teach

Creates stateful teaching workspaces with HTML lessons, glossaries, and learning records.

1|Updated Jul 23, 2026
One-click install
npx skills add https://github.com/sanjanb/my-agent-harness --skill teach-sanjanb
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: teach
Source: https://github.com/sanjanb/my-agent-harness/tree/main/skills/mp-teach
Command: npx skills add https://github.com/sanjanb/my-agent-harness --skill teach-sanjanb

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Learning a new topic with an AI is usually stateless and unstructured — each session starts from zero, lessons are not grounded in your goals, and nothing tracks what you already know. This Skill turns a directory into a persistent teaching workspace that remembers your mission, your progress, and your preferred resources across sessions. ## Core Features & Use Cases - Stateful Learning Workspace: Maintains MISSION.md, RESOURCES.md, GLOSSARY.md, NOTES.md, and sequentially numbered learning records to track what you know and why you are learning it. - HTML Lesson Generation: Produces short, self-contained, beautifully typeset lessons in ./lessons/ built from reusable components in ./assets/, each tied to your mission and zone of proximal development. - Evidence-Based Teaching: Grounds lessons in curated high-trust resources rather than parametric knowledge, and applies retrieval practice, spacing, and interleaving to build long-term retention. - Use Case: You want to learn strength training over several weeks. The Skill interviews you to define a concrete mission, curates trusted sources, then generates interactive quiz-based HTML lessons and a glossary, recording each demonstrated understanding so future sessions build on it. ## Quick Start Ask the agent to teach you a topic, for example: teach me the fundamentals of strength training in this workspace.

Frequently Asked Questions about teach

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I use an AI agent to learn a new topic over multiple sessions?

Invoke the teach skill in a dedicated directory and state what you want to learn. The agent creates a workspace with a mission file, curated resources, and numbered learning records, so each new session resumes from your demonstrated knowledge rather than starting over.

What files does the teach workspace create?

The workspace contains MISSION.md for your learning goal, RESOURCES.md for trusted sources, GLOSSARY.md for compressed terminology, NOTES.md for preferences, a learning-records directory with numbered markdown files, a lessons directory of HTML lessons, and an assets directory of reusable components.

How does the skill decide what to teach next?

It calculates your zone of proximal development by reading existing learning records and your mission. Lessons target material that is challenging but achievable, and records are only written when you demonstrate genuine understanding, not mere exposure.

Does the teach skill rely on the AI's built-in knowledge?

No. The skill explicitly distrusts parametric knowledge and prioritizes curating high-trust external resources in RESOURCES.md first. Lessons are expected to cite these sources, and gaps in available resources are flagged explicitly.

What format are the lessons in and can I review them later?

Lessons are self-contained HTML files saved in ./lessons/ with incrementing numbered filenames. They are designed with clean typography for printing and later review, link to each other and to reference documents, and reuse shared components from ./assets/.