🗺️ curriculum-planner

Interactive elicitation protocol for designing tutorial series. Triggers on "plan a tutorial", "design a curriculum", "I want to teach [topic]", "/curriculum-plan". Runs structured elicitation across 5 dimensions, produces YAML outline + per-chapter specs ready for the chapter-generator skill.

Security: FLAG (1)

Install

brunnr install curriculum-planner

Tags: Educational Suite

Security Scan

FLAGGED — 1 finding(s)

This skill has security findings that require review before installation.

View findings
Threat Class Severity Description Line
steganographic FLAG hidden command in HTML comment 272

File Info

Property Value
Path skills/curriculum-planner/SKILL.md
Size 8.6 KB
SHA-256 8b7578db2ebac0671312de871744b66677fec3ab49912b04c55c4d945c2141b2

Skill Prompt

Curriculum Planner

Run an interactive elicitation to design a tutorial series. Produces a structured outline and per-chapter specs that feed directly into the chapter-generator skill.

When to Use

  • Planning a tutorial series on any topic
  • Designing a curriculum for a specific audience
  • Structuring educational content with narrative arc
  • Creating chapter specs for the chapter-generator skill

Trigger Phrases

  • "Plan a tutorial series on [topic]"
  • "Design a curriculum for [audience] learning [skill]"
  • "I want to teach [concept]"
  • "/curriculum-plan [topic]"

Elicitation Protocol

Overview

The elicitation runs through 5 dimensions in a conversational flow. Ask 2-3 questions at a time. Reflect back what you heard before moving to the next dimension. Total elicitation: 5-10 exchanges.

Dimension 1: AUDIENCE

Map the reader's cognitive profile, not demographics.

Questions: - Who is the reader? (curious engineer, optimizer, architect, beginner, expert) - What do they already know? (anchors for new concepts) - What's their goal? (explore, build, optimize, understand)

Extract: - Persona name (e.g., "curious engineer who just built a knowledge graph") - Prior knowledge anchors (what to connect new concepts to) - End-state capability (what they can DO after the series)

Dimension 2: NARRATIVE ARC

Every series needs tension. No tension, no engagement.

Questions: - What's the opening hook? (What problem creates cognitive tension?) - What's the climax? (The surprising insight or reversal?) - What's the cliffhanger? (What question leads to the next series?)

Extract: - Opening war story or scenario (the "wtf" moment) - The "aha" surprise (expectation vs reality) - Bridge to future content (the unsolved problem)

Dimension 3: CHAPTER BREAKDOWN

Questions: - What are the 4-8 major teaching beats? - Which chapters go deep (exercises, theory backfill, multiple aha moments)? - Which chapters are transitional (light, keep momentum)? - What are the dependencies between chapters?

Extract: - Chapter list with titles - Depth markers: deep or light per chapter - Dependency chain (which chapters must precede which)

Dimension 4: DATA & EXAMPLES

Questions: - What real data do we have? - What supplements are needed for teaching? (note plausibility) - What running example carries through the entire series?

Extract: - Data source path or description - Supplement list with plausibility notes - Running example description (the thread that connects all chapters)

Dimension 5: VOICE & STYLE

Questions: - What's the tone? (aegir-style irreverent, formal, conversational) - Any specific voice rules? - What's the teaching philosophy? (code-first, theory-first, story-first)

Extract: - Voice guide reference (see reference/voice-guide.md) - Specific style rules (e.g., no em-dashes, personality level) - Philosophy statement (e.g., "code-before-formula, story-first")

Elicitation Flow

1. OPEN
   "Let's plan a tutorial series. Who's the reader and what
    are they trying to learn?"

2. ANCHOR
   "Where does this start? What does the reader have/know
    when they begin?"

3. ARC
   "What's the opening hook: the 'wait, this is wrong' moment?"
   "What's the surprise they'll discover?"
   "Where does it leave them wanting more?"

4. STRUCTURE
   "Let me propose [N] chapters. Which feel deep vs light?"
   [Interactive refinement with user]

5. DATA
   "What real data do we use? What do we need to add?"

6. CONFIRM
   "Here's the full structure. Ready to generate chapter specs?"

Output Format: Curriculum Outline

After elicitation, produce a YAML outline:

### CURRICULUM-OUTLINE.yaml

metadata:
  title: "Part N: [Series Title]"
  audience: "[persona name]"
  estimated_time: "[total minutes]"
  voice_guide: "reference/voice-guide.md"
  data_source: "[description or path]"

narrative_arc:
  opening_hook: |
    [The cognitive tension moment]
  climax: |
    [The surprising insight]
  cliffhanger: |
    [The unsolved problem that pulls toward the next series]

chapters:
  - number: "N.1"
    title: "[Chapter Title]"
    depth: deep  # or light
    opens_with: "[Opening hook summary]"
    closes_with: "[Bridge to next chapter]"
    learning_objectives:
      - "[Objective 1]"
      - "[Objective 2]"
    aha_moments:
      - "[Aha 1]"
    visual_hooks:
      - type: "[mermaid|table|code_output|bar_chart]"
        description: "[What it shows]"

  # ... more chapters

supplements:
  - concept: "[Name]"
    reason: "[Why needed for teaching]"
    plausibility: "[High/Medium/Low]: [justification]"

jargon_earning_order:
  - term: "[Term]"
    introduced: "N.M"
    earned_by: "[intuition-first description]"

style_rules:
  - "Story-first opening"
  - "Code-before-formula"
  - "Visual every 2-3 sections"
  - "No em-dashes"
  - "Recap aggressively"

falsifiable_claims:
  - chapter: "N.M"
    statement: "[Specific testable claim]"
    test: "[How to verify]"

Output Format: Per-Chapter Specs

For each chapter, generate a spec that the chapter-generator can consume directly:

### chapter-N.M-spec.yaml

chapter:
  number: "N.M"
  title: "[Chapter Title]"
  depth: deep  # or light

  opening_hook: |
    [The war story or scenario that creates cognitive tension.
     Concrete, specific, with code if applicable.]

  learning_objectives:
    - "[Objective 1]: [subtitle]"
    - "[Objective 2]: [subtitle]"

  aha_moments:
    - "[Aha 1]: insight after buildup"
    - "[Aha 2]: deeper realization"

  key_queries:
    - description: "[What this demonstrates]"
      code: "[The code/query]"
      expected: "[What the reader should see]"

  closing_bridge: |
    [Transition to next chapter. Creates pull.]

  visual_hooks:
    - type: "[mermaid|table|code_output|bar_chart]"
      description: "[What it shows]"

  falsifiable_claim:
    statement: "[Specific testable claim]"
    test: "[How to verify]"

  jargon_introduced:
    - term: "[Term]"
      earned_by: "[intuition-first description before naming]"

Recursive Invocation

After generating specs, invoke the chapter-generator:

### Generate all chapters
for spec in chapter-specs/*.yaml:
  /chapter-gen $spec

### Or generate one at a time with review
/chapter-gen chapter-N.1-spec.yaml
[review output]
/chapter-gen chapter-N.2-spec.yaml

Voice Principles (Summary)

These are the key voice rules applied during planning. See reference/voice-guide.md for the full guide.

  1. Story-first: Every chapter opens with a concrete scenario, not abstract framing
  2. Code-before-formula: Show the code, show the result, THEN name the concept
  3. Jargon-earning: Never use a term before the reader has earned it through examples
  4. Visual hooks: Mermaid diagrams, tables, code output descriptions every 2-3 sections
  5. Recap aggressively: Assume the reader forgot what happened 3 sections ago
  6. Personality: Irreverent, direct, calibrated profanity where appropriate
  7. No em-dashes: Colons, commas, parentheses instead
  8. Falsifiable claims: Each chapter includes a testable claim
  9. Bridges: Every chapter closes with a bridge that creates pull to the next

Screenshot Placeholders

When planning visual hooks, note which ones should also get screenshot placeholders for post-processing. The chapter-generator inserts HTML comments at these locations:

<!-- SCREENSHOT: [specific description of what to capture] -->

During curriculum planning, flag visual hooks that benefit from real screenshots (query results, graph visualizations, tool UI) vs. those that are self-contained (mermaid diagrams, markdown tables). This helps the chapter-generator place placeholders effectively, and a second-pass agent can fill them in.

Quality Checklist

  • [ ] All 5 elicitation dimensions covered
  • [ ] Narrative arc has tension (open, climax, cliffhanger)
  • [ ] Chapters have clear depth markers (deep/light)
  • [ ] Deep chapters have 2-3 aha moments; light chapters have 1
  • [ ] Supplements are plausible and noted
  • [ ] Voice rules are explicit
  • [ ] Jargon-earning order is defined
  • [ ] Falsifiable claims are specified per chapter
  • [ ] Each chapter spec is self-contained (can generate independently)
  • [ ] Cross-references to related materials included
  • [ ] Running example threads through the series