Infinite Gratitude 🐾

無限貓報恩 | 無限の恩返し Multi-agent research that keeps bringing gifts back — like cats! 🐱

Quick Reference

Option	Values	Default
topic	Required	-
--depth	quick / normal / deep	normal
--agents	1-10	5

Usage

/infinite-gratitude "pet AI recognition"
/infinite-gratitude "RAG best practices" --depth deep
/infinite-gratitude "React state management" --agents 3

Behavior

Step 1: Split Directions

Split {topic} into 5 parallel research directions:

1. GitHub projects
2. HuggingFace models
3. Papers / articles
4. Competitors
5. Best practices

Step 2: Dispatch Agents

Task(
    prompt="Research {direction} for {topic}...",
    subagent_type="research-scout",
    model="haiku",
    run_in_background=True
)

Step 3: Collect Gifts

Compile all findings into structured report.

Step 4: Loop

If follow-up questions exist → Ask user → Continue? → Back to Step 2

Step 5: Final Report

Example Output

🐾 Infinite Gratitude!

📋 Topic: "pet AI recognition"
🐱 Dispatching 5 agents...

━━━━━━━━━━━━━━━━━━━━━━
🎁 Wave 1
━━━━━━━━━━━━━━━━━━━━━━

🐱 GitHub: MegaDescriptor, wildlife-datasets...
🐱 HuggingFace: DINOv2, CLIP...
🐱 Papers: Petnow uses Siamese Network...
🐱 Competitors: Petnow 99%...
🐱 Tutorials: ArcFace > Triplet Loss...

💡 Key: Data volume is everything!

🔍 New questions:
   - How to implement ArcFace?
   - How to use MegaDescriptor?

Continue? (y/n)

🐾 by washinmura.jp

Notes

• Uses haiku model to save cost
• Max 5 agents per wave
• Deep mode loops until satisfied

Additional Resources

• For agent configuration, see references/agent-config.md

Related Skills

• ai-dojo — Foundation for AI coding agents
• research-scout — Single-agent research

---

Part of 🥋 AI Dojo Series by Washin Village 🐾