Doceo — Design skill for Claude Code

doceo: the Claude Code skill that teaches you anything in one screen

*doceo* (Latin): **"I teach."**

An open-source **AI tutor skill for Claude Code**. Point it at anything you don't understand and instead of a wall of text you get a **one-screen visual lesson**: one plain-language answer, one diagram, one analogy, and a tiny self-quiz. Every lesson is saved to your notes, so the next lesson already knows what you learned. And it improves itself from your feedback.

/doceo what is an LLM?
/doceo how does OAuth actually work
/doceo https://github.com/some/repo
/doceo src/services/billing/
/doceo            <- bare: "teach me the confusing thing we're discussing right now"

The pain it solves

You ask an AI to explain something new. It answers with 1,500 technically correct words. You read all of it, understand half of it, and remember none of it by tomorrow. Learning a new domain this way feels like being assigned a book per day.

Three things are broken in that workflow:

1. No shape. Explanations arrive as prose. Your brain wants a picture, an analogy, and one sentence that resolves the confusion.
2. No memory of you. Every chat starts from zero. The AI re-explains what you already know and skips what you actually miss.
3. No feedback loop. When an explanation fails you, nothing changes. The next one fails the same way.

Doceo fixes all three: fixed lesson shape, persistent learner memory, and a self-improvement loop that turns your feedback into permanent teaching rules.

What a lesson looks like

Every lesson has the same skeleton, capped at roughly one screen:

Part	What it gives you
The answer	One plain sentence at the top. Read only this and you still leave smarter.
The picture	One diagram: boxes, arrows, a "you are here" map. Drawn as inline SVG, sketchnote style.
The analogy	One real-world comparison picked from your world (your stack, your daily life).
Where this fits	Why