infinite-gratitude

by @sstklen in AI & LLM

# Install this skill:

npx skills add sstklen/infinite-gratitude

Or install specific skill: npx add-skill https://github.com/sstklen/infinite-gratitude

# Description

Multi-agent research that keeps bringing gifts back — like cats! Dispatch multiple agents to research a topic in parallel, compile findings, and iterate on new discoveries.

# SKILL.md

name: infinite-gratitude
description: Multi-agent research that keeps bringing gifts back — like cats! Dispatch multiple agents to research a topic in parallel, compile findings, and iterate on new discoveries.
argument-hint: "" [--depth quick|normal|deep] [--agents 1-10]

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

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

Part of 🥋 AI Dojo Series by Washin Village 🐾

# README.md

🐾 無限貓報恩 | Infinite Gratitude | 無限の恩返し

Dispatch 10 parallel research agents — like having a team of researchers working for you simultaneously

⚡ Quick Start

# Install (one command!)
curl -sSL https://raw.githubusercontent.com/sstklen/infinite-gratitude/main/infinite-gratitude.skill.md \
  -o ~/.claude/skills/infinite-gratitude.skill.md

# Use in Claude Code
/infinite-gratitude "your research topic"

💡 What It Does

Problem: Deep research takes hours. Reading papers, comparing tools, analyzing competitors — one person can only do so much.

Solution: Dispatch multiple AI agents in parallel. Each agent researches a different angle, then brings findings back.

You: "Research pet AI recognition"
     ↓
🐱🐱🐱🐱🐱 5 agents go out (parallel)
     ↓
📊📊📊📊📊 Each brings back a report
     ↓
You: "Great! Now go deeper on ArcFace..."
     ↓
🔄 Loop until satisfied

Like cats bringing gifts home — mice, bugs, leaves. This skill keeps bringing research findings until you say stop.

📊 Real Results: Pet AI Research

We used this skill to research building an AI system for recognizing 28 cats & dogs.

Metric	Result
Research Topics	12
Agents Deployed	10 (parallel)
Reports Generated	9
Time	30 minutes (vs 20+ hours manual)
Key Discovery	Petnow's 99% accuracy secret

Reports Produced

#	Report	Key Finding
1	Competitor Analysis	Petnow leads with 99% accuracy
2	Dataset Survey	Oxford-IIIT Pet is commercially safe
3	Technical Roadmap	ArcFace > Triplet Loss for stability
4	GitHub Projects	MegaDescriptor is the best pretrained model
5	HuggingFace Models	DINOv2 for general, MegaDescriptor for animals
6	Petnow Deep Dive	Siamese + Self-Attention + 200K data
7	Loss Function Guide	ArcFace vs Triplet comparison
8	Business Model	Pet insurance is the money maker
9	Data Formula	10K→85%, 50K→92%, 200K→99%

Outcome: Achieved 77.6% accuracy, with clear roadmap to 90%+.

🔧 Configuration

# Basic usage
/infinite-gratitude "topic"

# Deep research (more thorough)
/infinite-gratitude "RAG best practices" --depth deep

# Control agent count
/infinite-gratitude "vector databases" --agents 10

# Multiple waves
/infinite-gratitude "embedding models" --waves 5

Parameter	Default	Description
`--depth`	`normal`	`quick`, `normal`, `deep`
`--agents`	`5`	Parallel agents (1-10)
`--waves`	`3`	Research iterations

🎯 Best Use Cases

Use Case	Why It Works
Technical Research	Compare 10 tools/libraries simultaneously
Competitor Analysis	Each agent analyzes a different competitor
Literature Review	Parallel paper reading and summarization
Market Research	Multi-angle industry analysis
Due Diligence	Comprehensive background checks

📁 Files

├── infinite-gratitude.skill.md   # ← Install this!
├── infinite-gratitude-story.md   # Full origin story
└── docs/                         # Additional documentation

🐾 Origin Story

In Japan's Boso Peninsula, Washin Village is home to 28 cats and dogs. While building their AI recognition platform, there was too much research for one person.

So we made AI agents work like village cats: go out, bring gifts back, repeat.

The name "Infinite Gratitude" (無限報恩) comes from cats bringing "gifts" home — their way of saying thanks.

Full story: infinite-gratitude-story.md

📜 License

MIT License

Made with 🐾 by Washin Village — 和牠一起，療癒全世界

# Supported AI Coding Agents

This skill is compatible with the SKILL.md standard and works with all major AI coding agents:

⚡ Amp 🚀 Antigravity 🤖 Claude Code 🦀 Clawdbot 📝 Codex ▶️ Cursor 🤖 Droid 💎 Gemini CLI 🐙 GitHub Copilot 🪿 Goose 📊 Kilo Code 🔧 Kiro CLI 💻 OpenCode 🦘 Roo Code 🌲 Trae 🏄 Windsurf

Learn more about the SKILL.md standard and how to use these skills with your preferred AI coding agent.