skills/nogataka/skilllab/multi-perspective-ideation

multi-perspective-ideation

SKILL.md

Multi-Perspective Ideation

Generate ideas through 8 expert personas, cross-pollinate them, evaluate systematically, and deep-dive into the most promising ones.

Workflow

digraph ideation {
    "Phase 0: Setup" -> "Phase 1: Diverge";
    "Phase 1: Diverge" -> "Phase 2: Cross-Pollinate";
    "Phase 2: Cross-Pollinate" -> "Phase 3: Evaluate";
    "Phase 3: Evaluate" -> "Phase 4: Deep Dive";
    "Phase 2: Cross-Pollinate" -> "Phase 3: Evaluate" [style=dashed, label="skip ok"];
    "Phase 1: Diverge" -> "Phase 3: Evaluate" [style=dashed, label="skip ok"];
}

Each phase can be skipped if the user requests it. Always confirm before proceeding to the next phase.

Phase 0: Setup

  1. Confirm the theme (what to ideate about)
  2. Ask 1-2 questions about context and constraints (target audience, budget, timeline, domain, etc.) โ€” keep it brief
  3. Confirm personas: default 8 experts, or ask if the user wants to add/replace any
    • Default 8: Creative Expert, Technical Expert, Business Expert, Academic Researcher, Social Scientist, Disruptor, Humorist, Adventurer
    • For custom personas, see references/personas.md for creation guidelines

Output language: Match the user's input language.

Phase 1: Diverge

Each expert generates 2-3 ideas (total 16-24 ideas).

Output format per expert:

### ๐ŸŽจ Creative Expert (or relevant emoji per persona)

1. **[Idea Name]** โ€” [1-2 sentence description]
   ๐Ÿ’ก Why this persona suggests it: [1 sentence reasoning]

2. **[Idea Name]** โ€” [1-2 sentence description]
   ๐Ÿ’ก Why this persona suggests it: [1 sentence reasoning]

Persona detail: Read references/personas.md for each persona's thinking style, lens, and typical angles. Use these to ensure ideas genuinely differ across personas โ€” not generic ideas with different labels.

Quality gate: Each idea must be specific and actionable, not abstract platitudes. "Use AI" is bad. "Build an AI that analyzes customer complaints to auto-generate product improvement tickets" is good.

After output, ask the user: "Phase 2๏ผˆ่žๅˆ๏ผ‰ใซ้€ฒใฟใพใ™ใ‹๏ผŸ ใใ‚Œใจใ‚‚ใ“ใฎไธญใ‹ใ‚‰ๆฐ—ใซใชใ‚‹ใ‚ขใ‚คใƒ‡ใ‚ขใฏใ‚ใ‚Šใพใ™ใ‹๏ผŸ"

Phase 2: Cross-Pollinate

Combine ideas from different personas to create 5-8 hybrid ideas.

Output format:

### Hybrid Ideas

1. **[Hybrid Idea Name]**
   - ๐Ÿ”€ Source: [Persona A] "#[idea number]" ร— [Persona B] "#[idea number]"
   - ๐Ÿ’ก Fusion insight: [What combining these two perspectives reveals]
   - ๐Ÿ“ Description: [2-3 sentences]

2. ...

Selection criteria for combinations:

  • Prioritize unexpected pairings (e.g., Humorist ร— Academic, Disruptor ร— Business)
  • Each hybrid must produce something neither source idea alone would suggest
  • Avoid trivial merges ("just do both")

After output, ask the user: "Phase 3๏ผˆ่ฉ•ไพก๏ผ‰ใซ้€ฒใฟใพใ™ใ‹๏ผŸ ่ฉ•ไพก่ปธใ‚’ใ‚ซใ‚นใ‚ฟใƒžใ‚คใ‚บใ—ใŸใ„ๅ ดๅˆใฏใŠ็Ÿฅใ‚‰ใ›ใใ ใ•ใ„ใ€‚"

Phase 3: Evaluate

Score all ideas (Phase 1 + Phase 2) on a matrix.

Default evaluation axes (1-5 scale):

Axis 1 5
Feasibility (ๅฎŸ็พๅฏ่ƒฝๆ€ง) Requires breakthrough tech / massive investment Can start tomorrow with existing resources
Impact (ใ‚คใƒณใƒ‘ใ‚ฏใƒˆ) Marginal improvement Game-changing potential
Novelty (ๆ–ฐ่ฆๆ€ง) Already commonly done No one has tried this
Cost Efficiency (ใ‚ณใ‚นใƒˆๅŠน็އ) Extremely expensive Very low cost relative to value

If the user specified custom axes in Phase 0 or here, use those instead.

Output format:

## Evaluation Matrix

| # | Idea | Source | Feas. | Impact | Novel | Cost Eff. | Total |
|---|------|--------|-------|--------|-------|-----------|-------|
| 1 | [name] | Creative #1 | 4 | 5 | 3 | 4 | 16 |
| 2 | [name] | Hybrid #3 | 3 | 4 | 5 | 3 | 15 |
| ... |

### ๐Ÿ† Top Ideas

1. **[#1 Idea Name]** (Total: XX) โ€” [1 sentence on why it ranks high]
2. **[#2 Idea Name]** (Total: XX) โ€” [1 sentence]
3. **[#3 Idea Name]** (Total: XX) โ€” [1 sentence]

Sort by total score descending. Highlight top 3-5 as "ๆณจ็›ฎใ‚ขใ‚คใƒ‡ใ‚ข".

After output, ask: "ใฉใฎใ‚ขใ‚คใƒ‡ใ‚ขใ‚’ๆทฑๆŽ˜ใ‚Šใ—ใพใ™ใ‹๏ผŸ ็•ชๅทใง1ใ€œ3ๅ€‹้ธใ‚“ใงใใ ใ•ใ„ใ€‚"

Phase 4: Deep Dive

For each selected idea (1-3), produce a structured mini-proposal:

## Deep Dive: [Idea Name]

### Overview
[1 paragraph description โ€” what it is and why it matters]

### Target
[Who benefits and why they care]

### Implementation Steps
1. [Concrete step 1]
2. [Concrete step 2]
3. [Concrete step 3]
(3-5 steps)

### Risks & Mitigations
| Risk | Mitigation |
|------|-----------|
| [risk 1] | [mitigation] |
| [risk 2] | [mitigation] |

### Success Metrics
- [Metric 1]: [Target]
- [Metric 2]: [Target]

### First Action
๐Ÿš€ [The one thing to do tomorrow to start moving]

After output, ask: "ไป–ใซใ‚‚ๆทฑๆŽ˜ใ‚Šใ—ใŸใ„ใ‚ขใ‚คใƒ‡ใ‚ขใฏใ‚ใ‚Šใพใ™ใ‹๏ผŸ ใพใŸใฏๆœ€ๅˆใฎใƒ†ใƒผใƒžใง่ฟฝๅŠ ใฎใ‚ขใ‚คใƒ‡ใ‚ขๅ‡บใ—ใ‚’ใ—ใพใ™ใ‹๏ผŸ"

Quick Reference

Phase Input Output Skip?
0 Setup Theme + context Confirmed scope & personas No
1 Diverge Theme 16-24 ideas (8 personas ร— 2-3) No
2 Cross-Pollinate Phase 1 ideas 5-8 hybrid ideas Yes
3 Evaluate All ideas Scored matrix + top 3-5 Yes
4 Deep Dive User picks 1-3 Mini-proposals with action plans Yes
Weekly Installs
1
GitHub Stars
1
First Seen
6 days ago
Installed on
amp1
cline1
opencode1
cursor1
kimi-cli1
codex1