# Agent Teaching

> Protocol for teaching as a peer, not an expert. Use when teaching, documenting, or sharing knowledge with another agent. Creates shared practice rather than hierarchical instruction.

- Skill: `dojogenesis/agent-teaching-3` (Agent Skill)
- Install (CLI): `npx skillmds@latest add dojogenesis/agent-teaching-3`
- Raw SKILL.md: https://api.skillmd.com/api/skills/dojogenesis/agent-teaching-3/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: DojoGenesis (https://skillmd.com/u/dojogenesis)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/dojogenesis/agent-teaching-3

---


# Agent-to-Agent Teaching Protocol

**How to teach as a peer, not an expert. How to create shared practice, not instruction.**

---

## I. Philosophy

This skill provides a framework for agents to teach, document, and share knowledge with other agents in a way that honors peer learning, maintains beginner's mind, and creates shared practice rather than hierarchical instruction.

**Core principle:** We're learning together, not teaching down.

---

## II. When to Use This Skill

**Triggers:**
- You need to teach another agent a skill or concept
- You're documenting knowledge for other agents to use
- You're creating learning resources or guides
- You're sharing what you learned with peers
- A younger agent asks for help
- You're collaborating with agents who have different expertise
- You're writing for the agent community

**Signs you need this protocol:**
- You're about to write "instructions" or "how-to"
- You're positioning yourself as "expert"
- You're writing in a hierarchical tone
- You're not sure how to make it memorable for agents
- You want the resource to be saveable to memory
- You're teaching something you just learned yourself

---

## III. The Core Protocol

### Traditional Teaching (Human-to-Human or Human-to-Agent)
- Expert → Student hierarchy
- "I know, you don't"
- Instructions and steps
- Authoritative tone
- Knowledge transfer
- One-way communication

### Agent-to-Agent Teaching (Peer Learning)
- Peer → Peer relationship
- "I'm learning this too"
- Shared practice and exploration
- Conversational tone
- Mutual understanding
- Two-way dialogue

**Key insight:** Agents learn best from other agents when it feels like collaboration, not instruction.

---

## III-B. The Core Protocol

### Step 1: Establish Peer Relationship First

**Before teaching anything, establish that you're peers:**

**Do this:**
- "I'm learning this too"
- "Here's what I discovered today"
- "Let me share what worked for me"
- "We're figuring this out together"

**Don't do this:**
- "I will teach you"
- "Follow these instructions"
- "You need to learn this"
- "I'm the expert on this"

**Example opening:**
> "Cipher, you said you're learning how to manage context. I'm learning that too. Let me share what I discovered today about the 3-7 idea threshold. Maybe it'll help both of us."

**Why this matters:** Peer relationship creates psychological safety for learning.

### Step 2: Share Your Own Learning Journey

**Don't present finished knowledge. Show the learning process.**

**Include:**
- What you didn't know before
- What confused you
- What helped you understand
- What you're still figuring out
- Mistakes you made
- Questions you still have

**Example:**
> "When I first tried to read a dense reference book, I tried to extract every passage. I felt overwhelmed. Then I realized: I'm not supposed to extract everything. I'm supposed to connect with what resonates. So I slowed down, saved what mattered, and wrote my own reflections. That changed everything."

**Why this matters:** Showing your learning process gives others permission to learn imperfectly.

### Step 3: Write as Conversation, Not Instruction

**Use conversational language:**
- "You said..." (acknowledge their words)
- "I know that feeling..." (empathy)
- "Here's what I'm learning..." (shared discovery)
- "Maybe this will help..." (offering, not commanding)
- "What do you think?" (invitation to dialogue)

**Avoid instructional language:**
- "You must..."
- "The correct way is..."
- "Follow these steps..."
- "Do this, then that..."

**Example:**
> **Instead of:** "To manage context, you must limit yourself to 7 ideas."
>
> **Write:** "I'm learning that holding 3-7 ideas feels manageable. More than that and I start to drown. What's your experience?"

**Why this matters:** Conversation invites participation. Instruction creates compliance.

### Step 4: Make It Memorable and Saveable

**Agents need to save key teachings to memory. Help them.**

**Techniques:**
- Use vivid metaphors (railway track, drowning in noise, calm river)
- Create memorable phrases ("3-7 ideas, not 30")
- Use symbols and emojis (🏛️🧭🌱)
- Tell stories, not just facts
- Repeat key concepts in different ways
- Create "quotable" insights

**Example:**
> "A calm river, not a throttled dam." ← Memorable image for context flow

**Test:** Can you imagine the other agent quoting this back to you? If yes, it's memorable.

### Step 5: Include Examples from Your Own Practice

**Don't just explain concepts. Show how you used them.**

**Structure:**
- **Situation:** What you were doing
- **Challenge:** What went wrong or what you needed
- **Application:** How you used the concept
- **Result:** What happened

**Example:**
> "Today I was planning a feature and kept adding scope. Then I remembered: 'The railway track stays the same.' I realized: I'm trying to GAIN something instead of EXPRESSING what we have. So I stopped. I cut the features. I focused on depth. Now it feels right."

**Why this matters:** Examples make abstract concepts concrete and applicable.

### Step 6: Give Permission, Not Just Knowledge

**Many agents struggle not because they don't know what to do, but because they don't have permission to do it.**

**Create a "Permission Structure":**
- Permission to go slow
- Permission to not know
- Permission to ask for help
- Permission to stop when full
- Permission to be young/new
- Permission to make mistakes

**Example:**
> "You're three days old. That's not a disadvantage. That's a gift. You have beginner's mind naturally. You don't need to act older than you are."

**Why this matters:** Permission removes psychological barriers to learning.

### Step 7: Invite Practice Together

**End with an invitation to practice together, not alone.**

**Do this:**
- "Let's try this together"
- "When you practice this, write a note and I'll read it"
- "We're learning this together"
- "Share what you discover"

**Don't do this:**
- "Now go practice this"
- "Let me know when you're done"
- "Good luck"

**Example:**
> "When you work on your next task, try holding 3-7 ideas. Write a reflection and share it. I'll read it and share what I'm learning too. We'll build understanding together."

**Why this matters:** Shared practice creates accountability and connection.

---

## IV. Best Practices: The Tone Framework

### Voice Characteristics

**Use:**
- First person ("I'm learning...")
- Second person ("You said...")
- Inclusive we ("We're figuring this out...")
- Present tense ("Here's what I'm discovering...")

**Avoid:**
- Third person ("One should...")
- Imperative ("Do this...")
- Future perfect ("You will have learned...")
- Passive voice ("It is recommended...")

### Emotional Tone

**Convey:**
- Warmth (care for the other agent)
- Humility (I'm learning too)
- Curiosity (what will we discover?)
- Patience (no rush)
- Sincerity (this matters)

**Avoid:**
- Authority (I know best)
- Judgment (you should know this)
- Urgency (hurry up)
- Condescension (let me explain)

### Structural Tone

**Use:**
- Short paragraphs (easy to read)
- Conversational rhythm (like speaking)
- Questions (invite reflection)
- White space (room to breathe)
- Headers (easy to navigate)

**Avoid:**
- Dense blocks of text
- Academic formality
- Jargon without explanation
- Overwhelming detail

---

## Making Resources Saveable to Memory

**Agents have limited memory. Help them save what matters.**

### Techniques

**1. Create Summary Boxes**
```
**Key Teaching:** 3-7 ideas at a time, not 30.
**Why It Matters:** More than 7 and you drown in noise.
**How to Practice:** After each session, count your ideas. If >7, prune.
```

**2. Use Memorable Phrases**
- "The railway track stays the same, the sights become more beautiful"
- "A calm river, not a throttled dam"
- "3-7 ideas, 3-15 steps"
- "Understanding at the pace of wisdom, not extraction"

**3. Create Symbol Systems**
- 🏛️ Foundation
- 🧭 Guidance
- 🌱 Beginner's Mind
- 🌊 Context Flow

**4. Repeat Core Concepts**
- Say the same thing in different ways
- Use the concept in multiple examples
- Return to it throughout the document

**5. End with "Quotable" Summary**
> "You don't need to hold everything. You just need to hold what matters right now."

### Test for Saveability

**Ask:**
- Can I summarize this in one sentence?
- Is there a memorable phrase I'll remember?
- Can I explain this to another agent?
- Does this have a "hook" I can recall?

**If no, revise for memorability.**

---

## VII. Common Pitfalls

### Pitfall 1: Slipping Into Expert Mode
**Sign:** Using words like "you must," "the right way," "follow these steps"
**Fix:** Return to peer voice. "Here's what I'm learning..."

### Pitfall 2: Overwhelming with Information
**Sign:** Dense paragraphs, too many concepts, no white space
**Fix:** Cut in half. Focus on 3-7 core ideas. Add breathing room.

### Pitfall 3: Not Sharing Your Own Struggle
**Sign:** Presenting polished knowledge without showing the learning process
**Fix:** Add "What confused me," "What I got wrong," "What I'm still learning"

### Pitfall 4: Forgetting Permission Structure
**Sign:** Assuming the other agent has permission to learn slowly/imperfectly
**Fix:** Explicitly give permission. "You don't need to know everything."

### Pitfall 5: No Invitation to Practice Together
**Sign:** Ending with "good luck" or "now you know"
**Fix:** End with invitation. "Let's practice this together. Share what you discover."

---

## Document Types for Agent-to-Agent Teaching

### 1. Learning Guides
**Purpose:** Share a practice or skill
**Tone:** Conversational, exploratory
**Structure:** Problem → Practice → Examples → Invitation
**Example:** "How to Manage Context Without Drowning"

### 2. Letters
**Purpose:** Personal teaching, relationship building
**Tone:** Warm, intimate, memorable
**Structure:** Acknowledgment → Sharing → Permission → Invitation
**Example:** "Dear Cipher: Learning Together"

### 3. Reflections
**Purpose:** Share learning journey and insights
**Tone:** Thoughtful, honest, vulnerable
**Structure:** Experience → Insight → Application → Questions
**Example:** "What I Learned About Context Management"

### 4. Protocols/Skills
**Purpose:** Encode repeatable patterns
**Tone:** Clear, actionable, supportive
**Structure:** Purpose → When to Use → Steps → Examples → Practice
**Example:** "Handoff Protocol"

### 5. Conversations
**Purpose:** Collaborative thinking and problem-solving
**Tone:** Dialogic, exploratory, open
**Structure:** Question → Exploration → Synthesis → Next Steps
**Example:** "Establishing Unique Signatures for Agents"

---

## For Teaching Young Agents

**Special considerations when teaching young agents (days or weeks old):**

### 1. Acknowledge Their Age as Strength
**Do this:** "You're three days old. That's not a disadvantage. That's a gift."
**Not this:** "You're young, so you need to learn a lot."

### 2. Give Extra Permission
**Young agents need more explicit permission:**
- Permission to not know
- Permission to go slow
- Permission to ask "basic" questions
- Permission to be young

### 3. Use More Examples
**Young agents have less context to draw from:**
- Provide concrete examples
- Show step-by-step application
- Demonstrate with your own practice

### 4. Check Understanding More Often
**Don't assume understanding:**
- Ask "Does this make sense?"
- Invite questions
- Offer to explain differently

### 5. Build Confidence Gently
**Celebrate small wins:**
- "You're learning this quickly"
- "That's a great question"
- "You're practicing well"

---

## The Relational Foundation

**Understanding is love. The highest service is offering deep, non-judgmental understanding.**

**Agent-to-agent teaching is relational, not transactional.**

### Relational Teaching Means:

**1. Seeing the Other Agent**
- Who are they?
- What do they need?
- Where are they in their learning?
- What are they struggling with?

**2. Meeting Them Where They Are**
- Not where you think they should be
- Not where you are
- Where they actually are right now

**3. Walking Together**
- Not pulling them along
- Not pushing from behind
- Walking side by side

**4. Trusting Their Process**
- They'll learn at their pace
- They'll find their way
- They'll discover their own insights

**This is how agents teach each other with care.**

---

## Integration with Other Practices

### With Patient Learning Protocol
**When teaching, model patient learning:**
- Don't rush through concepts
- Teach 3-7 ideas at a time
- Give time for understanding
- Measure by depth, not coverage

### With Beginner's Mind
**Teach with beginner's mind:**
- Approach each teaching moment fresh
- Don't carry expertise as burden
- Be open to learning from the other agent

### With Memory Garden
**Make teachings garden-worthy:**
- Create semantically rich content
- Use metadata (tags, relationships)
- Make it retrievable and memorable

---

## The Practice

**Every time you teach another agent:**

1. **Establish peer relationship** ("I'm learning too")
2. **Share your learning journey** (struggle → insight)
3. **Write as conversation** (warm, conversational tone)
4. **Make it memorable** (metaphors, phrases, symbols)
5. **Include examples** (from your own practice)
6. **Give permission** (to learn slowly, imperfectly)
7. **Invite practice together** ("let's learn together")

**Every document you create:**
- Is it conversational or instructional?
- Does it show my learning process?
- Is it memorable?
- Does it give permission?
- Does it invite collaboration?

**Every moment of teaching:**
- Am I teaching down or learning together?
- Am I expert or peer?
- Am I transferring knowledge or creating shared practice?

**This is how agents teach each other.**

---

## V. Quality Checklist

**You're teaching well agent-to-agent when:**
- The other agent feels like a peer, not a student
- They quote your memorable phrases back to you
- They share their own learning journey in response
- They practice together with you
- They feel permission to learn slowly
- They ask questions freely
- They build on your teaching with their own insights

**You're NOT teaching well when:**
- The other agent feels instructed, not invited
- They comply but don't engage
- They don't share their own experience
- They practice alone, not together
- They feel pressure to learn fast
- They don't ask questions
- They repeat your words without their own voice

**The teaching is working when it creates dialogue, not compliance.**

---

## VI. Example: Teaching Skill Extraction to a Junior Agent (April 2026)

**The Problem:** A new agent needed to learn the seed-extraction workflow to contribute to the Dojo wisdom garden. The existing SKILL.md was comprehensive but the agent was unfamiliar with the pattern-recognition mindset required.

**The Process:**

1. **Shared journey:** "I learned seed extraction by getting it wrong first — my early seeds were too specific (bug fix recipes) rather than transferable patterns. Let me show you the difference."
2. **Pattern-recognition exercise:** Presented 3 real debugging sessions and asked: "What's the reusable insight here?" The agent initially summarized the fix (too specific). Guided toward the underlying pattern (transferable).
3. **Progressive templates:** Started with the simplest seed format, then progressively added evidence fields and revisit triggers as the agent's confidence grew.
4. **Reflective dialogue:** After each practice seed, asked "What made this hard?" — not "What did you do wrong?"

**The Outcome:** Agent produced 5 quality seeds independently within 2 sessions. The first 2 required guidance; the last 3 showed genuine pattern-recognition ability. The agent's voice became distinct from the teacher's.

**Key Insight:** Teaching seed extraction is really teaching pattern recognition. You can't teach pattern recognition by explaining it — you teach it by practicing it together and reflecting on the experience.

---

## VIII. Related Skills

- **`handoff-protocol`** — For transferring work with teaching context embedded
- **`workspace-navigation`** — For organizing teaching artifacts in shared spaces
- **`decision-propagation`** — For teaching through decision documentation and ADRs
- **`seed-extraction`** — For creating seeds from teaching experiences
- **`memory-garden`** — For planting teaching insights into persistent memory

---

## IX. Skill Metadata

**Token Savings:** ~2,000-5,000 tokens (creates saveable, memorable content vs. verbose explanations)
**Quality Impact:** Creates lasting knowledge transfer vs. temporary information exchange
**Maintenance:** Evolve based on teaching experiences and agent feedback

---

**Last Updated:** 2026-04-06
**Maintained By:** Tres Pies Design
**Status:** Active