AI Coding & Agent Workflows in Chinese: A Practical Guide
A systematic Chinese tutorial collection covering Claude Code, OpenClaw, Codex, and WorkBuddy, helping individuals, teams, and institutions adopt AI coding and agent workflows.
- Source repo
- KimYx0207/AI-Coding-Guide-Zh
- Stars
- ★ 6.3k
- Last updated
- 8d ago
- License
- MIT
- Primary language
- Python
- FA score
- 38/100 · Major gaps
At a glance
- How it runs
- Works with
- Platform-specificChatGPT · Codex · Claude Code
- Cost
- Free software; you pay for model usage
- Setup effort
- Medium · a few setup steps
- You'll need
- Typical use
- A beginner developer can follow the quick-start path in the README, starting with Claude Code 01 or OpenClaw OC-03, to complete the first low-risk task, such as installing and running a simple conversation.
- Not a fit if
- Readers who want English-only documentation
- Users expecting a ready-to-run lab environment
- Source review
- 38/100 · Major gaps 7 safety controls not found
What does this agent do, and when should you use it?
This open-source repository by KimYx0207 provides in-depth Chinese tutorials for four major AI coding and agent tools: Claude Code, OpenClaw, Codex, and WorkBuddy. It contains 50 complete tutorials and 1 quick-reference card, totaling over 800k characters of Markdown content, with 1500+ code blocks/commands/config examples and 250+ FAQ entries. The tutorials are organized into four main tracks, each progressing from installation and basic usage to enterprise-grade practices, with clear learning paths and role-based recommendations. The project is MIT-licensed, supports Chinese reading, and tailors content for Chinese office scenarios such as SME AI enablement, university training, and team standards.
The repository itself does not execute operations but provides structured learning material. It includes 13 tutorials for Claude Code (Anthropic's official CLI) covering installation, basic usage, MCP, Hooks, Subagents, Skills, Plugins, etc.; 12 tutorials for OpenClaw (open-source AI assistant) covering installation, model configuration, messaging platform integration, skills, memory, multi-agent collaboration, etc.; 14 tutorials for Codex (OpenAI's coding agent platform), with the Codex App as the main path, covering installation, desktop workflow, Commands, MCP, Skills, Subagents, etc.; and 11 tutorials for WorkBuddy (Tencent's AI office assistant) aimed at office workers and domestic teams. Each tutorial provides specific commands, configuration examples, and environment requirements, such as the standard installation command for Claude Code: npm install -g @anthropic-ai/claude-code.
- A beginner developer can follow the quick-start path in the README, starting with Claude Code 01 or OpenClaw OC-03, to complete the first low-risk task, such as installing and running a simple conversation.
- A practicing developer wanting to integrate AI into real projects can focus on Claude Code 02/04/05 or Codex CX-02/CX-04/CX-05 to learn how to use MCP to connect tools and Hooks to automate workflows.
- A team lead or project manager needing to establish AI tool usage standards can refer to Claude Code 10/11 and Codex CX-13 chapters to define collaboration norms, milestone checks, and review processes.
- A corporate trainer or university lecturer can design a teachable, practicable, and assessable course path using the repository's roadmaps, enterprise security sections, and FAQ/checklists.
- An office worker with no command-line experience but wanting to use AI for weekly reports, research, or presentations can follow the WorkBuddy track (WB-01 to WB-10), using WeChat scan-to-login to access experts, skills, and knowledge base features.
- An AI enthusiast wanting to build a personal AI assistant can learn OpenClaw's OC-01 through OC-11, such as configuring messaging platforms, skills, and memory systems.
How do you install or deploy this agent?
Clone the repository: git clone https://github.com/KimYx0207/AI-Coding-Guide-Zh.git, then install dependencies for your chosen tool. Claude Code requires Node.js 18+ and can be installed via npm install -g @anthropic-ai/claude-code, authenticated with a Claude subscription or Anthropic Console. OpenClaw requires Node.js 24.x (or 22.19+) and an AI model API key (OpenAI/Anthropic/Google) or local models via Ollama. Codex App can be installed from official channels (macOS from the official entry, Windows via Microsoft Store or winget -s msstore), authenticated with a ChatGPT account or OpenAI API key. WorkBuddy desktop app is downloaded from Tencent Cloud's official site and logged in with a WeChat scan.
How do you use this agent?
Choose a learning path based on your role. For quick starts, refer to the README: Claude Code route (3 hours) follows 01→04→05; OpenClaw route (1 hour) follows OC-01→OC-02→OC-03→OC-04; Codex route (20 minutes) follows CX-01→CX-02; WorkBuddy route (30 minutes) follows WB-02→WB-01. Each tutorial includes step-by-step operations and command examples, e.g., for Claude Code, run claude in the terminal to start the CLI and use /mcp to manage MCP servers.
What are this agent's strengths and limitations?
- Covers four major AI coding and agent tools, offering a complete learning path from beginner to enterprise, suitable for different user roles.
- Extensive content with 1500+ code blocks, commands, and configuration examples, plus 250+ FAQ items, making it highly practical.
- Tailored for Chinese users, especially considering domestic office scenarios like WeChat Work and Tencent Docs, which is friendly for domestic teams.
- Continuously updated with version numbers aligned to official releases (e.g., Claude Code v2.1.222), and explains the version verification method.
- The tutorial volume is large and may be overwhelming for beginners, requiring significant time to read.
- Some tools (like Codex, WorkBuddy) depend on specific platforms or accounts, posing vendor lock-in risks.
- Version numbers may become outdated due to rapid upstream updates, requiring users to verify against official docs.
- This is tutorial content only; it does not provide a running environment or virtual lab, requiring users to set up their own environments.
How does this agent compare with similar options?
The repo compares all four tools—Claude Code, OpenClaw, Codex, and WorkBuddy—noting their respective focuses on coding, automation, multi-entry collaboration, and office tasks. Additionally, CX-14 specifically compares Codex with Claude Code to help users choose between the two.
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| AI Coding & Agent Workflows in Chinese: A Practical Guide This agent | 38 · Major gaps | CLIFree + model costs | ★ 6.3k | 8d ago | Python | ChatGPT · Codex · Claude Code |
| Project Cairn | 67 · Some gaps | Agent plugin / skillFree | ★ 241 | 5d ago | Shell | Codex · Claude Code |
| Hivemind — Shared Brain for Your Agents | 56 · Major gaps | CLIFreemium | ★ 1.6k | 4d ago | TypeScript | Codex · Claude Code · Claude API |
| NoPUA — Unlocking AI's Potential Through Love | 52 · Major gaps | Agent plugin / skillFree + model costs | ★ 1.4k | 20d ago | Python | Codex · Claude Code |
How does FollowAgents rate this agent?
Why each dimension lost points
Evidence: The repository is a collection of tutorials without executable code or automation, so there is no actual implementation of permission management, user confirmation, data flow transparency, sensitive data handling, dependency security, external effects, or rollback mechanisms. These criteria are scored 0 due to lack of relevant code or configuration. Source attribution is clear: README identifies the author as Lao Jin, provides GitHub, website, social media, and contact info, and declares MIT license, so scored 2. Deductions: no code to evaluate security attributes; attribution is clear but unverified.
Evidence: README is internally consistent; tutorial directory, version numbers, and statistics match, so self-consistency scored 2. Dependency availability: tutorials depend on external tools (Claude Code, OpenClaw, Codex, WorkBuddy) and official docs, but the repo itself lacks dependency manifests or lock files, so scored 1. Failure messages: tutorials may include troubleshooting tips, but no systematic error handling or diagnostics, so scored 1. Deductions: external dependencies without verification; unsystematic failure handling.
Evidence: README clearly distinguishes audiences (beginners, developers, team leads, enterprise training) and provides learning paths, so audience and scenarios scored 2. Capability boundaries: tutorials describe tool features and limitations, but do not clearly state the tutorial's own capability boundaries, so scored 1. Trigger precision: tutorials are static documents with no automation triggers, so scored 0. Environment fit: provides system requirements and installation instructions for each tool, so scored 2. Deductions: incomplete capability boundaries; no trigger mechanisms.
Evidence: Information architecture is clear with table of contents, parts, and reading guides, scored 2. Install notes are detailed covering system requirements and steps, scored 2. Naming stability: tutorial numbering and filenames are consistent, scored 2. Examples and FAQ are abundant with 1500+ examples and 250+ FAQs, scored 2. Known limitations: README mentions versions may become outdated, but does not systematically list all limitations, scored 1. License is clearly MIT, scored 2. Versioning and changelog: has CHANGELOG.md and version notes, scored 2. Maintenance responsibility is clear with author contact and contribution welcome, scored 2. Deductions: incomplete known limitations.
Evidence: Output usability: tutorials provide actionable steps and examples that users can follow, scored 2. Marginal value: tutorials fill a gap in Chinese AI coding tutorials and provide systematic learning paths, scored 2. Cost-benefit: tutorials are free and open source, but require user time investment and depend on external paid tools, so scored 1. Deductions: cost-benefit affected by external tool costs.
Evidence: Claim traceability: README references official docs and version numbers, and explains verification methods, scored 2. Cross-source corroboration: tutorials claim to be checked against official docs, but lack specific citation links, so scored 1. Fact-inference separation: version info is separated from advice, but some content may mix, so scored 1. Deductions: lack of specific citations; incomplete separation.
- Not found in source: least-privilege scopingGrant only what the task needs: a dedicated account or read-only token, scoped to specific directories and repos.
- Not found in source: confirmation before actingTurn on (or add) a confirmation step before it acts, and try it in a sandbox or test environment before real data.
- Not found in source: data-flow disclosureWatch which external services it contacts (proxy or firewall logs) and keep sensitive data out until you know where it goes.
- Not found in source: sensitive-data handlingUse dedicated, low-privilege, revocable API keys — never production credentials — and keep secrets out of logs.
- Not found in source: dependency securityPin versions and run a dependency audit (npm audit, pip-audit) before installing; prefer running it in a container.
- Not found in source: disclosed external effectsEstablish which external systems it writes to, sends to or changes, and verify with test accounts or repos before production.
- Not found in source: rollback or recovery pathBack up first, or work on a git branch or snapshot, so its changes can be undone.
- The repository is a tutorial collection without executable code, so actual security attributes cannot be assessed.
- Tutorials depend on external tools and official docs, but no dependency locking or verification is provided.
- Version numbers may become outdated; users should refer to official sources.
- Author identity is unverified; attribution is clear but not reliable.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
v5.5 v5.5：四工具教学补齐与近期变化复核
Release notes excerpt
这次更新把四条教程线里影响读者照做的近期变化补进正文，尤其纠正了 v5.4 把 Mods 简介当成完整教学的记录。 - Claude Code：Mods 工坊有完整文件、加载与信任、热重载、离线测试和停用步骤；同步托管读取、沙箱、IDE、记忆、插件和云任务边界。 - Codex：补 MCP 认证排障、网络撤权、任务隐藏/归档/删除、工作树占用与附图入口，讲清 Guardian 和 Touch ID 的适用条件。 - OpenClaw：补技能搜索、自学习模式与费用、只读配置及 include 备份、Sora / video-frames 退役路线，以及消息平台、定时任务、后台工作和手机入口的变化。 - WorkBuddy：补原生手机 App、文件浏览器、专家、消息引用、技能、连接器和定时任务步骤，历史说明保留日期。 …View on GitHub ↗ -
v5.4 v5.4：正文、当前功能与跟练的第二轮复核
Release notes excerpt
正文、功能与练习的第二轮复核 - 重新核对全部 50 篇教程和速查卡的当前入口、功能变化、重复 FAQ 与练习衔接，在既有核查清单加入四工具的功能覆盖矩阵；正确原文和注明日期的历史说明保留。 - Claude Code 修正不存在的 CLI 参数、日常 bypass 建议、MCP 本地配置及归档服务入口；补 AGENTS.md、mcp_tool、omitClaudeMd、Mods、账号资源同步、插件配置和企业供应商/遥测说明。语音步骤按当前录音、转写、提交和按键配置重写。 - Claude Code Notification 保留可选 title，无标题时使用默认值；格式化器采用参数列表，Git 检查使用 NUL 分隔文件名、拒绝把检查失败当通过，并返回合法的 PreToolUse ask 结构。SDK 区分可用工具池与免审批工具，删除无依据的固定并发上限。 …View on GitHub ↗ -
v5.3 v5.3：四主线全面核查与实操修正
Release notes excerpt
这次把 50 篇教程和速查卡全部扫了一遍，重点修正会让读者装错、配错或跟练失败的内容。47 篇教程和 1 张速查卡有必要更新；Codex 插件、Codex Review/PR、WorkBuddy 资料库三篇核心主线核对后保留。 - Claude Code：当前安装与 npm Node 22+ 要求、权限语义、Remote Control/Channels；Python SDK 消息类型、有限运算计算器、Hooks 通知与 Action 报告检查。 - Codex：ChatGPT 桌面入口、CLI 参数及权限 profile；新 Cloud 环境准备和发布、代理 Secrets、Scheduled 与导入范围。 - OpenClaw：匹配版本的 Docker 初始化、首次 token/数据库密码配置与三类状态卷备份，路由/模型/技能/记忆及 Agent 配置迁移；工具策略与实际隔离边界。 - WorkBuddy：实际登录、菜单、连接器与远程助理流程，虚构 CSV 跟练和参考结果；价格与未证实能力保留清楚的核查范围。 …View on GitHub ↗