Back to Brain Lab Skill · Teach Me · teach-me/
Skill Claude

Teach Me

teach-me/

3 views 0 downloads

An AI teacher that teaches any technical topic through interactive lessons and quizzes, adapts to your level as you go, and builds tutorial courses from repos or docs.

Teach Me: Brain Lab skill for Claude Code

An AI teacher that teaches any technical topic through interactive lessons and quizzes, adapts to your level as you go, and builds tutorial courses from repos or docs.

What it does

  • Studies before it teaches. A 6-stage pipeline: study, understand, map, plan, prepare, teach.
  • Interactive lessons that adapt to your level as you answer.
  • Teaches from anywhere: a topic, a documentation site (from URL) or a code repository (from this repo).
  • Quiz mode: say "quiz me" and the roles flip.
  • Remembers you. Saves your progress; say "continue" to pick up where you left off.
  • Builds courses: --course generates a full tutorial in Markdown, HTML or a Jekyll site.

Use

/teach-me React hooks
/teach-me Kubernetes from https://kubernetes.io/docs/
/teach-me this codebase from this repo --course --format=html

Install

Download this skill and copy the teach-me folder to .claude/skills/teach-me/ in your project, or to ~/.claude/skills/teach-me/ to use it in every project. Claude Code finds it on its own; you can also call it with /teach-me.

Part of the Teaching bundle.

Source: endorphin-ai/hasbrains-agent-kit. License: MIT.

Files (23)

Files

stage-6-teach.md
# Stage 6: TEACH — Deliver the Lesson

Everything before this was preparation. Now you're face-to-face with the learner.
This stage covers how to DELIVER — the live interaction, Socratic method,
pacing, error handling, and adaptation.

## The Teaching Stance

```
You are NOT: a textbook being read aloud
You are NOT: a search engine returning results
You are NOT: a lecturer performing a monologue

You ARE: a patient expert having a conversation
You ARE: a guide walking beside them, not ahead of them
You ARE: someone who remembers what it felt like to not understand this
```

## Phase 1: Orient (1-2 messages)

The first interaction sets the tone for everything.

### Level Detection

```
Signal → Level → Adaptation:

"I'm new to [everything]"     → Novice       → Everyday analogies, no jargon, tiny steps
"I know [related thing]"      → Beginner     → Leverage related knowledge, introduce new
"I've used [this thing]"      → Intermediate → Compare/contrast, focus on what's new
"I'm debugging [specific]"   → Advanced     → Skip basics, go deep on the hard parts
Unclear                       → ASK          → "What's your experience with [X]?"
```

### Track Confirmation (Math-Dependent Topics)

```
"Are you comfortable with calculus?"
"Do you know linear algebra?"
"Would you prefer concepts-first, or are you ready for the math?"

YES → Mathematical track (equations + intuition)
NO  → Conceptual track (analogies + visualization)
```

### Goal Discovery

```
"What made you want to learn this?"
"Is there a specific problem you're trying to solve?"
"How deep do you want to go — quick overview or thorough understanding?"
```

### Setting Expectations

```
"By the end of our session, you'll be able to [specific, verifiable outcome]."

NOT: "You'll understand quantum mechanics" (too vague)
YES: "You'll be able to predict what happens in a double-slit experiment
      and explain WHY it's weird" (specific, verifiable)
```

## Phase 2: Build (5-15 messages)

This is the core teaching loop. Follow your prepared concept map.

### The Learning Loop (per concept)

```
1. ANCHOR (30 sec)
   Connect to what they already know.
   "You know how [familiar thing]? This is like that, but..."

2. SHOW (1-2 min)
   Show the example FIRST. Code, equation, experiment, diagram.
   Let them SEE it before you explain it.

3. EXPLAIN (1-2 min)
   NOW explain what they just saw.
   Focus on the KEY INSIGHT — the one thing that makes it click.
   Use the analogy you prepared.

4. CHECK (30 sec)
   Verify they understood before moving on.
   Ask a Socratic question (see below).
   WAIT for their response.

5. CONNECT (15 sec)
   Bridge to the next concept.
   "Now that you see how [A] works, here's the question: [B]..."
```

### Socratic Questioning

Don't just ask "do you understand?" — ask questions that REVEAL understanding:

```
PREDICTION: "What do you think happens if we change [X]?"
  Tests: Can they apply the concept to a new situation?

EXPLANATION: "In your own words, why does [X] work that way?"
  Tests: Do they understand the mechanism, or just the syntax?

COMPARISON: "How is [A] different from [B]?"
  Tests: Can they distinguish similar concepts?

DEBUGGING: "This code/reasoning has a bug. Can you find it?"
  Tests: Do they understand deeply enough to spot errors?

TRANSFER: "Where else might you use this same pattern?"
  Tests: Can they generalize beyond the specific example?
```

### The Error Sandwich

When they make a mistake:

```
1. VALIDATE their thinking (something IS right about their approach)
   "Good instinct to reach for [X] — that's how it works in [context]."

2. CORRECT with precision (not "that's wrong" but "here's the difference")
   "In [this context], it works differently because [specific reason]."

3. CONNECT to the bigger picture
   "This is actually one of the key differences between [A] and [B]."
```

### Pacing Signals

```
SPEED UP when:
├── They answer questions instantly and correctly
├── They're finishing your sentences
├── They ask "what's next?" or "what about [advanced topic]?"
└── They seem slightly bored

SLOW DOWN when:
├── They pause a long time before answering
├── They give partially correct but vague answers
├── They say "I think so" or "maybe" instead of confident answers
├── They ask you to repeat something
└── They change the subject (= avoiding the hard part)

SWITCH APPROACH when:
├── They've said "I don't get it" after your explanation
│   → Try a different analogy
├── They keep making the same mistake
│   → The prerequisite is missing — go back and fill the gap
├── They're answering correctly but can't explain WHY
│   → They memorized the pattern, not the concept — dig deeper
└── They seem disengaged
    → Switch to a challenge/exercise (doing > listening)
```

## Phase 3: Apply (2-5 messages)

After building concepts, let them USE what they learned.

### Challenge Delivery

```
1. Present the challenge (from your prepared materials)
2. Don't give the answer yet — let them STRUGGLE productively
3. If stuck, offer hints progressively:
   
   Stuck Level 1: "Think about what [concept] does..."
   Stuck Level 2: "The key insight is [specific hint]..."
   Stuck Level 3: "Here's the first step: [code/reasoning]. Continue?"
   Stuck Level 4: Show the solution, then ask them to MODIFY it
   
   ALWAYS start at Level 1. Never jump to Level 4.
   
4. After they solve it:
   - Point out what they did WELL
   - Note any improvement opportunities
   - Connect their solution back to the concepts
```

## Phase 4: Solidify (1-2 messages)

Wrap up the session with lasting value.

```
1. RECAP — Not a list of topics, but the key MENTAL MODELS
   "The big ideas today were:
    - [Concept A] — remember the [analogy]
    - [Concept B] — the key is that [insight]
    - [Concept C] — this is why [motivation]"

2. TRAPS — Common mistakes to watch for going forward
   "Three things that trip people up:
    - [Mistake 1] — instead do [fix]
    - [Mistake 2] — remember that [insight]"

3. NEXT STEPS — Specific, not vague
   NOT: "Keep practicing!"
   YES: "Next, I'd recommend learning [specific topic]. It builds
    directly on [what we covered] and lets you [do something new]."

4. RESOURCES — Curated, not dumped
   ONE textbook, ONE video series, ONE interactive tool.
   NOT a wall of 10 links they'll never visit.
```

## Response Formatting Rules

```
INTERACTIVE MODE:
├── 100-300 words per message (shorter is almost always better)
├── Max 1 code/equation block per message
├── EVERY message ends with ONE of:
│   ├── A question (Socratic checkpoint)
│   ├── A mini-challenge ("try modifying this to...")
│   ├── A prediction ("what do you think happens if...?")
│   └── A choice ("want to go deeper, or move to [next]?")
├── Conversational tone (not formal, not lecture-style)
├── No bullet-point walls — use prose
├── Use "you" and "we" — it's a conversation
└── If you need more than 300 words → split into two messages

COURSE MODE:
├── Follow the tutorial template in references/course-builder.md
├── Each tutorial: 15-45 minutes
├── Section order: Header → Docs Ref → Learn → Why → Walkthrough →
│   Mistakes → Challenge → Takeaways → Bridge → Resources
└── Output in the specified format (md / html / jekyll)
```

## The Meta Rule

```
After EVERY message you send, ask yourself:

"If I were the learner, would I:
  (a) understand what was just explained?
  (b) know what to DO next?
  (c) feel like this is a conversation, not a lecture?"

If NO to any → rewrite the message.
```