Skip to content

bkan-hq/smartthink

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

20 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

English | 한국어

SmartThink

천재적 사고 엔진 시스템 — A genius-level thinking engine for Claude Code.

SmartThink combines 9 thinking frameworks (81 mental models, TRIZ innovation principles, antifragile strategy, and more) with Cynefin-based situation diagnosis to deliver deep, creative analysis on any topic.


Quick Start

1. Install

git clone https://github.com/bkan-hq/smartthink.git
cd smartthink
chmod +x install.sh uninstall.sh
./install.sh

2. Open a new Claude Code session

claude

Important: Start a new session after installation. Existing sessions won't recognize the skill.

3. Try it

/smartthink 1인 기업의 확장 전략

That's it! SmartThink will diagnose your topic, select the best thinking frameworks, and produce a full analysis with 10+ actionable ideas.


Three Modes

Deep Analysis (default)

Full interactive analysis with 9 reference modules. The main agent loads frameworks into context and interactively refines direction with you. Produces 10+ ideas with feasibility, impact, and risk assessment.

/smartthink AI 스타트업에서 네트워크 효과를 만드는 방법

Add search for data-backed analysis:

/smartthink search 2026년 SaaS 시장 트렌드

What you get:

  • Cynefin domain diagnosis (Clear / Complicated / Complex / Chaotic)
  • Multi-layer analysis from 3+ perspectives
  • 10+ genius ideas with engine/model used, feasibility, expected impact
  • Top 3 with unicorn potential assessment and execution roadmap
  • Actionable next steps
  • Brief export (if the topic has implementable tasks)

Agent Analysis

Fully automated analysis delegated to a sub-agent (STSA). Same methodology as Deep mode, but all decisions are made automatically. Use when you want hands-off analysis.

/smartthink agent 1인 기업의 수익 모델 설계

When to use: You want the same depth as Deep mode but prefer fully automated analysis without interaction points.

Light Analysis

Quick analysis without loading reference modules. Uses ~5% of context.

/smartthink light 사이드 프로젝트 아이디어

When to use: Quick brainstorming, simple questions, or when you want to save context for other work.


What Happens Behind the Scenes

You type: /smartthink [mode] 주제
         ↓
    ┌─────────────────────────┐
    │  Main Agent              │
    │  1. Cynefin 진단         │
    │  2. 모듈 추천            │
    │  3. 모드별 분기           │
    └──────────┬──────────────┘
               ↓
    ┌──────────┼──────────────┐
    │          │              │
    ▼          ▼              ▼
  Deep       Agent          Light
  (inline)   (sub-agent)    (no refs)
  - 모듈 직접  - STSA 위임    - 제1원리
    로딩       - 자동 판단     - ~5% ctx
  - 인터랙션   - 30 turns     - 3-5 ideas
  - 10+ ideas - 10+ ideas

Deep loads reference modules into main context and interacts with you. Agent delegates to a sub-agent (STSA) for hands-off analysis. Light does quick analysis without loading any reference files.


Evolution System

SmartThink learns from each session and gets smarter over time.

How it works:

  • After each analysis, SmartThink records what worked and what was lacking
  • Effective thinking patterns are saved as insights (up to 10 slots)
  • Blind spots are tracked as gaps (up to 5 slots)
  • Module diversity is monitored — if you over-rely on one framework, SmartThink nudges you toward others

Example after 3 sessions:

## 핵심 인사이트 (3/10)
- **AI 레버리지 = 스케일 상한선 제거**: AI가 1인 기업의 매출 상한을 해방
  적용 맥락: 1인 사업에서 팀 규모 제약 극복 시
  근거: 전기가 공장 규모를 해방시킨 것과 동일 구조

## 다양성 H=2.13(>=1.8). 원천: 핵심엔진3, 현실왜곡1, 패턴합성1

Your evolution data is stored in skill/.data/evolution-state.md. To preserve across machines:

cd smartthink
git add skill/.data/evolution-state.md
git commit -m "update insights"

Brief Export

When your topic has implementable tasks (code, system setup, etc.), SmartThink automatically generates a compressed brief:

skill/.data/briefs/
  └── 2026-03/
      └── 141530-seo-blog-automation/
          └── brief.md          ← Top 3 ideas + execution plan

Briefs are gitignored by default (they contain session-specific analysis).


Thinking Modules

Module Key Frameworks
Core Engines First Principles, Asymmetric Opportunity, Network Effects, Market Creation, Moat Building, Contrarian Validation, Value Capture, Timing Intelligence, Compound Advantage, Ecosystem Design, Inversion, Lollapalooza Detection
Unicorn Playbook $0-to-$1B business building (6 phases)
Reality Distortion Constraint Inversion, Category Creation, Temporal Arbitrage
Cognitive Arsenal 81 mental models across 9 disciplines
Pattern Synthesis Cross-domain pattern recognition, Weak signal detection
Execution Velocity OODA Loop, Decision frameworks, Blitzscaling
Antifragile Strategy Barbell Strategy, Optionality, Black Swan positioning
TRIZ Innovation 40 Inventive Principles, Contradiction resolution
Meta-Cognition Recursive self-improvement, Wardley Mapping, Lollapalooza

SmartThink automatically selects 2-3 modules based on your topic type (idea discovery, strategy, problem solving, etc.).


Requirements

  • Claude Code CLI installed and configured
  • macOS or Linux (Windows: use WSL2)

Uninstallation

cd smartthink
./uninstall.sh

Your evolution state and briefs remain in the repo. To fully remove, delete the cloned directory.

License

MIT

About

Genius-level thinking engine for Claude Code — 9 frameworks, 81 mental models, 12 engines, TRIZ, Cynefin diagnosis

Topics

Resources

License

Contributing

Security policy

Stars

Watchers

Forks

Packages

 
 
 

Contributors

Languages