits\n  - Added performance threshold constants for execution time and output size monitoring\n  - Claude Code specific output limits (10k character truncation threshold)\n  - Token estimation utilities for output size analysis\n\n### Fixed\n- **File-guard Pattern Matching**: Improved ignore file processing and pattern handling\n  - Fixed file-guard to check all supported ignore files (6 different AI tool formats)\n  - Enhanced pattern-to-filepath conversion for more accurate testing\n  - Improved glob pattern handling for nested directories and wildcards\n- **Hook Output Capture**: Enhanced hook execution output collection\n  - Fixed hook profiling to properly capture both stdout and stderr streams\n  - Improved output size limiting with 10MB buffer for large hook outputs\n  - Enhanced error handling for hook execution failures during profiling\n- **Test Infrastructure**: Resolved testing framework issues\n  - Fixed Date constructor mocking in tests to avoid breaking Date functionality\n  - Improved mock fs-extra implementation to use proper Date constructors\n  - Enhanced test stability and reliability across different execution environments\n\n### Changed\n- **Hook Runner Architecture**: Enhanced hook execution and output management\n  - Updated HookRunner with standalone hook execution function for profiling integration\n  - Improved output capture system with memory limits and truncation handling\n  - Enhanced hook execution error handling and graceful failure management\n- **Hook CLI Interface**: Extended command-line interface capabilities\n  - Added profile subcommand to hooks CLI with iterations parameter support\n  - Enhanced argument validation for profile command parameters\n  - Improved CLI help text and command structure for profiling features\n\n## Links\n\n- 📖 [Full Changelog](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fblob\u002Fmain\u002FCHANGELOG.md)\n- 🔗 [NPM Package](https:\u002F\u002Fwww.npmjs.com\u002Fpackage\u002Fclaudekit\u002Fv\u002F0.8.2)\n- 📦 [All Releases](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Freleases)\n\n---\n\n*For the complete list of changes across all versions, see the [CHANGELOG.md](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fblob\u002Fmain\u002FCHANGELOG.md) file.*\n","2025-08-25T20:40:39",{"id":250,"version":251,"summary_zh":252,"released_at":253},107633,"v0.8.1","## Installation\n\n```bash\nnpm install -g claudekit@0.8.1\n```\n\n## What's Changed\n\n\n### Fixed\n- **Agent Registry Consistency**: Fixed agent registry to use consistent naming conventions\n  - Updated testing agent reference from `vitest-expert` to `vitest-testing-expert` in agent registry\n  - Ensures proper agent resolution and availability detection in spec:execute workflows\n  - Aligns with standardized naming convention where framework-specific testing experts use `-testing-` suffix\n\n### Added\n- **Iterative Code Review Workflow**: Enhanced spec:execute command with comprehensive quality assurance\n  - Added Phase 2: Test Writing with automatic selection of appropriate testing experts\n  - Added Phase 3: Code Review with code-review-expert integration for comprehensive analysis\n  - Added Phase 4: Iterative Improvement with specialist agent selection for fixing identified issues\n  - Added Phase 5: Atomic commit creation with standardized commit message format\n  - Added Phase 6: Progress tracking with quality metrics monitoring\n- **Dynamic Agent Availability Detection**: Improved agent orchestration in spec:execute\n  - Added real-time agent availability checks using `claudekit list agents`\n  - Dynamic selection of specialized agents based on actual availability rather than hardcoded lists\n  - Fallback to general-purpose agents when specialized experts are unavailable\n  - Enhanced agent matching logic for optimal task assignment\n\n### Changed\n- **Spec Execution Architecture**: Restructured spec:execute for better quality and maintainability\n  - Reorganized workflow into 6 distinct phases with clear quality gates\n  - Enhanced task management integration with both STM and TodoWrite systems\n  - Improved error handling and blocked task management\n  - Added comprehensive success criteria validation\n  - Enhanced progress tracking with status monitoring and quality metrics\n\n## Links\n\n- 📖 [Full Changelog](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fblob\u002Fmain\u002FCHANGELOG.md)\n- 🔗 [NPM Package](https:\u002F\u002Fwww.npmjs.com\u002Fpackage\u002Fclaudekit\u002Fv\u002F0.8.1)\n- 📦 [All Releases](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Freleases)\n\n---\n\n*For the complete list of changes across all versions, see the [CHANGELOG.md](https:\u002F\u002Fgithub.com\u002Fcarlrannaberg\u002Fclaudekit\u002Fblob\u002Fmain\u002FCHANGELOG.md) file.*\n","2025-08-24T20:04:20"]