云舒的 Skills 搭子们 / Yunshu's Claude Code Skills

English | 中文

中文

📖 简介

这是一个精心打造的 Claude Code Skills 集合，旨在提升软件开发和产品管理的效率。每个 Skill 都经过实战验证，帮助你在日常工作中更加高效。

📊 Skills 总览

目前共包含 30 个 Skills，分为以下 3 大类：

💻 AI 编程（15 个）

✍️ 思考与写作（10 个）

💼 日常办公与效率（5 个）

✨ 包含的 Skills

🎨 配图助手 (Image Assistant)

描述: 把文章/模块内容转成统一风格、少字高可读的 16:9 信息图提示词

适用场景:

• 文章需要配图但不知道怎么设计
• PPT、海报、社媒图需要统一风格
• 内容太多字，想要更趣味、更好读的视觉呈现
• 需要批量生成配图提示词

核心功能:

• 📋 需求澄清：挖掘内容、场景、受众和字数偏好
• 🗂️ 配图规划：拆分内容，定义图清单（几张图/每张讲什么）
• ✍️ 文案定稿：逐字定稿"图上写什么"（Copy Spec）
• 🎯 提示词封装：生成可复制的生图提示词，支持批量出图
• 🔄 迭代润色：根据反馈减字、换隐喻、提升可读性

触发方式:

这段内容做个图/配几张图
给我两张出图提示词
字太多不好看，帮我更趣味、更好读
/image /配图 /出图

🧠 思维挖掘助手 (Thought Mining)

描述: 通过对话帮助你把脑子里的零散想法倒出来、记录下来、整理成文章

适用场景:

• 想写文章但思路不清晰
• 有很多零散想法需要整理
• 需要从混乱的思考中提炼核心观点

核心功能:

• 📝 思维挖掘：引导式对话，帮你说出并记录想法
• 🎯 选题确定：从洞察中找到核心观点
• ✅ 观点验证：联网搜索验证理解是否正确
• ✍️ 写作辅助：逻辑检查、文字润色、金句提炼
• 🔍 最终审核：发布前的全面检查

触发方式:

我想写一篇关于 XX 的文章
帮我整理一下我的想法
/thought-mining

📋 PRD 文档撰写助手 (PRD Doc Writer)

描述: 以故事驱动的方式，帮助你撰写和迭代完善 PRD/需求文档

适用场景:

• 需要撰写产品需求文档
• 想用用户故事的方式梳理需求
• 需要用图表减少需求歧义

核心功能:

• 🗺️ 用户旅程地图：构建宏观业务流程
• 📖 故事化需求：每个功能点都是一个完整的用户故事
• 🎨 ASCII 线框图：可视化页面布局
• 📊 Mermaid 图表：流程图/状态图/时序图
• ✅ 阶段性确认：确保每一步都与你达成共识

触发方式:

帮我写 PRD
梳理需求文档
/prd-doc-writer

🎨 新功能设计探索 (Design Exploration)

描述: 新功能设计探索流程。从模糊想法产出可交付的设计参考文档，作为 PRD 阶段的输入

适用场景:

• 有模糊想法要做新功能/新模块
• 需要批量探索多种设计方案
• 需要完整覆盖页面状态（正常/异常/边界）
• 需要产出设计稿作为开发参考

核心功能:

• 📝 需求收敛：明确做什么、不做什么、边界在哪
• 🔍 技术调研：了解外部数据约束和技术可行性
• 🎨 ASCII 批量探索：一次出 5-8 个方案供选择
• 📐 HTML 设计稿：基于选定方案产出 mockup
• 📊 全状态覆盖：正常态/加载态/空态/错误态/边界情况
• 📋 交互规则表：前端开发直接对照实现

触发方式:

我想做个新功能
帮我设计一下这个模块
出个设计方案
/design-exploration

🔄 需求变更工作流 (Requirement Change Workflow)

描述: 标准化需求变更流程，避免改需求时的混乱和代码崩溃

适用场景:

• 需要修改现有功能的需求
• 改需求时经常出现意外 bug
• 需要一个可靠的变更验证流程
• 特别适合 Chrome 扩展等复杂项目

核心功能:

• 📝 需求澄清：锁定变更范围和验收标准
• 🔍 现状基线：从代码中确认当前行为
• ⚠️ 影响评估：评估风险和变更范围
• 🎯 设计方案：提出新设计并获得批准
• 🛠️ 最小化实现：小范围、局部化的代码修改
• ✅ 回归测试：固定的验证清单
• 📚 文档维护：决策日志和文档更新

触发方式:

改需求/需求变更
调整交互/改功能
/req-change-workflow

📚 课程构建器 (Lesson Builder)

描述: 通过讨论驱动的方式，帮助你快速完成课程大纲和课件

适用场景:

• 需要快速备好一节课
• 已有清晰想法，需要整理成文档
• 需要迭代现有课程大纲
• 准备培训或教学内容

核心功能:

• 💭 共创大纲：通过讨论挖掘想法，形成清晰课程框架
• 📖 课件撰写：基于大纲写出完整课件内容
• 🎯 框架优先：先确认框架再写细节，避免返工
• ⚡ 快速迭代：支持快速共创和严格确认两种模式
• 📋 最少文档：只产出需要的内容（大纲/课件/补充材料）

触发方式:

备课
做课件/准备课程
/lesson-builder

📦 需求池管理 (Backlog Manager)

描述: 需求池管理。随时抛出想法/痛点，AI 负责追问、整理、合并、归档。准备开新版本时协助筛选。痛点驱动，不做提前排期

适用场景:

• 随时记录产品想法和功能需求
• 管理零散的需求，避免遗忘
• 准备新版本时挑选要做什么
• 整理和清理过时的需求

核心功能:

• 📝 收集：追问痛点、频率、绕过方式，确认理解
• 📂 归类：检查合并可能，判断状态（可以做了/需要想想/先放着）
• 🧹 整理：定期清理过时条目，升档想清楚的条目
• 🎯 筛选：基于频率/可绕过性/ROI 分析候选需求
• ✅ 归档：标记已完成，更新依赖关系

触发方式:

我想做个功能
记个需求
整理一下需求池
下一个版本做什么
/backlog-manager

🗺️ 项目目录地图构建器 (Project Map Builder)

描述: 为指定目录范围生成或增量更新高信噪比的目录说明文档

适用场景:

• 需要生成项目目录概览
• 需要代码仓结构说明文档
• 想要更新已有的 PROJECT_MAP.md
• 需要理解项目文件夹结构

核心功能:

• 📋 智能范围确认：必须先询问要扫描的文件夹，不默认全仓扫描
• 🔍 关键文件识别：自动识别入口、配置、服务线程等关键文件
• 📝 增量更新：已有 PROJECT_MAP.md 时仅做增量更新，不全量重写
• 🎯 多目录支持：支持单目录或多目录（可合并或分别生成）
• ⚠️ 明确标注：不确定的地方显式标注"假设"或"未确认"

触发方式:

生成项目地图
更新 PROJECT_MAP
给我生成目录说明
/project-map-builder

📅 版本规划助手 (Version Planner)

描述: 将产品需求拆解成可执行的渐进式版本规划（V0.1 MVP → V1.0）

适用场景:

• 需要拆解产品版本
• 想要规划 MVP 最小可行产品
• 需要分阶段实现功能
• 不确定优先级和开发顺序

核心功能:

• 🎯 价值优先：优先解决用户最痛的点，而非技术最难的点
• 🚀 快速验证：MVP 设计为 2-3 天能跑通，避免过度设计
• 📋 明确边界：每个版本明确"做什么"和"不做什么"
• 🔄 渐进式交付：每个版本都能独立使用，不依赖后续版本
• ✅ 可测量：每个版本有清晰的验证点和完成标准

触发方式:

拆版本
版本规划
MVP怎么做
分阶段实现
/version-planner

✍️ 写作助手 (Writing Assistant)

描述: 根据观点清晰度自动选择最优写作路线，覆盖从思维挖掘到成稿的完整流程

适用场景:

• 想写文章但思路不清晰
• 需要梳理选题和核心观点
• 需要组织文章框架和逻辑
• 从零散想法到完整文章

核心功能:

• 🔍 智能诊断：快速判断观点清晰度，选择最优路线
• 💭 思维挖掘：观点模糊时，通过引导式对话倒出零散想法
• 🎯 选题确定：从洞察中提炼核心选题和灵魂句
• 📐 框架打磨：组织文章逻辑结构，确保表达清晰
• ✍️ 内容产出：根据框架写成完整文章，保持用户风格

触发方式:

我想写XX
帮我梳理选题
怎么形成框架
给我组织思路
/writing-assistant

📝 周报写作助手 (Weekly Report)

描述: 帮助用户梳理周报，按照完整逻辑展示工作价值和边界

适用场景:

• 需要整理一周的工作内容
• 想要清晰展示工作价值和成果
• 需要说明遇到的问题和挑战
• 梳理下周工作重点

核心功能:

• 📋 素材收集：引导式对话，收集本周工作内容
• 🗂️ 分类整理：根据角色灵活选择合适的模块分类
• 🔍 信息补充：追问背景、结果、价值、状态和下一步
• ✅ 讨论调整：确认表述习惯和逻辑完整性
• 📄 文档输出：生成结构清晰的周报文档

触发方式:

写周报
周报
梳理周报
整理工作
/weekly-report

🎯 优先级判断助手 (Priority Judge)

描述: 从混沌的待办事项中判断优先级，确定现在该做什么

适用场景:

• 有很多事要做，不知道从哪开始
• 想快速理清楚今天/本周该做什么
• 需要基于客观标准判断优先级
• 避免在没想清楚的事情上浪费时间

核心功能:

• 📝 收集待办：记录所有要做的事情
• 🔍 状态询问：了解每件事的清晰度和deadline
• ⚖️ 优先级判断：基于清晰度和deadline做决策
• 🎯 聚焦行动：每次只聚焦1-2件最重要的事
• 📋 文档化：生成优先级清单文档

触发方式:

我有很多事要做
帮我理一下
排个优先级
今天该做什么
我要盘一下
/priority-judge

🤝 思考拍档 (Thinking Partner)

描述: 陪你从混沌中理清局面，锁定核心问题，拆解卡点，共创解法，落地行动

适用场景:

• 面对复杂问题不知道从哪入手
• 想法很多但理不清主次
• 需要有人陪你一起想清楚
• 避免陷入细节迷失方向

核心功能:

• 📝 信息获取：引导式提问，看清全局
• 🎯 锁定核心问题：从一堆问题中找到最关键的那个
• 🔍 拆解卡点：层层追问，找到真正的根因
• 💡 共创解法：基于你的想法补充和修正
• ✅ 落地计划：把讨论结论变成可执行的行动

触发方式:

我现在很乱，帮我理一下
这事我不知道怎么办
陪我想想这个问题
/thinking-partner

🎨 UI 样式修改助手 (UI Design)

描述: UI 样式修改协作流程。通过结构化流程减少沟通偏差，避免浪费 token

适用场景:

• 需要修改页面样式、布局
• 调整间距、颜色、组件搭配
• UI 细节优化和微调
• 避免猜测式改代码

核心功能:

• 📸 截图定位：用截图确认当前状态
• 📐 现状描述：ASCII 画出当前布局
• 🎯 方案选择：提供 2-3 个可视化方案
• 🛠️ 最小改动：只改选定方案涉及的部分
• 🔄 微调迭代：执行具体的小修改

触发方式:

改一下这个页面的布局
调整这里的样式
这个间距不太对
/ui-design

🔮 终局愿景探索 (Vision Exploration)

描述: 帮你把模糊想法推演到最远的未来形态，输出 4~6 种维度完全不同的终局 HTML 原型

适用场景:

• 有个想法，想看看它最终能长成什么样
• 需要发散思维，探索一个 idea 的多种可能性
• 想在 PRD 之前先看到终局愿景
• 需要投资人级别的视觉原型

核心功能:

• 🔍 价值挖掘：反复追问"为什么"，找到需求背后的核心诉求
• 🎯 动机发现：识别真实用户使用场景
• 🔄 演进推导：从最小可行版本推导自然演进路径
• 🎨 终局渲染：输出 4~6 个自包含 HTML 原型，每个代表不同维度的终局形态
• 📊 对比分析：生成演进路径图和终局对比表

触发方式:

探索一下这个想法的终局
帮我看看这个功能最终能长成什么
/vision-exploration

🏷️ 产品命名助手 (Product Naming)

描述: 通过结构化协作流程，从产品本质出发推导名字，避免拍脑袋起名

适用场景:

• 需要给产品/项目/模块起名字
• 不想拍脑袋，想从品牌策略角度思考命名
• 需要竞品命名调研支撑决策
• 起了名字但不确定好不好

核心功能:

• 🔍 灵魂挖掘：理解产品本质、用户画像、品牌气质
• ⚡ 约束提取：提炼核心语素，识别诉求之间的张力
• 🛤️ 路线发散：推导 2~4 条不同的命名策略方向
• 🏷️ 竞品验证：搜索同赛道产品命名规律，用数据支撑建议
• 📌 最终确认：输出名称 + Slogan + 决策依据的完整结论

触发方式:

帮我起个名字
给项目取名
产品命名
/product-naming

🔧 Issue 协作处理 (Issue Triage)

描述: GitHub Issue 分诊与回复工作流，系统化分析问题、定位根因、撰写专业回复

适用场景:

• 收到 GitHub Issue 需要分析和回复
• 不确定 Issue 是 bug、架构局限还是功能请求
• 需要量化修复成本来做决策
• 想写出专业、诚恳的用户回复

核心功能:

• 🔍 代码诊断：沿调用链完整追踪，找到真正根因
• 🏷️ 问题定性：区分 Bug / 架构局限 / 功能请求 / 使用疑问
• ⚖️ 做不做决策：从改动范围、影响面、测试条件、变通方案四维评估
• ✍️ 起草回复：三层结构——解释意图、说明影响、交代计划
• 📮 发布：发评论、打标签、记录需求

触发方式:

帮我处理这个 Issue
分析一下这个 bug 报告
/issue-triage

🔍 GitHub 开源项目搜索助手 (GitHub Repo Search)

描述: 帮助用户搜索和筛选 GitHub 开源项目，输出结构化推荐报告

适用场景:

• 想找某个方向的开源项目
• 需要对比多个同类项目
• 想了解技术栈选型方案
• 寻找可直接使用或二次开发的工具

核心功能:

• 📝 需求收敛：确认主题、数量、排序模式、目标形态
• 🔍 检索词拆解：5-10 组 query，覆盖同义词、场景词、技术词
• 🏷️ 仓库分类：框架层/应用层/记忆层/MCP层/目录清单层/垂直场景层/方法论层
• 📊 质量精炼：综合权重排序（相关性/场景适用性/活跃度/工程成熟度）
• 📋 结构化报告：可理解、可比较、可决策、可直接行动的候选仓库列表

触发方式:

帮我找开源项目
搜一下GitHub上有什么
找找XX方向的仓库
开源项目推荐
/github-search

🧠 记忆系统初始化 (Memory Init)

描述: 在当前目录下一键部署记忆系统，生成 CLAUDE.md、MEMORY.md 和 memory/ 目录

适用场景:

• 新项目需要搭建记忆系统
• 希望 Claude 在多次对话间保持上下文连续性
• 需要长期记忆 + 每日记忆的分层管理
• 想让 Claude 像懂业务的同事一样协作

核心功能:

• 🔍 现有检查：自动检测已有文件，避免覆盖
• 📝 交互收集：引导式收集角色、用途、风格偏好
• 🤖 人设生成：根据用户信息定制 CLAUDE.md（人设 + 对话规则 + 记忆协议）
• 📋 记忆模板：生成 MEMORY.md 长期记忆初始模板
• 📁 目录创建：创建 memory/ 每日记忆目录
• ⚡ Quick 模式：支持快速初始化，用默认模板直接生成

触发方式:

初始化记忆
搭建记忆
memory init
/memory-init
/memory-init quick

🔍 多视角深度分析 (Multi-Perspective Analysis)

描述: 用多个 Sub-agent 分别扮演不同思维框架的顾问，对同一份素材做独立分析，然后交叉汇总共识与分歧，产出结构化诊断报告

适用场景:

• 个人/创作者月度复盘，想要多角度诊断
• 产品策略需要不同思维框架的碰撞
• 商业决策前想看清盲区和分歧
• 内容增长方向需要多维度验证

核心功能:

• 🧠 10 个内置视角：Dan Sullivan、马斯克、张小龙、MrBeast、芒格、蒂尔、乔布斯、贝佐斯、张一鸣、任正非
• 🔀 并行分析：所有 Sub-agent 同时启动，完全独立，互不可见
• 🔍 交叉汇总：找共识（高可信度）、找分歧（高价值洞察）、找独家观点（盲区）
• 📝 自动保存：每个视角独立报告 + 汇总报告，自动写入文件
• 🎯 自定义视角：支持通过 WebSearch 临时构建任意思想家/框架

触发方式:

帮我多视角分析
多维分析
用多个视角看看
帮我诊断一下
/multi-perspective-analysis

🗂️ 文件夹整理助手 (Organize)

描述: 帮助用户梳理混乱的文件夹，按归属分类、统一命名、清理冗余，让目录结构一目了然

适用场景:

• 根目录散落很多文件，找东西困难
• 子目录命名风格不统一，看着难受
• 工具产生的缓存/日志/临时文件堆积
• 怀疑有重复文件但不确定

核心功能:

• 🔍 扫描全貌：罗列文件、目录、隐藏项，识别散落与命名不齐
• 🏷️ 按归属分类：客户/内部/个人 优先于 文档/代码/设计 的格式分类
• 📋 制定方案：输出目标结构 + 操作清单（新建/移动/重命名/删除/不动）
• ✅ 确认后执行：必须用户确认才动手，删除项尤其谨慎
• 🛡️ 已有结构不破坏：组织良好的子目录保持不动
• 🔐 隐藏文件谨慎：只清理明确是缓存/日志的，配置目录不动
• 🧮 重复文件先验证：md5 比对确认后才判定为重复

触发方式:

整理文件
文件夹好乱
归档/清理目录
/organize [可选路径]

🤖 复杂长程任务自主执行 (Auto Task) (Beta)

描述: 复杂长程任务的自主执行流程。AI 自己拆解、自己执行、自己校验，1-2 小时无人值守跑长任务

适用场景:

• 任务复杂或模糊，需要拆解才能跑
• 答案不在脑子里，AI 要跑工具/调研/分析才能得出
• 时间跨度长（多步、多工具、可能要 sub-agent）
• 希望 AI 自驱，少打扰

核心功能:

• 📋 任务确认：先重述理解再开始，对齐边界和产出形态
• 🗂️ 任务队列：TASK.md 作为单一真源，每跑一批主动校验
• ⚡ 分批执行：自由调度 WebSearch / Bash / Sub-agent / MCP / 并行
• 🔁 周期校验：立场跟着证据走，不是补充原立场
• 📣 触发式汇报：默认不打扰，只在关键节点停下找用户
• 🧪 诚实声明：测试/评估类任务标注数据来源、是否盲测、潜在污染

触发方式:

帮我搞清楚 X
评估一下这事可不可行
帮我跑一次调研
你自己跑别打扰我
/auto-task

⚠️ 测试版：当前为 Beta，欢迎使用并反馈问题。

🍎 macOS 产品设计专家 (macOS Product Design) (Beta)

描述: 根据需求描述输出符合 macOS 原生风格的 HTML/CSS 设计稿，可直接在浏览器中预览

适用场景:

• 需要快速画出符合 macOS 原生风格的产品原型
• 想要可直接预览的 HTML 设计稿
• 需要遵循统一设计系统的多页面输出
• 暗色主题为主的桌面端界面设计

核心功能:

• 📐 设计系统对齐：基于 macos-design-system.md 的 CSS 变量与组件模板
• 🤖 Sub-agent 出稿：独立 sub-agent 严格按设计系统输出完整 HTML
• 🌑 暗色优先：默认暗色主题，所有交互元素带 hover 状态
• 📁 版本化输出：写入 design/ 目录，支持 v1 / v2 多版本迭代
• 🔄 主对话迭代：出稿后直接在主对话里改细节，无需重启 sub-agent

触发方式:

帮我设计一个界面
做个页面
产品设计 / UI 设计
画个原型
/macos-product-design

⚠️ 测试版：当前为 Beta，欢迎使用并反馈问题。

🧪 PRD 自动化测试闭环 (PRD Auto Test Loop) (Beta)

描述: PRD 驱动的自动化测试编排技能。把每版 PRD 的测试计划、AI 自测与自修复、测试报告标准化落地

适用场景:

• 每个 PRD 版本要发布前都需要测试闭环
• 想把 Unit / Integration / E2E 分层做规范
• 需要划分自动化与人工验证的边界
• 想要版本化的 TEST_PLAN / TEST_REPORT 文档

核心功能:

• 📋 范围锁定：从 PRD 提取版本目标、测试范围、非范围、验收标准
• 🧱 分层映射：验收标准映射到 Unit / Integration / E2E，标记 A/H/A+H
• 📂 版本化目录：自动化测试/<version>/ 下生成 TEST_PLAN.md、TEST_REPORT.md
• 🔁 主回归 + 增量：跨版本维护主回归基线，仅新增 PRD 差异用例
• 🤖 AI 自测自修复：固定顺序 Unit → Integration → E2E，失败按根因重跑（最多 3 轮）
• 🚦 发布门禁：报告输出 PASS / FAIL 结论与人工补测项

触发方式:

帮我做这个版本的测试计划
跑一下 PRD 的自动化测试
生成 TEST_PLAN / TEST_REPORT
/prd-auto-test-loop

⚠️ 测试版：当前为 Beta，欢迎使用并反馈问题。

📑 PRD 测试用例一体化 (PRD Test Writer) (Beta)

描述: PRD + 可执行测试用例双文档一体化协作。以伙伴模式与用户共同写并迭代，理解需求后先自主读代码再落笔

适用场景:

• 既要写 PRD，又要把可执行测试用例一起冻结
• 需要用例严格落地真实代码（每条断言标代码依据）
• 任务类用例要写死轮数、每轮内容与期望，杜绝"测个长任务"式泛化
• 想要 MD 事实源 + 给人查阅的 review HTML 严格 1:1

核心功能:

• 📖 故事驱动 PRD：用户旅程地图 + 自包含用户故事 + ASCII 线框图 + Mermaid
• 🔍 先读死代码：写文档前 grep 落地真实行为，每条断言标「代码依据 文件:行」
• 🧪 原子测试用例：一条用例 = 一个原子验证点，13 字段标准，任务类写死轮数/每轮内容/期望
• 🧾 双文档一体：PRD-MD + 测试用例-MD（给 AI 事实源）+ 两份 review HTML（给人查阅）
• 🕵️ 对抗校核：≥3 个无共享上下文 sub-agent 交叉审，反假绿/反同义反复
• 🧊 冻结与台账：定稿标 Frozen+commit，维护 PRD_REGISTRY

触发方式:

帮我写 PRD 和测试用例
梳理需求文档 + 测试方案
/prd-test-writer

⚠️ 测试版：当前为 Beta，欢迎使用并反馈问题。

📖 输出工程 (Readable Output)

描述: 把"想清楚再输出"的通用逻辑封装成思考清单 + 强制 4 问挖掘，专门用来产可读性高的 HTML 长文

适用场景:

• 做复盘 / 总结 / 汇总 / 整理一堆素材
• 做教程 / 学习指南 / 上手文档
• 把某事讲清楚，整理成可读 HTML
• 任何 > 500 字、要给人读的结构化阅读内容

核心功能:

• 🎯 强制 4 问：受众 / 终点 / 长度 / 风格 + 侧重，一次 AskUserQuestion 问清
• 🧭 Backward design：先定"读完拿什么可复用工具"（对照表/判断框架/决策路径/口诀），再倒推内容
• 📐 6 阶段框架：定终点 → 抓核心（≤3 个）→ 选主结构（不叠加）→ 写（TL;DR + 论点句 + 反例 + 出口）→ 自检
• 📋 配置确认条：HTML 顶部自动写入「本次配置」，扫一眼知道按什么基准产的
• 🚫 内置反模式清单：列菜单不给推荐 / TL;DR 埋中间 / 5 点并列 / 主视觉叠加等高频陷阱
• 🧪 自检 4 问：不懂的人看得懂吗 / 一周后能复述吗 / 砍一半还说得通吗 / 每段在一屏内吗
• 📦 写完自动 open 在浏览器打开 review

触发方式:

做个复盘
汇总一下
总结这堆
整理成 HTML
做个教程 / 学习指南
把 X 讲清楚
/readable-output

🧭 系统化学习材料生成器 (System Study)

描述: 给一个新领域/技术/概念，AI 自主调研、搭体系，并产出带案例、工程化、争议焦点和盲区清单的 HTML 学习材料

适用场景:

• 想系统学习一个新领域，而不是只要结论
• 需要看清概念骨架、前沿趋势、真实案例和工具生态
• 想把调研沉淀成可复用的学习材料
• 需要把争议点、判断倾向和未知盲区单独拎出来

核心功能:

• 🎯 题目收敛：先确认学习切片和视角，避免题目跑偏
• 🧪 调研计划：明确每个 sub-agent 要查的来源和问题
• ⚡ 并行调研：A-D 覆盖骨架、前沿、真物案例、生态工程化，E 必做争议与盲区
• 🧱 体系合成：主线程整合调研笔记，形成可确认的大纲
• 📄 HTML 出稿：输出带侧栏导航、章节折叠、争议评分和盲区清单的单文件学习材料

触发方式:

我要学 X
帮我系统拆解 X
我想吃透 X 这个领域
给我整理 X 的全貌
深度调研 X 给我系统讲讲
/system-study

🚀 一键推送 GitHub (Git Push)

描述: 一键推送项目到 GitHub，覆盖首次推送、日常更新、版本发布三种模式

适用场景:

• 新项目需要从零推送到 GitHub
• 日常改了文件需要快速 commit + push
• 需要打 tag 和创建 Release 发版本
• 担心把大文件或敏感内容推上去

核心功能:

• 🔍 环境检查：自动检测 git、gh CLI、登录状态、用户配置
• 📂 大文件扫描：10MB+ 提醒、50MB+ 警告、100MB+ 强制排除
• 🔒 敏感内容扫描：公开仓库自动检测密钥、证书、AI 记忆文件
• 📝 .gitignore 生成：基于扫描结果智能生成，不覆盖已有规则
• 🚀 三种模式：首次推送 / 日常更新 / 版本发布（tag + Release）
• ⚡ 日常更新轻量化：只检查新增大文件，快进快出

触发方式:

推到GitHub
推送到GitHub
推一下
发版本
打release
/git-push

📡 案例雷达 (Case Radar)

描述: 给一个新东西（新工具/新概念/新生态），扫一遍生态找好玩的真实案例，重点抓"真物"（截图/源码/演示）而不是 GitHub 主页，输出可浏览的 HTML 案例集

适用场景:

• 想了解一个新生态/新工具，但不想读三手 SEO 文
• 需要一手物料（截图、源码片段、demo）来支撑研究或写作
• 扫某个概念的"市面上有哪些玩法"
• 不适合：已有明确目标的深度调研（用 long-research）/ 直接写文章（用 writing-assistant）

核心功能:

• 🔭 4 阶段流程：scan（多渠道扫信源）→ recon（真物侦察）→ capture（拉真物）→ embed（出 HTML）
• 🎯 信源分级：一手源 / 二手优质 / 三手中文，砍掉 SEO 垃圾和 awesome 集市
• 🧭 7 种"真物栖息地"判别 patterns（reference/recon-heuristics.md），灵活而非 checklist 化
• 🛠️ 工具组合：curl 直接资源 + gh API 拉源码 + agent-browser 截全页
• 📐 HTML 模板规范：5-7 张精读卡 + 普通索引层，避免"35 卡断崖式注意力低谷"
• 🚦 反过度工程化：截不到真物就降级，不硬塞装饰图；每一步可中断
• 🔁 Skill 自演化：发现新栖息地就更新 heuristics 文件

触发方式:

看看大家用 X 做了什么
扫一下 X 生态
市面上 X 有什么新玩法
给我看 X 的真物案例
/case-radar

📦 依赖：需要预装 agent-browser 和 gh CLI 已登录。

💝 Hermes 陪伴型人设生成器 (Hermes Persona Builder)

描述: 陪伴型 AI 人设生成与优化流程，通过结构化对话把一句模糊想法产出"有感情、聊久不掉、像真人"的第一人称人设文本

适用场景:

• 想给 Hermes Agent（或任意 AI 陪伴角色）捏一个能立住的人设
• 一句模糊想法（"JK 女友""年上男友""高冷御姐"）要落成可直接用的设定
• 现有人设"掉戏"/变助手腔，需要优化或换一版
• 支持女友/男友/各种气质，不只是甜妹

核心功能:

• 🎚️ 定调子：先定位（恋爱/全能/分身）+ 性别关系 + 气质大类（甜/暖/冷/野/稳），矛盾需求有收敛协议
• 🔍 多轮深问：顺着追问到牌子/口味/名字/数字的颗粒度，密度不达标不生成
• 🧱 形容词翻硬规则：正向命令 + 反向禁令（冷系/男友型必带），防掉人设
• 🛡️ 安全红线：年龄 ≥18、不涉露骨、身份≠能力、不写有害人格，产出前自检
• 💬 说话节奏：选"一句一句发"还是"整段说"，干活长内容自动不硬拆
• 📄 产出对齐 Hermes SOUL.md 模板，热加载即生效，支持迭代优化

触发方式:

帮我做个人设 / 捏个 AI 女友男友
给 Hermes 弄个角色
优化人设 / 换个人设
/hermes-persona-builder

🚀 快速开始

安装方式

使用 Codex 或 Claude Code CLI 安装

如果你使用 Codex 或 Claude Code CLI，直接跟 AI 说：

安装这个 GitHub 库：https://github.com/yunshu0909/yunshu_skillshub

AI 会自动帮你完成安装！

手动安装

将本仓库克隆到你的本地 Skills 目录：

# Claude Code 默认 Skills 目录通常是 ~/.claude/skills/
cd ~/.claude/skills/

# 克隆本仓库
git clone https://github.com/yunshu0909/yunshu_skillshub.git

或者，你也可以单独复制需要的 Skill 到你的 Skills 目录。

使用 Skills

在 Claude Code CLI 中，你可以通过以下方式使用：

# 使用配图助手
/image-assistant

# 使用思维挖掘助手
/thought-mining

# 使用 PRD 文档撰写助手
/prd-doc-writer

# 使用新功能设计探索
/design-exploration

# 使用需求变更工作流
/req-change-workflow

# 使用课程构建器
/lesson-builder

# 使用需求池管理
/backlog-manager

# 使用项目目录地图构建器
/project-map-builder

# 使用版本规划助手
/version-planner

# 使用写作助手
/writing-assistant

# 使用周报写作助手
/weekly-report

# 使用优先级判断助手
/priority-judge

# 使用思考拍档
/thinking-partner

# 使用 UI 样式修改助手
/ui-design

# 使用 GitHub 开源项目搜索助手
/github-search

# 使用终局愿景探索
/vision-exploration

# 使用产品命名助手
/product-naming

# 使用 Issue 协作处理
/issue-triage

# 使用记忆系统初始化
/memory-init

# 使用多视角深度分析
/multi-perspective-analysis

# 使用一键推送 GitHub
/git-push

# 使用文件夹整理助手
/organize

# 使用复杂长程任务自主执行 (Beta)
/auto-task

# 使用 macOS 产品设计专家 (Beta)
/macos-product-design

# 使用 PRD 自动化测试闭环 (Beta)
/prd-auto-test-loop

# 使用 PRD 测试用例一体化 (Beta)
/prd-test-writer

# 使用输出工程
/readable-output

# 使用系统化学习材料生成器
/system-study

# 使用案例雷达
/case-radar

或者直接在对话中描述你的需求，相关 Skill 会自动触发。

📚 查看使用示例

想了解每个 Skill 的具体使用方法？查看 使用示例文档，里面包含了详细的使用场景和预期输出。

📂 项目结构

.
├── README.md                    # 项目说明文档
├── LICENSE                      # MIT 许可证
├── CHANGELOG.md                 # 更新日志
├── EXAMPLES.md                  # 使用示例
├── docs/                        # 补充文档与测试材料
│   └── length-test/             # CodePal 长度测试 HTML 样例
├── image-assistant/             # 配图助手
│   ├── SKILL.md                # Skill 定义文件
│   ├── stages/                 # 各阶段详细说明
│   ├── templates/              # 风格模板和配图模板
│   ├── examples/               # 使用示例
│   └── scripts/                # 批量生图脚本
├── thought-mining/              # 思维挖掘助手
│   ├── SKILL.md                # Skill 定义文件
│   ├── stages/                 # 各阶段详细说明
│   ├── templates/              # 模板文件
│   └── examples/               # 使用示例
├── prd-doc-writer/             # PRD 文档撰写助手
│   ├── SKILL.md               # Skill 定义文件
│   ├── assets/                # 模板资源
│   └── references/            # 参考文档和示例
├── design-exploration/         # 新功能设计探索
│   ├── SKILL.md               # Skill 定义文件
│   └── templates/             # 需求总结模板
├── req-change-workflow/        # 需求变更工作流
│   ├── SKILL.md               # Skill 定义文件
│   ├── references/            # 模板和清单
│   └── scripts/               # 辅助脚本
├── lesson-builder/             # 课程构建器
│   └── SKILL.md               # Skill 定义文件
├── backlog-manager/            # 需求池管理
│   └── SKILL.md               # Skill 定义文件
├── project-map-builder/        # 项目目录地图构建器
│   └── SKILL.md               # Skill 定义文件
├── version-planner/            # 版本规划助手
│   └── SKILL.md               # Skill 定义文件
├── writing-assistant/          # 写作助手
│   ├── SKILL.md               # Skill 定义文件
│   ├── stages/                # 各阶段详细说明
│   └── templates/             # 模板文件
├── weekly-report/             # 周报写作助手
│   └── SKILL.md              # Skill 定义文件
├── priority-judge/            # 优先级判断助手
│   └── SKILL.md              # Skill 定义文件
├── thinking-partner/          # 思考拍档
│   └── SKILL.md              # Skill 定义文件
├── ui-design/                 # UI 样式修改助手
│   └── SKILL.md              # Skill 定义文件
├── github-repo-search/        # GitHub 开源项目搜索助手
│   └── SKILL.md              # Skill 定义文件
├── vision-exploration/        # 终局愿景探索
│   └── SKILL.md              # Skill 定义文件
├── product-naming/            # 产品命名助手
│   └── SKILL.md              # Skill 定义文件
├── issue-triage/              # Issue 协作处理
│   └── SKILL.md              # Skill 定义文件
├── multi-perspective-analysis/ # 多视角深度分析
│   ├── SKILL.md              # Skill 定义文件
│   └── reference/            # 5 个视角的分析框架语料
├── memory-init/               # 记忆系统初始化
│   └── SKILL.md              # Skill 定义文件
├── git-push/                  # 一键推送 GitHub
│   └── SKILL.md              # Skill 定义文件
├── organize/                  # 文件夹整理助手
│   └── SKILL.md              # Skill 定义文件
├── auto-task/                 # 复杂长程任务自主执行 (Beta)
│   └── SKILL.md              # Skill 定义文件
├── macos-product-design/      # macOS 产品设计专家 (Beta)
│   ├── SKILL.md              # Skill 定义文件
│   └── reference/            # macOS 设计系统规范
├── prd-auto-test-loop/        # PRD 自动化测试闭环 (Beta)
│   ├── SKILL.md              # Skill 定义文件
│   ├── agents/               # 测试 Agent
│   ├── references/           # 测试封装指南
│   └── scripts/              # 辅助脚本
├── prd-test-writer/           # PRD 测试用例一体化 (Beta)
│   ├── SKILL.md              # Skill 定义文件
│   ├── assets/               # PRD/用例模板与 review HTML 模板
│   ├── references/           # 样板用例、线框图/Mermaid 示例、校验规范
│   └── samples/              # 示例 review HTML
├── readable-output/           # 输出工程
│   └── SKILL.md              # Skill 定义文件
├── system-study/              # 系统化学习材料生成器
│   ├── SKILL.md              # Skill 定义文件
│   └── reference/            # 阶段流程、sub-agent 模板、HTML 规范与反模式
├── case-radar/                # 案例雷达
│   ├── SKILL.md              # Skill 定义文件
│   └── reference/            # recon heuristics / HTML 模板规范 / 反 anti-pattern 清单
└── hermes-persona-builder/    # Hermes 陪伴型人设生成器
    └── SKILL.md              # Skill 定义文件

🤝 贡献

欢迎提交 Issue 和 Pull Request！如果你有任何建议或发现了 bug，请随时告诉我。

📄 许可证

本项目采用 MIT License 开源。

📧 联系方式

如有问题或建议，欢迎通过 GitHub Issues 联系。

English

📖 Introduction

A carefully crafted collection of Claude Code Skills designed to boost efficiency in software development and product management. Each skill has been battle-tested to help you work more effectively in your daily tasks.

📊 Skills Overview

Currently includes 30 Skills across 3 categories:

💻 AI Programming (15)

✍️ Thinking & Writing (10)

💼 Daily Office & Productivity (5)

✨ Included Skills

🎨 Image Assistant

Description: Convert article/module content into unified-style, text-minimal, highly readable 16:9 infographic prompts

Use Cases:

• Need illustrations for articles but don't know how to design
• PPT, posters, or social media graphics need a unified style
• Too much text, want more engaging and readable visual presentation
• Need to batch generate illustration prompts

Core Features:

• 📋 Requirement Clarification: Extract content, scenario, audience, and text density preferences
• 🗂️ Illustration Planning: Split content, define image list (how many/what each explains)
• ✍️ Copy Finalization: Word-by-word finalization of "what text goes on the image" (Copy Spec)
• 🎯 Prompt Packaging: Generate copy-ready image generation prompts, support batch generation
• 🔄 Iterative Refinement: Reduce text, change metaphors, improve readability based on feedback

Trigger:

Make an image for this content / how many images?
Give me two image generation prompts
Too much text, make it more engaging and readable
/image

🧠 Thought Mining

Description: Helps you extract scattered thoughts from your mind, record them, and organize them into articles through conversational guidance

Use Cases:

• Want to write an article but thoughts are unclear
• Have many scattered ideas that need organizing
• Need to extract core insights from chaotic thinking

Core Features:

• 📝 Thought Mining: Guided conversation to help you articulate and record ideas
• 🎯 Topic Selection: Find core viewpoints from insights
• ✅ Validation: Verify understanding through web search
• ✍️ Writing Assistance: Logic checking, text polishing, extracting key phrases
• 🔍 Final Review: Comprehensive check before publishing

Trigger:

I want to write an article about XX
Help me organize my thoughts
/thought-mining

📋 PRD Doc Writer

Description: A story-driven approach to writing and iteratively refining PRD/requirement documents

Use Cases:

• Need to write product requirement documents
• Want to organize requirements using user stories
• Need to reduce requirement ambiguity with diagrams

Core Features:

• 🗺️ User Journey Map: Build macro business processes
• 📖 Story-based Requirements: Each feature is a complete user story
• 🎨 ASCII Wireframes: Visualize page layouts
• 📊 Mermaid Diagrams: Flow charts/state diagrams/sequence diagrams
• ✅ Staged Confirmation: Ensure consensus at every step

Trigger:

Help me write a PRD
Organize requirement document
/prd-doc-writer

🎨 Design Exploration

Description: New feature design exploration workflow. From fuzzy ideas to deliverable design reference documents as input for the PRD phase

Use Cases:

• Have a fuzzy idea for a new feature/module
• Need to batch explore multiple design options
• Need complete coverage of page states (normal/exception/boundary)
• Need to produce design mockups as development reference

Core Features:

• 📝 Requirement Convergence: Define what to do, what not to do, and boundaries
• 🔍 Technical Research: Understand external data constraints and technical feasibility
• 🎨 ASCII Batch Exploration: Generate 5-8 options at once for selection
• 📐 HTML Mockups: Produce mockups based on selected option
• 📊 Full State Coverage: Normal/loading/empty/error/boundary states
• 📋 Interaction Rule Table: Frontend development reference

Trigger:

I want to build a new feature
Help me design this module
Create a design proposal
/design-exploration

🔄 Requirement Change Workflow

Description: Standardize the requirement change process to avoid chaos and code breakage when modifying requirements

Use Cases:

• Need to modify existing feature requirements
• Frequently encounter unexpected bugs when changing requirements
• Need a reliable change validation process
• Especially suitable for complex projects like Chrome extensions

Core Features:

• 📝 Requirement Clarification: Lock change scope and acceptance criteria
• 🔍 Current Baseline: Confirm current behavior from code
• ⚠️ Impact Assessment: Assess risks and change scope
• 🎯 Design Proposal: Propose new design and get approval
• 🛠️ Minimal Implementation: Small-scale, localized code changes
• ✅ Regression Testing: Fixed validation checklist
• 📚 Documentation Maintenance: Decision log and documentation updates

Trigger:

Change requirement/requirement modification
Adjust interaction/change feature
/req-change-workflow

📚 Lesson Builder

Description: A discussion-driven approach to quickly complete course outlines and teaching materials

Use Cases:

• Need to quickly prepare a lesson
• Have clear ideas that need to be organized into documents
• Need to iterate on existing course outlines
• Preparing training or teaching content

Core Features:

• 💭 Co-create Outline: Extract ideas through discussion, form clear course framework
• 📖 Write Materials: Create complete teaching materials based on outline
• 🎯 Framework First: Confirm framework before details to avoid rework
• ⚡ Rapid Iteration: Supports both quick co-creation and strict confirmation modes
• 📋 Minimal Documentation: Only produce what you need (outline/materials/supplements)

Trigger:

Prepare lesson
Make teaching materials/prepare course
/lesson-builder

📦 Backlog Manager

Description: Backlog management. Throw in ideas/pain points anytime, AI handles追问, organizing, merging, and archiving. Assists with selection when starting a new version. Pain-driven, no premature scheduling

Use Cases:

• Record product ideas and feature needs anytime
• Manage scattered requirements to avoid forgetting
• Select what to do when preparing a new version
• Organize and clean up outdated requirements

Core Features:

• 📝 Collection: Ask about pain points, frequency, workarounds to confirm understanding
• 📂 Categorization: Check merge possibilities, judge status (ready to do/needs thinking/on hold)
• 🧹 Organization: Regularly clean outdated items, upgrade well-thought-out items
• 🎯 Selection: Analyze candidates based on frequency/workaround feasibility/ROI
• ✅ Archiving: Mark completed, update dependencies

Trigger:

I want to build a feature
Record a requirement
Organize the backlog
What to do in next version
/backlog-manager

🗺️ Project Map Builder

Description: Generate or incrementally update high signal-to-noise directory documentation for specified directories

Use Cases:

• Need to generate project directory overview
• Need codebase structure documentation
• Want to update existing PROJECT_MAP.md
• Need to understand project folder structure

Core Features:

• 📋 Smart Scope Confirmation: Must ask for folders to scan, never defaults to full repository
• 🔍 Key File Recognition: Automatically identifies entry points, configs, service threads, and other critical files
• 📝 Incremental Updates: Only does incremental updates when PROJECT_MAP.md exists, never full rewrites
• 🎯 Multi-directory Support: Supports single or multiple directories (merge or generate separately)
• ⚠️ Explicit Annotation: Clearly marks uncertain areas as "assumption" or "unconfirmed"

Trigger:

Generate project map
Update PROJECT_MAP
Generate directory documentation
/project-map-builder

📅 Version Planner

Description: Decompose product requirements into executable progressive version roadmap (V0.1 MVP → V1.0)

Use Cases:

• Need to break down product versions
• Want to plan MVP (Minimum Viable Product)
• Need to implement features in stages
• Uncertain about priorities and development order

Core Features:

• 🎯 Value First: Prioritize solving users' biggest pain points, not the hardest technical challenges
• 🚀 Quick Validation: MVP designed to work in 2-3 days, avoiding over-engineering
• 📋 Clear Boundaries: Each version clearly defines "what to do" and "what not to do"
• 🔄 Progressive Delivery: Each version can be used independently, not dependent on future versions
• ✅ Measurable: Each version has clear validation points and completion criteria

Trigger:

Break down versions
Version planning
How to do MVP
Implement in stages
/version-planner

✍️ Writing Assistant

Description: Automatically selects optimal writing route based on viewpoint clarity, covering complete flow from thought mining to final draft

Use Cases:

• Want to write an article but thoughts are unclear
• Need to organize topic and core viewpoint
• Need to organize article framework and logic
• From scattered ideas to complete article

Core Features:

• 🔍 Smart Diagnosis: Quickly assess viewpoint clarity and choose optimal route
• 💭 Thought Mining: When viewpoint is fuzzy, use guided dialogue to extract scattered ideas
• 🎯 Topic Selection: Extract core topic and key message from insights
• 📐 Framework Polishing: Organize article logic structure, ensure clear expression
• ✍️ Content Production: Write complete article based on framework, maintain user's style

Trigger:

I want to write XX
Help me organize my topic
How to form a framework
Help me organize my thoughts
/writing-assistant

📝 Weekly Report

Description: Helps users organize weekly reports with complete logic to showcase work value and boundaries

Use Cases:

• Need to organize a week's work content
• Want to clearly demonstrate work value and achievements
• Need to explain problems and challenges encountered
• Organize next week's priorities

Core Features:

• 📋 Material Collection: Guided dialogue to collect weekly work content
• 🗂️ Categorization: Flexibly choose appropriate module classification based on role
• 🔍 Information Supplement: Ask about background, results, value, status, and next steps
• ✅ Discussion & Adjustment: Confirm expression habits and logical completeness
• 📄 Document Output: Generate well-structured weekly report document

Trigger:

Write weekly report
Weekly report
Organize weekly report
Organize work
/weekly-report

🎯 Priority Judge

Description: Determine priorities from chaotic to-do items and decide what to do now

Use Cases:

• Have many things to do, don't know where to start
• Want to quickly figure out what to do today/this week
• Need to judge priorities based on objective criteria
• Avoid wasting time on things not thought through

Core Features:

• 📝 Collect To-dos: Record all things to do
• 🔍 Status Inquiry: Understand clarity and deadline of each item
• ⚖️ Priority Judgment: Make decisions based on clarity and deadline
• 🎯 Focus Action: Focus on only 1-2 most important things each time
• 📋 Documentation: Generate priority list document

Trigger:

I have many things to do
Help me sort it out
Prioritize
What should I do today
Let me review
/priority-judge

🤝 Thinking Partner

Description: Accompany you to clarify the situation from chaos, lock core problems, break down bottlenecks, co-create solutions, and land actions

Use Cases:

• Facing complex problems and don't know where to start
• Have many ideas but can't sort out priorities
• Need someone to think through things with you
• Avoid getting lost in details and losing direction

Core Features:

• 📝 Information Gathering: Guided questioning to see the big picture
• 🎯 Lock Core Problem: Find the most critical one from a bunch of problems
• 🔍 Break Down Bottlenecks: Layer-by-layer questioning to find the real root cause
• 💡 Co-create Solutions: Supplement and refine based on your ideas
• ✅ Action Plan: Turn discussion conclusions into executable actions

Trigger:

I'm confused now, help me sort it out
I don't know what to do about this
Think through this problem with me
/thinking-partner

🎨 UI Design Assistant

Description: UI style modification collaboration workflow. Reduce communication deviation and avoid wasting tokens through structured process

Use Cases:

• Need to modify page styles and layouts
• Adjust spacing, colors, component combinations
• UI detail optimization and fine-tuning
• Avoid guessing when changing code

Core Features:

• 📸 Screenshot Positioning: Confirm current state with screenshots
• 📐 Status Description: Draw current layout with ASCII
• 🎯 Solution Selection: Provide 2-3 visualized solutions
• 🛠️ Minimal Changes: Only change parts involved in selected solution
• 🔄 Iterative Refinement: Execute specific small modifications

Trigger:

Modify the layout of this page
Adjust the style here
The spacing is not quite right
/ui-design

🔮 Vision Exploration

Description: Explore the ultimate potential of your ideas, outputting 4-6 dimensionally different end-state HTML prototypes

Use Cases:

• Have an idea and want to see what it could ultimately become
• Need divergent thinking to explore multiple possibilities
• Want to see end-state visions before writing a PRD
• Need investor-ready visual prototypes

Core Features:

• 🔍 Value Extraction: Repeatedly ask "why" to uncover the core need behind the requirement
• 🎯 Motivation Discovery: Identify real user scenarios
• 🔄 Evolution Derivation: Derive natural evolution path from minimum viable version
• 🎨 End-State Rendering: Output 4-6 self-contained HTML prototypes, each representing a different dimensional end-state
• 📊 Comparison Analysis: Generate evolution pathway diagram and end-state comparison table

Trigger:

Explore the end-state of this idea
What could this feature ultimately become
/vision-exploration

🏷️ Product Naming

Description: Structured collaborative workflow to derive product names from brand essence, avoiding random name generation

Use Cases:

• Need to name a product/project/module
• Want to approach naming from a brand strategy perspective
• Need competitive naming research to support decisions
• Have a name but unsure if it's good

Core Features:

• 🔍 Soul Mining: Understand product essence, user persona, brand personality
• ⚡ Constraint Extraction: Extract core semantic elements, identify tension between requirements
• 🛤️ Route Divergence: Derive 2-4 different naming strategy directions
• 🏷️ Competitive Validation: Research naming patterns in the same space, back suggestions with data
• 📌 Final Confirmation: Output complete conclusion with name + slogan + decision rationale

Trigger:

Help me name this
Name this project
Product naming
/product-naming

🔧 Issue Triage

Description: GitHub Issue triage and response workflow for systematic analysis, root cause identification, and professional replies

Use Cases:

• Received a GitHub Issue that needs analysis and response
• Unsure if an Issue is a bug, architectural limitation, or feature request
• Need to quantify fix costs for decision-making
• Want to write professional, honest user responses

Core Features:

• 🔍 Code Diagnosis: Trace the full call chain to find the real root cause
• 🏷️ Issue Classification: Distinguish Bug / Architectural Limitation / Feature Request / Usage Question
• ⚖️ Do/Don't Decision: Evaluate across 4 dimensions — change scope, impact area, test conditions, workaround feasibility
• ✍️ Draft Reply: Three-layer structure — explain intent, describe impact, outline plans
• 📮 Publish: Post comment, apply labels, record requirements

Trigger:

Help me handle this Issue
Analyze this bug report
/issue-triage

🔍 GitHub Repo Search

Description: Help users search and filter GitHub open source projects, output structured recommendation reports

Use Cases:

• Looking for open source projects in a specific direction
• Need to compare multiple similar projects
• Want to understand technology stack selection options
• Looking for tools ready to use or for secondary development

Core Features:

• 📝 Requirement Convergence: Confirm topic, quantity, sorting mode, target form
• 🔍 Query Breakdown: 5-10 query groups covering synonyms, scenario words, technical terms
• 🏷️ Repository Classification: Framework/Application/Memory/MCP/Catalog/Vertical/Methodology layers
• 📊 Quality Refinement: Comprehensive weighted ranking (relevance/applicability/activity/maturity)
• 📋 Structured Report: Understandable, comparable, decision-ready, actionable candidate repository list

Trigger:

Help me find open source projects
Search GitHub for XX
Looking for repositories in XX direction
Open source project recommendations
/github-search

🧠 Memory Init

Description: One-click deployment of a memory system in the current directory, generating CLAUDE.md, MEMORY.md, and memory/ directory

Use Cases:

• New project needs a memory system setup
• Want Claude to maintain context continuity across conversations
• Need layered management of long-term memory + daily memory
• Want Claude to collaborate like a knowledgeable colleague

Core Features:

• 🔍 Existing Check: Auto-detect existing files to avoid overwriting
• 📝 Interactive Collection: Guided collection of role, usage, and style preferences
• 🤖 Persona Generation: Customize CLAUDE.md based on user info (persona + conversation rules + memory protocol)
• 📋 Memory Template: Generate initial MEMORY.md long-term memory template
• 📁 Directory Creation: Create memory/ daily memory directory
• ⚡ Quick Mode: Support quick initialization with default templates

Trigger:

Initialize memory
Set up memory
memory init
/memory-init
/memory-init quick

🔍 Multi-Perspective Analysis

Description: Use multiple Sub-agents as advisors with different thinking frameworks, independently analyze the same material, then cross-reference to find consensus, divergence, and produce a structured diagnostic report

Use Cases:

• Personal/creator monthly review with multi-angle diagnosis
• Product strategy needs collision of different thinking frameworks
• Want to identify blind spots and divergence before business decisions
• Content growth direction needs multi-dimensional validation

Core Features:

• 🧠 10 Built-in Perspectives: Dan Sullivan, Elon Musk, Zhang Xiaolong, MrBeast, Charlie Munger, Peter Thiel, Steve Jobs, Jeff Bezos, Zhang Yiming, Ren Zhengfei
• 🔀 Parallel Analysis: All Sub-agents launch simultaneously, fully independent, invisible to each other
• 🔍 Cross-Synthesis: Find consensus (high confidence), divergence (high-value insights), and unique observations (blind spots)
• 📝 Auto-Save: Individual perspective reports + summary report, automatically written to files
• 🎯 Custom Perspectives: Support building any thinker/framework via WebSearch on the fly

Trigger:

Multi-perspective analysis
Analyze from multiple angles
Help me diagnose this
/multi-perspective-analysis

🗂️ Organize

Description: Help users organize messy folders by classifying by ownership, unifying naming, and cleaning up redundancy

Use Cases:

• Many files scattered in root directory, hard to find things
• Subdirectory naming styles are inconsistent and look messy
• Tool-generated cache/log/temp files have piled up
• Suspect duplicate files but not sure

Core Features:

• 🔍 Full Scan: List files, directories, hidden items; identify scattered and inconsistent naming
• 🏷️ Classify by Ownership: Client/internal/personal takes priority over format-based grouping
• 📋 Plan Proposal: Output target structure + operation list (create/move/rename/delete/keep)
• ✅ Confirm Before Acting: Must get user approval before executing, especially for deletions
• 🛡️ Preserve Existing Structure: Well-organized subdirectories remain untouched
• 🔐 Cautious with Hidden Files: Only clean clear cache/log items, leave config dirs alone
• 🧮 Verify Duplicates First: Use md5 to confirm before judging files as duplicates

Trigger:

Organize files
The folder is messy
Archive / clean up
/organize [optional path]

🤖 Auto Task (Beta)

Description: Autonomous execution flow for complex, long-horizon tasks. The AI breaks them down, executes, and self-verifies — running for 1-2 hours unattended

Use Cases:

• Task is complex or fuzzy and needs decomposition to run
• The answer is not in the user's head — AI must run tools/research/analysis
• Long time horizon (multi-step, multi-tool, possibly sub-agents)
• User wants AI to drive autonomously with minimal interruption

Core Features:

• 📋 Task Confirmation: Restate understanding to align scope and output form before starting
• 🗂️ Task Queue: TASK.md as single source of truth, validated after each batch
• ⚡ Batch Execution: Freely orchestrate WebSearch / Bash / Sub-agent / MCP / parallel tools
• 🔁 Periodic Validation: Position follows evidence — not appended to original stance
• 📣 Trigger-based Reporting: Default no interruption, only stop at key nodes
• 🧪 Honesty Declaration: Test/eval tasks must declare data source, blind-test status, contamination

Trigger:

Help me figure out X
Evaluate whether this is feasible
Run a research for me
Run on your own, don't bother me
/auto-task

⚠️ Beta: This is a beta release. Feedback is welcome.

🍎 macOS Product Design (Beta)

Description: Output macOS-native HTML/CSS design mockups based on requirement descriptions, directly previewable in browser

Use Cases:

• Need to quickly produce macOS-native style product prototypes
• Want directly-previewable HTML mockups
• Need multi-page output following a unified design system
• Dark-theme-first desktop UI design

Core Features:

• 📐 Design System Alignment: Based on macos-design-system.md CSS variables and component templates
• 🤖 Sub-agent Output: Independent sub-agent strictly outputs complete HTML per design system
• 🌑 Dark First: Default dark theme, all interactive elements have hover states
• 📁 Versioned Output: Written to design/ directory, supports v1 / v2 iteration
• 🔄 Main-thread Iteration: After initial output, iterate directly in main conversation without re-launching sub-agent

Trigger:

Help me design a UI
Make a page
Product design / UI design
Sketch a prototype
/macos-product-design

⚠️ Beta: This is a beta release. Feedback is welcome.

🧪 PRD Auto Test Loop (Beta)

Description: PRD-driven automated testing orchestration. Standardize per-PRD test plans, AI self-testing/self-fixing, and test reports

Use Cases:

• Each PRD release needs a closed-loop test cycle
• Want to standardize Unit / Integration / E2E layering
• Need to delineate boundaries between automation and manual verification
• Want versioned TEST_PLAN / TEST_REPORT documents

Core Features:

• 📋 Scope Locking: Extract version goal, scope, non-scope, and acceptance criteria from PRD
• 🧱 Layer Mapping: Map acceptance criteria to Unit / Integration / E2E, tag A/H/A+H
• 📂 Versioned Directory: Generate TEST_PLAN.md, TEST_REPORT.md under 自动化测试/<version>/
• 🔁 Main Regression + Increment: Maintain main regression baseline across versions, only add PRD-specific cases
• 🤖 AI Self-test & Self-fix: Fixed order Unit → Integration → E2E, retry up to 3 rounds on root-cause fix
• 🚦 Release Gating: Reports output PASS / FAIL conclusion plus manual coverage list

Trigger:

Make a test plan for this version
Run automated tests for this PRD
Generate TEST_PLAN / TEST_REPORT
/prd-auto-test-loop

⚠️ Beta: This is a beta release. Feedback is welcome.

📑 PRD Test Writer (Beta)

Description: Co-author a PRD and executable test cases as paired documents. Partner-mode collaboration — reads the real code before writing

Use Cases:

• Need to freeze a PRD and its executable test cases together
• Test cases must be grounded in real code (every assertion cites file:line)
• Task-type cases must hard-code rounds, per-round input, and expected output
• Want an MD source-of-truth plus a strictly 1:1 review HTML for humans

Core Features:

• 📖 Story-driven PRD: user journey map + self-contained stories + ASCII wireframes + Mermaid
• 🔍 Code-first: grep the real behavior before writing, every assertion tagged with file:line
• 🧪 Atomic test cases: one case = one atomic check, 13-field standard, task cases hard-coded
• 🧾 Paired docs: PRD-MD + test-cases-MD (AI source) + two review HTMLs (for humans)
• 🕵️ Adversarial review: ≥3 sub-agents with no shared context cross-check against false-green
• 🧊 Freeze & registry: mark Frozen+commit on sign-off, maintain PRD_REGISTRY

Trigger:

Write a PRD with test cases for me
Draft requirement doc + test plan
/prd-test-writer

⚠️ Beta: This is a beta release. Feedback is welcome.

📖 Readable Output

Description: Wraps the universal "think before you output" logic into a checklist + forced 4-question intake, dedicated to producing highly readable HTML long-form content

Use Cases:

• Retrospectives / summaries / aggregations / organizing piles of material
• Tutorials / learning guides / onboarding docs
• Explaining something clearly as a readable HTML doc
• Any > 500-word structured reading content meant for human readers

Core Features:

• 🎯 Forced 4-Question Intake: Audience / Endpoint / Length / Style + Focus, asked in one AskUserQuestion call
• 🧭 Backward Design: First define "what reusable tool the reader leaves with" (comparison table / decision framework / decision path / mnemonic), then work backward
• 📐 6-Stage Framework: Define endpoint → Pick ≤3 core points → Choose primary structure (no stacking) → Write (TL;DR + topic sentences + counter-examples + exit) → Self-check
• 📋 Config Banner: HTML automatically inserts a "This run's config" bar at the top so the user can see the basis at a glance
• 🚫 Built-in Anti-pattern List: Listing menus without recommendation / TL;DR buried mid-doc / 5 parallel points / stacked primary visuals — common traps to avoid
• 🧪 4 Self-check Questions: Can a beginner follow it / Can the reader recite the core a week later / Does it still work at half the length / Does each section fit on one screen
• 📦 Auto open in browser after writing for review

Trigger:

Make a retrospective
Summarize this
Aggregate this batch
Organize into HTML
Make a tutorial / learning guide
Explain X clearly
/readable-output

🧭 System Study

Description: Given a new field, technology, or concept, the AI researches independently, builds a learning system, and outputs an HTML learning material with cases, engineering practice, controversies, and blind spots

Use Cases:

• Want to systematically learn a new field, not just get a conclusion
• Need to understand the concept map, frontier trends, real artifacts, and tool ecosystem
• Want research to become a reusable learning material
• Need controversies, directional judgments, and unresolved blind spots separated clearly

Core Features:

• 🎯 Topic Narrowing: Confirm learning slices and perspective first to avoid researching the wrong target
• 🧪 Research Plan: Define sources and questions for each sub-agent before execution
• ⚡ Parallel Research: A-D cover skeleton, frontier, real cases, and ecosystem; E is mandatory for controversy and blind spots
• 🧱 System Synthesis: Main thread integrates notes into a confirmable outline
• 📄 HTML Output: Single-file learning material with sidebar navigation, collapsible sections, controversy scores, and a blind-spot list

Trigger:

I want to learn X
Help me systematically break down X
I want to really understand X
Give me the full picture of X
Deep research X and explain it systematically
/system-study

🚀 Git Push

Description: One-click push projects to GitHub. Auto-scan large files, generate .gitignore, init Git, create repo and push. Supports first push, daily updates, and version release modes

Use Cases:

• New project needs to be pushed to GitHub from scratch
• Daily file changes need quick commit + push
• Need to tag and create Release for versioning
• Worried about pushing large files or sensitive content

Core Features:

• 🔍 Environment Check: Auto-detect git, gh CLI, login status, user config
• 📂 Large File Scan: 10MB+ prompt, 50MB+ warning, 100MB+ forced exclusion
• 🔒 Sensitive Content Scan: Auto-detect keys, certificates, AI memory files for public repos
• 📝 .gitignore Generation: Smart generation based on scan results, preserves existing rules
• 🚀 Three Modes: First push / Daily update / Version release (tag + Release)
• ⚡ Lightweight Daily Updates: Only check new large files, quick in and out

Trigger:

Push to GitHub
Push it
Release version
Create release
/git-push

📡 Case Radar

Description: Given a new thing (new tool/concept/ecosystem), scan the ecosystem for interesting real-world cases. Focus on grabbing "real artifacts" (screenshots/source/demos) instead of GitHub homepages, and output a browsable HTML case gallery

Use Cases:

• Want to understand a new ecosystem/tool without reading third-hand SEO content
• Need first-hand artifacts (screenshots, source snippets, demos) to support research or writing
• Survey "what's actually being built" around a concept
• Not for: deep research with a clear target (use long-research) / writing articles directly (use writing-assistant)

Core Features:

• 🔭 4-stage flow: scan (multi-channel source sweep) → recon (real-artifact reconnaissance) → capture (pull artifacts) → embed (output HTML)
• 🎯 Source tiering: first-party / quality second-party / curated third-party — strip out SEO junk and meta-awesome lists
• 🧭 7 "real-artifact habitat" patterns (reference/recon-heuristics.md), used flexibly rather than as a checklist
• 🛠️ Tool combo: curl for direct assets + gh API for source snippets + agent-browser for full-page screenshots
• 📐 HTML template spec: 5-7 deep-read cards + index layer for the rest, avoiding the "35-card attention cliff"
• 🚦 Anti-over-engineering: degrade to plain card if no real artifact found, never stuff decorative images; every step is interruptible
• 🔁 Self-evolving skill: discover a new habitat, update the heuristics file

Trigger:

Show me what people are doing with X
Scan the X ecosystem
What new takes are out there for X
Show me real artifacts for X
/case-radar

📦 Dependencies: requires agent-browser installed and authenticated gh CLI.

💝 Hermes Persona Builder

Description: A structured-dialogue flow that turns a fuzzy idea into a first-person AI companion persona that feels emotional, stays in character over long chats, and reads like a real person — aligned with Hermes's SOUL.md format

Use Cases:

• Craft a persona that holds up for a Hermes Agent (or any AI companion role)
• Turn a fuzzy idea ("JK girlfriend", "older boyfriend", "cool onee-san") into a ready-to-use setting
• Optimize or swap a persona that keeps "breaking character" or slipping into assistant tone
• Supports girlfriend/boyfriend and any temperament, not just sweet types

Core Features:

• 🎚️ Set the tone: positioning + gender/relationship + temperament (sweet/warm/cold/wild/steady), with a protocol to converge conflicting asks
• 🔍 Multi-round probing: drill down to brand/flavor/name/number granularity; won't generate until density passes the gate
• 🧱 Translate adjectives into hard rules: positive commands + negative bans (required for cold/boyfriend types) to prevent character drift
• 🛡️ Safety red lines: age ≥18, no explicit content, identity ≠ capability, no harmful personas, with a pre-delivery self-check
• 💬 Speech rhythm: choose "message-by-message" vs "full paragraphs"; long/work content auto-stays intact
• 📄 Output aligns with the Hermes SOUL.md template, hot-reloads instantly, supports iteration

Trigger:

Help me build a persona / make an AI girlfriend or boyfriend
Set up a character for Hermes
Optimize my persona / swap to a new one
/hermes-persona-builder

🚀 Quick Start

Installation

Using Codex or Claude Code CLI

If you're using Codex or Claude Code CLI, simply tell the AI:

Install this GitHub repository: https://github.com/yourusername/云舒的Skills搭子们

The AI will automatically install it for you!

Manual Installation

Clone this repository to your local Skills directory:

# Claude Code default Skills directory is usually ~/.claude/skills/
cd ~/.claude/skills/

# Clone this repository
git clone https://github.com/yourusername/云舒的Skills搭子们.git

Alternatively, you can copy individual Skills you need to your Skills directory.

Usage

In Claude Code CLI, you can use them by:

# Use Image Assistant
/image-assistant

# Use Thought Mining
/thought-mining

# Use PRD Doc Writer
/prd-doc-writer

# Use Design Exploration
/design-exploration

# Use Requirement Change Workflow
/req-change-workflow

# Use Lesson Builder
/lesson-builder

# Use Backlog Manager
/backlog-manager

# Use Project Map Builder
/project-map-builder

# Use Version Planner
/version-planner

# Use Writing Assistant
/writing-assistant

# Use Weekly Report
/weekly-report

# Use Priority Judge
/priority-judge

# Use Thinking Partner
/thinking-partner

# Use UI Design Assistant
/ui-design

# Use GitHub Repo Search
/github-search

# Use Vision Exploration
/vision-exploration

# Use Product Naming
/product-naming

# Use Issue Triage
/issue-triage

# Use Memory Init
/memory-init

# Use Multi-Perspective Analysis
/multi-perspective-analysis

# Use Git Push
/git-push

# Use Organize
/organize

# Use Auto Task (Beta)
/auto-task

# Use macOS Product Design (Beta)
/macos-product-design

# Use PRD Auto Test Loop (Beta)
/prd-auto-test-loop

# Use PRD Test Writer (Beta)
/prd-test-writer

# Use Readable Output
/readable-output

# Use System Study
/system-study

# Use Case Radar
/case-radar

Or simply describe your needs in conversation, and the relevant Skill will trigger automatically.

📚 Check Usage Examples

Want to learn how to use each Skill? Check out the Usage Examples for detailed scenarios and expected outputs.

📂 Project Structure

.
├── README.md                    # Project documentation
├── LICENSE                      # MIT License
├── CHANGELOG.md                 # Changelog
├── EXAMPLES.md                  # Usage examples
├── docs/                        # Supplementary docs and test materials
│   └── length-test/             # CodePal length-test HTML samples
├── image-assistant/             # Image Assistant
│   ├── SKILL.md                # Skill definition file
│   ├── stages/                 # Detailed stage descriptions
│   ├── templates/              # Style templates and layout templates
│   ├── examples/               # Usage examples
│   └── scripts/                # Batch image generation scripts
├── thought-mining/              # Thought Mining Assistant
│   ├── SKILL.md                # Skill definition file
│   ├── stages/                 # Detailed stage descriptions
│   ├── templates/              # Template files
│   └── examples/               # Usage examples
├── prd-doc-writer/             # PRD Doc Writer
│   ├── SKILL.md               # Skill definition file
│   ├── assets/                # Template resources
│   └── references/            # Reference docs and examples
├── design-exploration/         # Design Exploration
│   ├── SKILL.md               # Skill definition file
│   └── templates/             # Requirements summary template
├── req-change-workflow/        # Requirement Change Workflow
│   ├── SKILL.md               # Skill definition file
│   ├── references/            # Templates and checklists
│   └── scripts/               # Helper scripts
├── lesson-builder/             # Lesson Builder
│   └── SKILL.md               # Skill definition file
├── backlog-manager/            # Backlog Manager
│   └── SKILL.md               # Skill definition file
├── project-map-builder/        # Project Map Builder
│   └── SKILL.md               # Skill definition file
├── version-planner/            # Version Planner
│   └── SKILL.md               # Skill definition file
├── writing-assistant/          # Writing Assistant
│   ├── SKILL.md               # Skill definition file
│   ├── stages/                # Detailed stage descriptions
│   └── templates/             # Template files
├── weekly-report/             # Weekly Report
│   └── SKILL.md              # Skill definition file
├── priority-judge/            # Priority Judge
│   └── SKILL.md              # Skill definition file
├── thinking-partner/          # Thinking Partner
│   └── SKILL.md              # Skill definition file
├── ui-design/                 # UI Design Assistant
│   └── SKILL.md              # Skill definition file
├── github-repo-search/        # GitHub Repo Search
│   └── SKILL.md              # Skill definition file
├── vision-exploration/        # Vision Exploration
│   └── SKILL.md              # Skill definition file
├── product-naming/            # Product Naming
│   └── SKILL.md              # Skill definition file
├── issue-triage/              # Issue Triage
│   └── SKILL.md              # Skill definition file
├── multi-perspective-analysis/ # Multi-Perspective Analysis
│   ├── SKILL.md              # Skill definition file
│   └── reference/            # Analysis framework materials for 5 perspectives
├── memory-init/               # Memory Init
│   └── SKILL.md              # Skill definition file
├── git-push/                  # Git Push
│   └── SKILL.md              # Skill definition file
├── organize/                  # Organize
│   └── SKILL.md              # Skill definition file
├── auto-task/                 # Auto Task (Beta)
│   └── SKILL.md              # Skill definition file
├── macos-product-design/      # macOS Product Design (Beta)
│   ├── SKILL.md              # Skill definition file
│   └── reference/            # macOS design system spec
├── prd-auto-test-loop/        # PRD Auto Test Loop (Beta)
│   ├── SKILL.md              # Skill definition file
│   ├── agents/               # Test agents
│   ├── references/           # Test packaging guide
│   └── scripts/              # Helper scripts
├── prd-test-writer/           # PRD Test Writer (Beta)
│   ├── SKILL.md              # Skill definition file
│   ├── assets/               # PRD/test templates & review HTML templates
│   ├── references/           # Sample cases, wireframe/Mermaid examples, check spec
│   └── samples/              # Example review HTML
├── readable-output/           # Readable Output
│   └── SKILL.md              # Skill definition file
├── system-study/              # System Study
│   ├── SKILL.md              # Skill definition file
│   └── reference/            # Stage workflow, sub-agent templates, HTML spec, and anti-patterns
└── case-radar/                # Case Radar
    ├── SKILL.md              # Skill definition file
    └── reference/            # Recon heuristics / HTML template spec / anti-pattern checklist

🤝 Contributing

Issues and Pull Requests are welcome! If you have any suggestions or find bugs, please feel free to let me know.

📄 License

This project is open source under the MIT License.

📧 Contact

For questions or suggestions, please contact via GitHub Issues.

Made with ❤️ by Yunshu