Skip to content
Back to search

code-reader-skills

🧠 Cognitive science-based Claude Skills for deep code understanding | 基于认知科学的源代码深度理解工具

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/notlate-cn--code-reader-skills/badge.svg)](/listings/notlate-cn--code-reader-skills)

HTML

<a href="/listings/notlate-cn--code-reader-skills"><img src="/api/listings/notlate-cn--code-reader-skills/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

**方式一:复制到 Claude 目录(推荐)**

# 中文版
mkdir -p ~/.claude/skills
cp -r code-reader-zh ~/.claude/skills

# 英文版
mkdir -p ~/.claude/skills
cp -r code-reader-en ~/.claude/skills

**方式二:直接在对话中使用**

# 在分析代码前,直接粘贴 skill 内容
cat code-reader-zh/SKILL.md | pbcopy  # macOS
# 或者在对话中输入 @文件路径

**方式三:使用 API 的开发者**

如果你通过 Claude API 调用,可以将 skill 内容作为 System Prompt:

import anthropic

# 读取 skill 内容(中文版)
with open('code-reader-zh/SKILL.md', 'r', encoding='utf-8') as f:
    skill_content = f.read()

client = anthropic.Anthropic()
message = client.messages.create(
    model="claude-sonnet-4-20250514",  # 或使用最新可用模型
    system=skill_content,  # 将 skill 作为系统提示
    messages=[
        {"role": "user", "content": "深入分析这段代码:\n\n[你的代码]"}
    ]
)

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

README.md · README · 629 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. # 🧠 Code Deep Understanding Analyzer
  2. <div align="center">
  3. **基于认知科学的源代码深度理解工具 | Cognitive Science-Based Code Analysis Tool**
  4. [![License](https://img.shields.io/badge/license-MIT-blue.svg)](LICENSE)
  5. [![Version](https://img.shields.io/badge/version-2.4.0-green.svg)](https://github.com/notlate-cn/code-reader-skills/releases)
  6. [![Claude Skills](https://img.shields.io/badge/Claude-Skills-orange.svg)](https://claude.ai)
  7. [![Language](https://img.shields.io/badge/language-中文%20%7C%20English-red.svg)](#)
  8. [English](#english) | [中文](#中文)
  9. </div>
  10. ---
  11. ## 中文
  12. ### 📖 项目简介
  13. 这是一套专业的 Claude Skills,帮助开发者**真正理解**源代码,而不只是"看懂"。基于认知科学研究,融合精细询问法、自我解释测试和应用迁移验证,确保深度学习而非产生流畅幻觉。
  14. **核心理念:** 理解为什么 (WHY) > 知道是什么 (WHAT)
  15. ### ✨ 核心特性
  16. - 🎯 **精细询问法** - 每个概念强制回答 3 个 WHY
  17. - 🧪 **自我解释测试** - 验证真实理解程度
  18. - 🔗 **概念网络构建** - 建立知识连接,而非孤立记忆
  19. - 🚀 **应用迁移测试** - 检验能否在不同场景应用
  20. - 📚 **学术研究支撑** - 基于 Dunlosky, Chi, Karpicke 等认知科学研究
  21. - ⚡ **三种模式** - Quick/Standard/Deep 满足不同需求
  22. - 🤖 **智能并行** - Deep Mode 大型项目自动使用并行处理
  23. ### 📦 文件结构
  24. ```
  25. code-reader-skills/
  26. ├── code-reader-zh/ # 中文版 🇨🇳
  27. │ ├── SKILL.md # Skill 源文件
  28. │ └── code-reader-zh.skill # Skill 包
  29. │
  30. ├── code-reader-en/ # 英文版 🇺🇸
  31. │ ├── SKILL.md # Skill source file
  32. │ └── code-reader-en.skill # Skill package
  33. │
  34. ├── README.md # 项目说明(本文件)
  35. └── LICENSE # 开源许可证
  36. ```
  37. ### 🚀 快速开始
  38. #### 1. 下载 Skill 文件
  39. ```bash
  40. # 克隆仓库
  41. git clone https://github.com/notlate-cn/code-reader-skills.git
  42. cd code-reader-skills
  43. ```
  44. #### 2. 三种分析模式
  45. 支持三种分析深度,根据场景自动选择:
  46. | 模式 | 耗时 | 适用场景 | 触发词示例 |
  47. |------|------|---------|-----------|
  48. | **Quick** | 5-10 分钟 | 快速浏览、代码审查 | "快速看一下"、"这段代码干嘛的" |
  49. | **Standard** | 15-20 分钟 | 学习理解、技术调研 ⭐ | "分析一下"、"帮我理解"、"解释一下" |
  50. | **Deep** | 30+ 分钟 | 深度掌握、大型项目 🚀 | "彻底分析"、"完全掌握"、"深入研究" |
  51. **默认使用 Standard Mode**
  52. **🚀 Deep Mode 内部智能策略:代码 ≤ 2000 行使用渐进式生成,代码 > 2000 行自动启用并行处理。**
  53. #### 3. 安装 Skill
  54. **方式一:复制到 Claude 目录(推荐)**
  55. ```bash
  56. # 中文版
  57. mkdir -p ~/.claude/skills
  58. cp -r code-reader-zh ~/.claude/skills
  59. # 英文版
  60. mkdir -p ~/.claude/skills
  61. cp -r code-reader-en ~/.claude/skills
  62. ```
  63. **方式二:直接在对话中使用**
  64. ```bash
  65. # 在分析代码前,直接粘贴 skill 内容
  66. cat code-reader-zh/SKILL.md | pbcopy # macOS
  67. # 或者在对话中输入 @文件路径
  68. ```
  69. **方式三:使用 API 的开发者**
  70. 如果你通过 Claude API 调用,可以将 skill 内容作为 System Prompt:
  71. ```python
  72. import anthropic
  73. # 读取 skill 内容(中文版)
  74. with open('code-reader-zh/SKILL.md', 'r', encoding='utf-8') as f:
  75. skill_content = f.read()
  76. client = anthropic.Anthropic()
  77. message = client.messages.create(
  78. model="claude-sonnet-4-20250514", # 或使用最新可用模型
  79. system=skill_content, # 将 skill 作为系统提示
  80. messages=[
  81. {"role": "user", "content": "深入分析这段代码:\n\n[你的代码]"}
  82. ]
  83. )
  84. ```
  85. #### 4. 开始使用
  86. **Quick Mode 触发示例:**
  87. ```
  88. 你:/code-reader-v2-cn 快速分析这段代码
  89. [上传或粘贴代码]
  90. ```
  91. **Standard Mode 触发示例:**
  92. ```
  93. 你:/code-reader-v2-cn 深入分析这段代码,我想理解它的设计原理
  94. [上传或粘贴代码]
  95. ```
  96. **Deep Mode 触发示例:**
  97. ```
  98. 你:/code-reader-v2-cn 我需要彻底掌握这个算法,准备面试用
  99. [上传或粘贴代码]
  100. ```
  101. **大型项目自动触发并行处理:**
  102. ```
  103. 你:/code-reader-v2-cn 彻底分析这个项目
  104. [上传或粘贴代码,系统检测到 >2000 行时自动使用并行处理]
  105. ```
  106. ### 📊 输出结构
  107. #### Quick Mode(快速概览)
  108. 1. 👀 快速概览
  109. 2. 📝 功能说明
  110. 3. 📐 核心算法/设计
  111. 4. 💻 关键代码段
  112. 5. 📦 依赖关系
  113. 6. 🎯 快速使用示例
  114. #### Standard Mode(标准理解)⭐
  115. 1. 🧪 理解验证状态
  116. 2. 👀 快速概览
  117. 3. 🎯 背景与动机(精细询问)
  118. 4. 📌 核心概念说明
  119. 5. 📐 算法与理论
  120. 6. 🏗️ 设计模式
  121. 7. 💻 关键代码深度解析(6A 核心片段识别 + 6B 六节深度解读)
  122. 8. 📦 依赖与使用示例
  123. #### Deep Mode(完全掌握)🚀
  124. [包含 Standard Mode 所有内容,加上]
  125. - 🔗 概念网络图
  126. - 🔍 **核心片段识别**(6A):从大型代码中筛选 3-7 个最值得深读的片段
  127. - 💻 **六节深度解读**(6B):每片段按「整体作用 → 核心逻辑 → 逐行解析 → 设计点 → 三组示例 → 改进建议」解读
  128. - 🧪 **测试用例分析**(如有测试,选 3-5 个典型用例说明代码功能)
  129. - 🚀 应用迁移场景(至少 2 个)
  130. - ✅ 质量验证清单
  131. - 🤖 **智能策略选择**:代码 ≤ 2000 行使用渐进式生成,代码 > 2000 行自动启用并行处理
  132. - 并行处理时:主协调 Agent + 并行子 Agents + 自动汇总
  133. - 确保每个章节都有充分深度
  134. ### 💡 使用示例
  135. #### 示例 1:学习 JWT 认证(Standard Mode)
  136. <details>
  137. <summary>点击查看完整示例</summary>
  138. **输入:**
  139. ```python
  140. def authenticate_user(username, password):
  141. user = db.find_user(username)
  142. if not user:
  143. return None
  144. if verify_password(password, user.password_hash):
  145. return generate_token(user.id)
  146. return None
  147. ```
  148. **输出(节选):**
  149. ```markdown
  150. ## 背景与动机分析
  151. **WHY 需要这段代码?**
  152. - 系统需要验证用户身份,保护未授权访问
  153. **WHY 选择 JWT 方案?**
  154. - 无状态认证,服务器不需要存储 Session
  155. **WHY 不用 Session?**
  156. - Session 需要服务器存储,不利于水平扩展
  157. ```
  158. </details>
  159. ### 🔬 研究基础
  160. 基于以下认知科学研究:
  161. - **[Dunlosky et al. (2013)](https://www.aft.org/ae/fall2013/dunlosky)** - 精细询问法显著优于被动阅读
  162. - **[Chi et al. (1994)](https://onlinelibrary.wiley.com/doi/10.1207/s15516709cog1803_3)** - 自我解释者获得正确心智模型的概率更高
  163. - **[Karpicke & Roediger (2008)](https://science.sciencemag.org/content/319/5865/966)** - 检索练习优于重复阅读 250%
  164. ### 🤝 贡献指南
  165. 欢迎贡献!以下是参与方式:
  166. 1. **报告问题** - 通过 [Issues](https://github.com/notlate-cn/code-reader-skills/issues) 反馈
  167. 2. **提出建议** - 分享你的使用体验和改进想法
  168. 3. **提交 PR** - 改进文档或添加新功能
  169. 4. **分享案例** - 展示你的使用案例
  170. ### ❓ 常见问题
  171. <details>
  172. <summary><strong>Q: 三种模式有什么区别?</strong></summary>
  173. **A:**
  174. - **Quick**:快速了解代码结构和功能,5-10 分钟
  175. - **Standard**:理解 WHY 和设计思路,15-20 分钟(推荐)
  176. - **Deep**:应用迁移测试 + 质量验证,30+ 分钟
  177. </details>
  178. <details>
  179. <summary><strong>Q: 支持哪些编程语言?</strong></summary>
  180. **A:** 支持所有主流语言:
  181. - Python, JavaScript, TypeScript, Java, C++, Go, Rust
  182. - 以及各种框架和库
  183. </details>
  184. <details>
  185. <summary><strong>Q: 如何判断我真正理解了?</strong></summary>
  186. **A:** 使用 Deep Mode 的"四能"测试:
  187. 1. ✅ 能否不看代码解释设计思路?
  188. 2. ✅ 能否独立实现类似功能?
  189. 3. ✅ 能否应用到不同场景?
  190. 4. ✅ 能否向他人清晰解释?
  191. </details>
  192. ### 📝 更新日志
  193. #### v2.4.0 (2026-03-10)
  194. **新增功能:**
  195. - 🔍 **核心片段识别(6A)**:大型项目先按优先级筛选 3-7 个最值得深读的片段,避免"全部浅析"
  196. - 💻 **六节深度解读模板(6B)**:每个核心片段按「整体作用 / 核心逻辑 / 逐行解析 / 设计点 / 三组对比示例 / 改进建议」结构化解读
  197. - 🤖 **子 Agent 两阶段指令**:关键代码解析子 Agent 强制执行片段识别 → 深度解读工作流
  198. **重构:**
  199. - 📦 SKILL.md 从 2371 行精简至 688 行(精简 71%),删除冗余示例和重复结构,保留全部执行逻辑
  200. - 🌐 中英文版同步更新,结构完全一致
  201. #### v2.3.0 (2026-02-07)
  202. **重构优化:**
  203. - 🔄 合并 Parallel Deep Mode 到 Deep Mode,作为内部策略
  204. - ⚡ 优化触发词系统,丰富各模式触发词
  205. - 🤖 Deep Mode 内部智能选择:≤2000 行渐进式,>2000 行自动并行
  206. - 📖 简化为三种模式:Quick/Standard/Deep
  207. - 🧪 新增测试用例反向理解功能(Step 6.5)
  208. **改进:**
  209. - 🎯 用户体验更友好,无需手动选择并行模式
  210. - 📝 更新中英文 SKILL.md 和 README 文档
  211. - 🔬 支持通过测试用例反向验证和深化理解
  212. - 🛠️ 支持 C++、MLIR/LLVM 等多种语言测试格式
  213. **解决问题:**
  214. - ✅ 解决模式过于冗余的问题
  215. - ✅ 自动策略选择降低使用门槛
  216. - ✅ 通过测试发现代码中隐藏的行为和边界条件
  217. #### v2.2.0 (2026-02-04)
  218. **新增功能:**
  219. - ✨ 场景化注释风格(场景 N / 步骤 N)
  220. - ✨ Token 优化策略(直接写入文件)
  221. - ✨ 渐进式生成(Deep Mode 专用)
  222. - ✨ 多语言注释支持规范
  223. **改进:**
  224. - 📦 优化文档结构,删除重复内容
  225. - 📖 统一章节深度自检标准
  226. - 🎯 完善执行流示例格式
  227. **优化:**
  228. - 🔧 文件大小:1500 → 1342 行 (-10.5%)
  229. - ⚡ 删除子 Agent 使用规范
  230. - 📝 合并重复的模式选择说明
  231. #### v2.1.0 (2026-01-31)
  232. **新增功能:**
  233. - ✨ Quick/Standard/Deep 三种分析模式
  234. - ✨ 智能模式选择机制
  235. - ✨ 优化输出结构
  236. **改进:**
  237. - 📦 独立中英文目录结构
  238. - 📖 更新 README 说明
  239. - 🎯 精简触发词
  240. ### 📄 许可证
  241. 本项目采用 [MIT License](LICENSE) 开源。
  242. ### 🌟 Star History
  243. 如果这个项目对你有帮助,请给个 Star ⭐
  244. [![Star History Chart](https://api.star-history.com/svg?repos=notlate-cn/code-reader-skills&type=Date)](https://star-history.com/#notlate-cn/code-reader-skills&Date)
  245. ### 📧 联系方式
  246. - 问题反馈:[GitHub Issues](https://github.com/notlate-cn/code-reader-skills/issues)
  247. - 讨论交流:[GitHub Discussions](https://github.com/notlate-cn/code-reader-skills/discussions)
  248. ---
  249. ## English
  250. ### 📖 Project Introduction
  251. A professional Claude Skills set that helps developers **truly understand** source code, not just "get it." Based on cognitive science research, integrating elaborative interrogation, self-explanation testing, and application transfer verification to ensure deep learning rather than fluency illusion.
  252. **Core Philosophy:** Understanding WHY > Knowing WHAT
  253. ### ✨ Key Features
  254. - 🎯 **Elaborative Interrogation** - Force answering 3 WHYs for each concept
  255. - 🧪 **Self-Explanation Test** - Verify true understanding
  256. - 🔗 **Concept Network Construction** - Build knowledge connections, not isolated memories
  257. - 🚀 **Application Transfer Test** - Examine if applicable in different scenarios
  258. - 📚 **Academic Research Support** - Based on Dunlosky, Chi, Karpicke's cognitive science research
  259. - 🌐 **Bilingual Support** - Complete Chinese and English versions
  260. - ⚡ **Four Modes** - Quick/Standard/Deep/Parallel Deep for different needs
  261. - 🤖 **Parallel Analysis** - Sub-agents process chapters in parallel, ensuring depth for large projects
  262. ### 📦 File Structure
  263. ```
  264. code-reader-skills/
  265. ├── code-reader-zh/ # Chinese version 🇨🇳
  266. │ ├── SKILL.md # Skill source file
  267. │ └── code-reader-zh.skill # Skill package
  268. │
  269. ├── code-reader-en/ # English version 🇺🇸
  270. │ ├── SKILL.md # Skill source file
  271. │ └── code-reader-en.skill # Skill package
  272. │
  273. ├── README.md # Project documentation (this file)
  274. └── LICENSE # Open source license
  275. ```
  276. ### 🚀 Quick Start
  277. #### 1. Download Skill Files
  278. ```bash
  279. # Clone repository
  280. git clone https://github.com/notlate-cn/code-reader-skills.git
  281. cd code-reader-skills
  282. ```
  283. #### 2. Three Analysis Modes
  284. | Mode | Duration | Use Case | Trigger Examples |
  285. |------|----------|----------|------------------|
  286. | **Quick** | 5-10 min | Quick browse, code review | "quick look", "what does this do" |
  287. | **Standard** | 15-20 min | Learning, research ⭐ | "analyze", "help me understand", "explain" |
  288. | **Deep** | 30+ min | Deep mastery, large projects 🚀 | "thoroughly analyze", "completely master", "in-depth research" |
  289. **Default: Standard Mode**
  290. **🚀 Deep Mode internal smart strategy: Code ≤ 2000 lines uses progressive generation, code > 2000 lines auto-enables parallel processing.**
  291. #### 3. Install Skill
  292. **Method 1: Copy to Claude directory (Recommended)**
  293. ```bash
  294. # English version
  295. mkdir -p ~/.claude/skills
  296. cp -r code-reader-en ~/.claude/skills
  297. # Chinese version
  298. mkdir -p ~/.claude/skills
  299. cp -r code-reader-zh ~/.claude/skills
  300. ```
  301. **Method 2: Use directly in conversation**
  302. ```bash
  303. # Paste skill content before analyzing code
  304. cat code-reader-en/SKILL.md | pbcopy # macOS
  305. # Or use @file-path in conversation
  306. ```
  307. **Method 3: For API Users**
  308. If you're using Claude API, pass skill content as System Prompt:
  309. ```python
  310. import anthropic
  311. # Read skill content (English version)
  312. with open('code-reader-en/SKILL.md', 'r', encoding='utf-8') as f:
  313. skill_content = f.read()
  314. client = anthropic.Anthropic()
  315. message = client.messages.create(
  316. model="claude-sonnet-4-20250514",
  317. system=skill_content, # Use skill as system prompt
  318. messages=[
  319. {"role": "user", "content": "Deeply analyze this code:\n\n[Your code]"}
  320. ]
  321. )
  322. ```
  323. #### 4. Getting Started
  324. **Quick Mode Trigger Example:**
  325. ```
  326. You: /code-reader-v2-en Quickly analyze this code
  327. [Upload or paste code]
  328. ```
  329. **Standard Mode Trigger Example:**
  330. ```
  331. You: /code-reader-v2-en Deeply analyze this code, I want to understand its design principles
  332. [Upload or paste code]
  333. ```
  334. **Deep Mode Trigger Example:**
  335. ```
  336. You: /code-reader-v2-en I need to thoroughly master this algorithm for interview preparation
  337. [Upload or paste code]
  338. ```
  339. **Parallel Deep Mode Trigger Example:**
  340. ```
  341. You: /code-reader-v2-en Thoroughly analyze this large project
  342. [Upload or paste code, system will auto-use parallel processing when >2000 lines detected]
  343. ```
  344. ### 📊 Output Structure
  345. #### Quick Mode
  346. 1. Quick Overview
  347. 2. Functionality Description
  348. 3. Core Algorithm/Design
  349. 4. Key Code Segments
  350. 5. Dependencies
  351. 6. Quick Usage Example
  352. #### Standard Mode ⭐
  353. 1. Understanding Verification Status
  354. 2. Quick Overview
  355. 3. Background & Motivation (Elaborative Interrogation)
  356. 4. Core Concepts
  357. 5. Algorithm & Theory
  358. 6. Design Patterns
  359. 7. Key Code Deep Analysis (6A snippet identification + 6B six-section deep-dive)
  360. 8. Dependencies & Usage Examples
  361. #### Deep Mode 🚀
  362. [All Standard Mode content, plus]
  363. - Concept Network Diagram
  364. - **Core Snippet Identification (6A)**: Select 3-7 most important snippets from large codebases
  365. - **Six-Section Deep-Dive (6B)**: Each snippet analyzed as: Purpose → Logic → Line-by-line → Design points → 3 examples → Improvements
  366. - **Test Case Analysis** (if tests exist, select 3-5 typical test cases to illustrate functionality)
  367. - Application Transfer Scenarios (at least 2)
  368. - Quality Verification Checklist
  369. - **Smart Strategy Selection**: Code ≤ 2000 lines uses progressive, code > 2000 lines auto-enables parallel processing
  370. - Parallel processing: Master coordinator Agent + parallel sub-agents + auto-aggregation
  371. - Ensures sufficient depth for each chapter
  372. ### 🔬 Research Foundation
  373. Based on the following cognitive science research:
  374. - **[Dunlosky et al. (2013)](https://www.aft.org/ae/fall2013/dunlosky)** - Elaborative interrogation significantly outperforms passive reading
  375. - **[Chi et al. (1994)](https://onlinelibrary.wiley.com/doi/10.1207/s15516709cog1803_3)** - Self-explainers achieve correct mental models with higher probability
  376. - **[Karpicke & Roediger (2008)](https://science.sciencemag.org/content/319/5865/966)** - Retrieval practice outperforms re-reading by 250%
  377. ### 🤝 Contributing
  378. Contributions are welcome! Here's how to participate:
  379. 1. **Report Issues** - Provide feedback via [Issues](https://github.com/notlate-cn/code-reader-skills/issues)
  380. 2. **Suggest Improvements** - Share your experience and ideas
  381. 3. **Submit PRs** - Improve documentation or add new features
  382. 4. **Share Cases** - Showcase your use cases
  383. ### ❓ FAQ
  384. <details>
  385. <summary><strong>Q: What's the difference between three modes?</strong></summary>
  386. **A:**
  387. - **Quick**: Fast understanding of structure and functionality, 5-10 min
  388. - **Standard**: Understand WHY and design rationale, 15-20 min (Recommended)
  389. - **Deep**: Application transfer testing + quality verification, 30+ min. Auto-uses parallel processing for large projects (>2000 lines)
  390. </details>
  391. <details>
  392. <summary><strong>Q: Which programming languages are supported?</strong></summary>
  393. **A:** All mainstream languages:
  394. - Python, JavaScript, TypeScript, Java, C++, Go, Rust
  395. - And various frameworks and libraries
  396. </details>
  397. <details>
  398. <summary><strong>Q: How do I know if I truly understand?</strong></summary>
  399. **A:** Use Deep Mode's "four abilities" test:
  400. 1. ✅ Can you explain the design rationale without looking at code?
  401. 2. ✅ Can you independently implement similar functionality?
  402. 3. ✅ Can you apply it to different scenarios?
  403. 4. ✅ Can you clearly explain it to others?
  404. </details>
  405. ### 📝 Changelog
  406. #### v2.4.0 (2026-03-10)
  407. **New Features:**
  408. - 🔍 **Core Snippet Identification (6A)**: For large projects, prioritize and select 3-7 most important snippets before deep analysis
  409. - 💻 **Six-Section Deep-Dive Template (6B)**: Each snippet analyzed with: Purpose / Logic / Line-by-line / Design points / 3 comparison examples / Improvements
  410. - 🤖 **Two-phase sub-agent instruction**: Code analysis sub-agents enforce snippet identification → deep-dive workflow
  411. **Refactor:**
  412. - 📦 SKILL.md trimmed from 2371 to 688 lines (71% reduction), removing redundant examples and duplicate structures while preserving all execution logic
  413. - 🌐 Chinese and English versions synchronized with identical structure
  414. #### v2.3.0 (2026-02-07)
  415. **Refactor:**
  416. - 🔄 Merge Parallel Deep Mode into Deep Mode as internal strategy
  417. - ⚡ Optimize trigger word system, enrich trigger words for each mode
  418. - 🤖 Deep Mode smart selection: ≤2000 lines progressive, >2000 lines auto-parallel
  419. - 📖 Simplify to 3 modes: Quick/Standard/Deep
  420. - 🧪 Add test case reverse understanding feature (Step 6.5)
  421. **Improvements:**
  422. - 🎯 Better UX, no manual parallel mode selection needed
  423. - 📝 Updated Chinese/English SKILL.md and README
  424. - 🔬 Support reverse-understanding code through test cases
  425. - 🛠️ Support C++, MLIR/LLVM and other language test formats
  426. **Problems Solved:**
  427. - ✅ Fixed redundant modes issue
  428. - ✅ Auto strategy selection lowers usage barrier
  429. - ✅ Discover hidden behaviors and boundary conditions through tests
  430. #### v2.2.0 (2026-02-04)
  431. **New Features:**
  432. - ✨ Scenario-based comment style (Scenario N / Step N)
  433. - ✨ Token optimization strategy (direct file writing)
  434. - ✨ Progressive generation (Deep Mode only)
  435. - ✨ Multi-language comment support standards
  436. **Improvements:**
  437. - 📦 Optimized document structure, removed duplicates
  438. - 📖 Unified chapter depth self-check standards
  439. - 🎯 Enhanced execution flow example format
  440. **Optimizations:**
  441. - 🔧 File size: 1500 → 1342 lines (-10.5%)
  442. - ⚡ Removed Sub-Agent usage guidelines
  443. - 📝 Merged duplicate mode selection sections
  444. #### v2.1.0 (2026-01-31)
  445. **New Features:**
  446. - ✨ Quick/Standard/Deep analysis modes
  447. - ✨ Smart mode selection mechanism
  448. - ✨ Optimized output structure
  449. **Improvements:**
  450. - 📦 Separate Chinese/English directory structure
  451. - 📖 Updated README documentation
  452. - 🎯 Refined trigger words
  453. ### 📄 License
  454. This project is open-sourced under the [MIT License](LICENSE).
  455. ### 🌟 Star History
  456. If this project helps you, please give it a Star ⭐
  457. [![Star History Chart](https://api.star-history.com/svg?repos=notlate-cn/code-reader-skills&type=Date)](https://star-history.com/#notlate-cn/code-reader-skills&Date)
  458. ### 📧 Contact
  459. - Bug Reports: [GitHub Issues](https://github.com/notlate-cn/code-reader-skills/issues)
  460. - Discussions: [GitHub Discussions](https://github.com/notlate-cn/code-reader-skills/discussions)
  461. ---
  462. <div align="center">
  463. **Made with ❤️ and 🧠 for deeper code understanding**
  464. **基于 ❤️ 和 🧠 创建,助力深度理解代码**
  465. [⬆ Back to Top](#-code-deep-understanding-analyzer)
  466. </div>

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.