Definition
Think-Better is an "OS for clear thinking and better decisions". It is a tool that brings structured decision frameworks directly into AI prompts, turning generic AI responses into rigorous, structured analyses.
The decide skill handles choices such as compare, should I, and pros and cons. It can apply a reversibility filter, a weighted matrix, a pre-mortem, or a pros-cons-fixes pass, and it warns about bias. The solve skill is for debug and root cause work. It walks a seven-step McKinsey method and can split the problem with an issue tree or MECE. Setup targets Claude Code, Copilot, or Antigravity.
Installation
# macOS / Linux
curl -sSL https://raw.githubusercontent.com/HoangTheQuyen/think-better/main/install.sh | bash
# Windows (PowerShell)
irm https://raw.githubusercontent.com/HoangTheQuyen/think-better/main/install.ps1 | iex
# Alternative: Go install
go install github.com/HoangTheQuyen/think-better/cmd/make-decision@latest
Setup
think-better init --ai claude # Claude Code
think-better init --ai copilot # GitHub Copilot
think-better init --ai antigravity # Antigravity
The Two Main Skills
/decide — For Choices
Keywords: "choose", "compare", "should I", "pros and cons"
10 Decision Frameworks:
| Framework | Description |
|---|---|
| Reversibility Filter | Assess how reversible a decision is |
| Weighted Matrix | Weighted comparison |
| Hypothesis-Driven | Decisions based on hypotheses |
| Pre-Mortem | Analyze failure before making the decision |
| Pros-Cons-Fixes | Pros-Cons-Fixes with bias warnings |
12 Cognitive Bias Warnings:
- Overconfidence
- Anchoring
- Sunk Cost
- Status Quo
- Confirmation Bias
- And 7 other biases
# Usage
/decide "Should we migrate from React to Next.js?"
/decide.quick "Choose between AWS vs GCP"
/decide.deep "Enterprise migration decision"
/decide.exec "Board-level strategic decision"
/solve — For Problems
Keywords: "solve", "debug", "root cause", "I'm stuck"
7-Step McKinsey Method:
- Define → 2. Decompose → 3. Prioritize → 4. Analyze → 5. Synthesize → 6. Communicate → 7. Recommend
15 Decomposition Methods:
- Issue Tree
- MECE (Mutually Exclusive, Collectively Exhaustive)
- Hypothesis Tree
- Profitability Tree
- Systems Map
- And 10 other methods
12 Mental Models:
- First Principles
- Inversion
- Bayesian Updating
- Pareto Principle
10 Communication Patterns:
- Pyramid Principle
- BLUF (Bottom Line Up Front)
- SCR (Situation-Complication-Resolution)
- Action Titles
# Usage
/solve "Why is the API slow?"
/solve.quick "Quick debug this issue"
/solve.deep "Root cause analysis"
/solve.exec "Executive-level problem report"
Depth Levels
| Command | Depth | Records | Best For |
|---|---|---|---|
/solve.quick / /decide.quick |
Quick | 0.5× | Fast scan |
/solve / /decide |
Standard | 1.0× | Default |
/solve.deep / /decide.deep |
Deep | 1.7× | Complex decisions |
/solve.exec / /decide.exec |
Executive | 2.5× | Board reports |
CLI Commands
think-better init # Install skills
think-better list # Show installed skills
think-better check # Verify prerequisites (Python 3)
think-better uninstall # Remove skills
think-better version # Show version
How It Works
Architecture: a BM25 search engine querying 160 knowledge records with depth multipliers (0.5× to 2.5×), combined with an advisor engine:
- Classifies input (decision/problem)
- Applies relevant frameworks
- Detects cognitive biases
- Generates structured output with next-step suggestions
Auto-activation: Think Better activates automatically when it detects a decision or problem in the conversation.
The engine is a BM25 search over 160 knowledge records, scaled by depth multipliers from 0.5× to 2.5×. It classifies the input as a decision or a problem, applies a framework, and detects cognitive bias. It then returns a structured answer plus a next step. Quick mode is 0.5×, standard is 1.0×, deep is 1.7×, and executive is 2.5×. Ask it to save step-by-step and it writes a markdown workspace with an overview, decomposition, analysis, findings, synthesis, a recommendation, and bias warnings. A binary download has no extra requirement. Go install and source builds need Go 1.25 or later. Running the skills needs Python 3. It is a thinking framework only, so it stores no memory. It supports only Claude, GitHub Copilot, and Antigravity.
Example Flow
User: "Should we migrate from React to Next.js?"
Think Better:
1. Detects: Binary Choice
2. Applies: Reversibility Filter framework
3. Warns: Overconfidence Bias, Status Quo Bias
4. Output: Weighted comparison matrix with recommendations
Step-by-Step Workspace
Add "save step-by-step" to any prompt to generate a full markdown workspace:
- Overview
- Decomposition
- Analysis
- Findings
- Synthesis
- Recommendation
- Bias warnings
Requirements
| Method | Requirements |
|---|---|
| Binary download | None |
| go install | Go 1.25+ |
| Build from source | Go 1.25+ |
| Running skills | Python 3 |
Strengths
| Strength | Description |
|---|---|
| Structured thinking | Turns AI into a structured decision maker |
| Bias detection | Automatically detects 12 cognitive biases |
| Multiple frameworks | 10 decision frameworks + 15 decomposition methods |
| Depth control | 4 selectable levels of analysis |
| Easy integration | Simple install, integrates with Claude, Copilot, Antigravity |
| CLI tool | No code needed, runs directly |
Weaknesses
| Weakness | Description |
|---|---|
| Go-based | Requires Go 1.25+ to build |
| Limited LLM support | Only supports Claude, GitHub Copilot, Antigravity |
| Not a memory system | Only a thinking framework, stores no memory |
When to Use
- Complex decision making: When you need to make an important decision
- Problem solving: Analyze problems using the McKinsey method
- Bias prevention: Recognize and avoid cognitive biases
- Team workshops: As a framework for group thinking
- Strategic planning: With the Executive depth level
References:
See also: Superpowers · Memory Tool Comparison