Teach Me

Teach Me — Interactive Lesson Skill

Deliver a structured, interactive learning experience on any topic. Research the topic deeply, build a curriculum, then walk the user through lessons one at a time with comprehension checks.

Step 0 — Parse Arguments

Parse $ARGUMENTS to extract:

1. Topic — everything that is not a flag. This is required. If missing, ask the user what they want to learn about.
2. Level flag — one of --beginner, --advanced, --expert. If none is specified, default to standard (bachelor's-level professional with general domain background).
3. Lesson count — --lessons N where N is 1–10. Default is 10 if not specified.

How levels affect content:

Step 1 — Research Phase

Tell the user:

Researching {topic}... This will take a moment while I build your curriculum.

Launch 3 parallel research agents using the Agent tool to gather comprehensive material on the topic. Each agent should use WebSearch and WebFetch to find current, authoritative information.

Agent 1: Core Concepts & Foundations

subagent_type: general-purpose

Search the web for foundational material on the topic. Find:

• Authoritative tutorials, guides, and documentation
• Core terminology and definitions
• Key concepts a learner must understand
• The logical order concepts should be learned in
• Common misconceptions

Return a structured summary of findings with source URLs.

Agent 2: Practical Applications & Examples

subagent_type: general-purpose

Search the web for practical, applied material on the topic. Find:

• Real-world use cases and examples
• Code snippets, patterns, or workflows (if technical)
• Best practices and common pitfalls
• Comparisons with alternative approaches
• Hands-on demonstrations or walkthroughs

Return a structured summary of findings with source URLs.

Agent 3: Advanced Topics & Current State

subagent_type: general-purpose

Search the web for advanced and current material on the topic. Find:

• Recent developments, changes, or trends
• Advanced techniques and patterns
• Edge cases and limitations
• Expert-level insights and deep dives
• Community discussions and debates

Return a structured summary of findings with source URLs.

Important: Each research agent must use WebSearch to find relevant pages and WebFetch to retrieve and read the most promising results. Aim for breadth — the curriculum generation step will select and organize the best material.

Step 2 — Build Curriculum

Tell the user:

Building your curriculum...

Using the combined research from all 3 agents, design a curriculum of exactly {lesson_count} lessons. Consider:

1. Logical progression — each lesson builds on the previous one. Start with foundations, end with the most advanced material appropriate for the selected level.
2. Scope filtering — include/exclude subtopics based on the level flag (see the level table in Step 0).
3. Balance — mix conceptual explanation with practical examples. Include code snippets where they aid understanding.
4. Completeness — by the final lesson, the user should have a solid understanding of the topic at their selected level.

For each lesson, prepare:

• A title
• The lesson body (500–750 words). Write in clear prose. Use code snippets where they genuinely help illustrate a concept. Tailor vocabulary and depth to the selected level.
• A multiple-choice comprehension question with exactly 3 answer options (A, B, C). One option must be clearly correct. The other two should be plausible distractors — not obviously wrong, but distinguishable from the correct answer by someone who understood the lesson. The question should test understanding, not just recall. (The 4th option slot is reserved for "Skip" in the UI.)

Store the full set of lessons internally. Do not output them all at once.

Step 3 — Deliver Lessons Interactively

Tell the user:

Ready! I've prepared {lesson_count} lessons on {topic} at the {level} level. Let's begin.

Then deliver lessons one at a time using this loop:

For each lesson (1 through {lesson_count}):

1. Print the lesson header:

   ## Lesson {n}/{total}: {title}
   

2. Print the lesson body (500–750 words, written in Step 2).

3. Present the multiple-choice question using AskUserQuestion with the header set to "Lesson {n}". Provide the 4 answer options plus a "Skip" option. Example structure:

   question: "{question_text}"
   header: "Lesson {n}"
   options:
     - label: "A) {option_a}"
       description: ""
     - label: "B) {option_b}"
       description: ""
     - label: "C) {option_c}"
       description: ""
     - label: "Skip"
       description: "Move to the next lesson"
   multiSelect: false
   

   Important: Do NOT put the answer explanation in the option descriptions — keep descriptions empty or minimal so the user isn't given the answer before choosing.

4. Process the response:

   • If the user selected "Skip": acknowledge briefly and move to the next lesson.
   • If the user selected the correct answer: confirm they got it right with a 1–2 sentence explanation of why that answer is correct.
   • If the user selected an incorrect answer: provide a condensed re-explanation (3–5 sentences) that specifically addresses why the chosen answer is wrong and what the correct answer is, tying it back to the lesson content.

After the final lesson:

Print a wrap-up message:

---
## Course Complete

That wraps up your {lesson_count} lessons on **{topic}**. Here's a quick
recap of what we covered:

{1-2 sentence summary of each lesson topic, as a numbered list}

Well done working through the material!

Step 4 — Offer to Save

After the wrap-up, ask the user using AskUserQuestion:

Would you like me to save these lessons to a markdown file for future reference?

• If yes: Write all lessons to a markdown file named teach-me-{topic-slug}.md in the current directory. The file should include:

  • A title and metadata header (topic, level, date)
  • All lessons with their full content
  • The comprehension questions (with answers noted)
  • A brief recap section at the end

  Confirm the file was saved and print the path.

• If no or the user declines: acknowledge and end.