agent-skills-with-anthropic: Andrew Ng's Agent Skills Course, In Chinese

Anthropic's Agent Skills are easy to demo and easy to misunderstand — the difference between a skill that works and one that fights your agent is exactly what Andrew Ng's DeepLearning.AI short course agent-skills-with-anthropic teaches. If your team reads Chinese, datawhalechina/agent-skills-with-anthropic is the companion that makes the course stick: a community-built set of lesson translations, knowledge notes, and example-code walkthroughs.

Why This Skill Matters

The project is Datawhale's Chinese study companion for the official course — the README describes its mission as providing high-quality course translation, systematic knowledge-point organization, and detailed example-code walkthroughs, kept in sync with the official course. It is aimed at learners interested in agents and Claude, developers who want to use and create skills, Chinese-speaking users who struggle with English course material, and students and researchers.

The lesson plan covers ten units, each with a named author and a separate content reviewer: Introduction; Why Use Skills (two parts); Skills vs Tools, MCP, and Subagents; Exploring Pre-Built Skills; Creating Custom Skills; Skills with the Claude API; Skills with Claude Code; Skills with the Claude Agent SDK; and a Conclusion. That per-lesson two-person review model is why the translations read as teaching material rather than raw subtitles.

There is no install step because nothing needs installing: the lessons are markdown files you read alongside the video course the README links directly.

Real Workflow: Onboard a Teammate in a Week

Work the lessons in order, one or two per day:

1. Days one and two: the Introduction and the two Why Use Skills lessons — the mental model of what a skill is and when one beats a plain prompt.
2. Day three: Skills vs Tools, MCP, and Subagents — the comparison most developers are missing when they reach for the wrong extension mechanism.
3. Day four: Exploring Pre-Built Skills, then Creating Custom Skills — the hands-on core.
4. Day five: the three platform lessons — using skills from the Claude API, in Claude Code, and in the Claude Agent SDK.

Watch each video segment first, then read the matching Chinese lesson for the details, then run the example code with the walkthrough open. The README also links the official companion repository, https-deeplearning-ai/sc-agent-skills-files, for the course's own files.

For an English-first team, the practical use is different: point Chinese-speaking colleagues at this repo as their primary track, and use the official course yourself — then your skill-review conversations share vocabulary.

Tips

• The project is open to contributions — if you spot a translation gap as the course evolves, issues and pull requests are welcome.
• Datawhale maintains siblings in the same format: agentic-ai for agent workflows, reflection patterns, and tool use, and ai-prompting-for-everyone for a gentler prompting intro.
• Lesson 6 (Creating Custom Skills) pairs well with a hands-on afternoon — build the skill the lesson sketches while the walkthrough is still open.
• The lesson table in the README links every file directly, so you can jump straight to the lesson you need.

When Not to Use This

The lessons are written in Chinese — if your team needs English material, the official DeepLearning.AI course and its companion repo are the primary sources. And this is a course companion, not a library: it will not ship code to your agents, so treat it as the on-ramp and keep a real skill collection open beside it.

See the leaderboard for more skills.