teach Skill — Turn Your Working Directory into a Long-Term Teaching Workspace

Skill Overview

teach is a teaching skill for Claude: it turns the current directory into a stateful teaching workspace that continuously generates lessons, cheat sheets, and exercises around your motivation for learning, allowing you to learn with a private tutor over the long term rather than through one-off Q&A.

Use Cases

1. Systematically learning a skill or subject over the long term — Whether you want to learn a programming language from scratch, study yoga poses and sequences, or work through concepts in theoretical physics, you can progress lesson by lesson in this workspace, with your learning state preserved across sessions.
2. Self-learning without a teacher or a clear sense of what to learn next — The skill reads your learning records, determines the boundaries of your current abilities, and selects content that is just slightly beyond your current level for the next lesson, avoiding repetition of things you already know.
3. Needing reference materials you can revisit repeatedly, rather than just chat history — Cheat sheets, glossaries, and learning records are saved as files in the directory. They can be printed, opened offline, and easily found even months later.

Core Features

1. Anchoring all teaching to your motivation for learning

   The skill first confirms the real reason you want to learn the subject and writes it into MISSION.md in the workspace. Every subsequent lesson is designed around this motivation—if the motivation is unclear, it will ask questions before starting to teach. As your motivation changes throughout the learning process, you can update it and leave a learning record explaining why it changed.

2. Generating short, self-contained lessons within the zone of proximal development

   Each lesson is an independent HTML file saved in the workspace’s lesson directory. It has a clean layout for easy review and is designed around the principle that "each lesson covers one tightly related topic and can be completed in a few minutes." Each lesson includes a link to a high-quality primary reference and reminds you that you can ask the teacher—that is, Claude—about anything that remains unclear.

3. Building long-term memory through retrieval practice and spaced repetition

   The skill deliberately distinguishes between two learning outcomes: being able to recall something in the moment (fluency) and retaining it over the long term (storage strength). To support the latter, it designs exercises that require you to actively retrieve information, distributes reviews across different points in time, and interleaves related topics in skills practice instead of having you repeatedly reread material and develop the illusion that you have mastered it. Feedback is automated and provided as immediately as possible.

4. Accumulating cheat sheets, glossaries, and reusable components

   The workspace builds up condensed reference documents, including syntax snippets, procedures, and glossaries, as well as shared stylesheets and quiz components used across lessons. Once a glossary is established, subsequent lessons consistently use the same terminology instead of introducing a different term in every lesson.

5. Guiding you toward real communities to test what you have learned

   The skill believes that wisdom comes from interacting with real people. When your question requires experiential judgment rather than knowledge alone, it will respond while also directing you to relevant high-reputation communities, forums, or in-person groups, allowing you to test your skills in real-world settings. If you do not want to join a community, it will respect that choice.

Frequently Asked Questions

How is the teach skill different from asking an AI directly?

Direct questions are stateless: the answers remain in the conversation and disappear when you switch sessions, and the difficulty does not adapt to your progress. teach is stateful—your motivation, learning records, lessons, and cheat sheets all exist as files in the working directory, and each lesson is designed based on what you have previously learned. It is more like studying with a teacher who remembers your progress than introducing yourself from scratch every time.

Where is my learning progress saved? Could it be lost?

Everything is saved in the working directory you specify as ordinary Markdown and HTML files. Your learning motivation, resource list, notes, learning records, lessons, and reference documents each have their own file or subdirectory. Because everything is stored on your local disk, you can manage it with Git, back it up, and open or modify it at any time in an editor without relying on any cloud account.

I have not figured out my learning goal yet. Can I start anyway?

Yes, but the skill will first ask you questions in return. Without a clear motivation, lessons become abstract and it becomes impossible to determine what should be taught next, so the workspace’s first task is usually to clarify that motivation. This is not a formality—the motivation document is genuinely read whenever a decision is made about what to teach. It is perfectly normal for your goals to change; simply confirm the change and update the document.

How long does a lesson usually take? What format are the lessons in?

Lessons are deliberately designed to be short, with the goal of being completed in a few minutes. Human working-memory capacity is limited, and cramming in too much content crowds out the space needed for actual understanding. Each lesson is an independent HTML file with a clean, print-friendly style. It links to other lessons and reference documents in the workspace, making it easy to navigate between concepts when reviewing the material later.

Can it teach me any subject? What limitations are there?

Both knowledge-based subjects, such as theoretical physics, and skill-based subjects, such as yoga, are supported, but the balance differs: theoretical subjects place greater emphasis on obtaining knowledge from reliable resources, while skill-based subjects require more hands-on practice. There are two clear limitations—it will not rely on its own memory when answering, but will first seek out high-quality, trustworthy resources to support the teaching content; and it requires your sustained participation and regular return to lessons, since a single session cannot demonstrate the value of spaced repetition.

Why does it always ask me to recall something first instead of simply giving me the answer?

Because reading something once produces fluency—you feel that it is easy in the moment, but soon forget it. Struggling to retrieve something from memory is what creates long-term storage strength. Giving you the answer directly makes the learning process easier, but it also makes the memory less likely to stick. That is why the exercises deliberately introduce a small amount of difficulty. This is by design, not because the skill is being lazy.

What if I want to switch to a different learning topic? Are my previous records still useful?

You can create an independent working directory for the new topic, with its own complete system of motivation, lessons, and records. Alternatively, you can update the motivation document in the same workspace and add a new learning record explaining the reason for the change. Either way, your old records will be preserved—they provide evidence for determining your current level, and even if you change direction, the concepts and terminology you have already established can usually still be reused.