thinking-patterns (nakprc)

Generate sequential thinking pattern files to study and learn AI reasoning patterns.

🧠 See each AI reasoning step as a numbered file: 1think.md, 2think.md, 3think.md...

Quick Install

npm i llm-thinking-patterns-nakprc

Quick Start

# Generate thinking patterns for any topic
npx thinking-patterns generate "explain quantum computing"

# Analyze an AI response file
npx thinking-patterns analyze response.txt

# Demo with sample data
npx thinking-patterns demo

# List available patterns
npx thinking-patterns list-patterns

Output

Each run creates a topic-derived output directory (dynamic slug) with numbered files:

machine-learning/           ← derived from "machine learning"
├── 1think.md   # Context & Framing
├── 2think.md   # Analysis
├── 3think.md   # Synthesis
├── 4think.md   # Conclusion
└── index.md    # Summary of all files

Configuration

Everything is controlled by thinkingpatterns.nakprc.config.js:

export default {
  output: {
    dir: null,                // null = dynamic (derived from topic)
    filePrefix: 'think',
    naming: 'numbered',       // 'numbered' | 'named' | 'datetime'
  },

  defaultPattern: 'reverse_engineer',

  patterns: {
    reverse_engineer: {
      label: 'Reverse Engineer AI Thinking',
      steps: [
        { id: 'context', label: 'Context & Framing', desc: 'How the AI frames the problem' },
        { id: 'analysis', label: 'Analysis', desc: 'How the AI breaks down the problem' },
        { id: 'synthesis', label: 'Synthesis', desc: 'How the AI combines insights' },
        { id: 'conclusion', label: 'Conclusion', desc: 'How the AI reaches conclusions' },
      ],
    },
    guided: {
      label: 'Guided Thinking',
      steps: [
        { id: 'observe', label: 'Observe', desc: 'What do we see?' },
        { id: 'question', label: 'Question', desc: 'What are we unsure about?' },
        { id: 'hypothesize', label: 'Hypothesize', desc: 'What could be true?' },
        { id: 'test', label: 'Test', desc: 'How do we verify?' },
        { id: 'learn', label: 'Learn', desc: 'What did we discover?' },
      ],
    },
  },

  llm: {
    enabled: false,
    provider: 'openai',
    model: 'gpt-4o',
  },
}

CLI Commands

| Command | Description | |---------|------| | generate <topic> | Generate thinking patterns for a topic | | analyze <file> | Reverse-engineer thinking from an AI response file | | list-patterns | List all available thinking patterns | | config | Show current configuration | | demo | Run with sample data |

Available Patterns

• reverse_engineer — Extract and expose thinking steps hidden in any AI response
• guided — Structured scientific reasoning (observe → question → hypothesize → test → learn)
• custom — Define your own thinking steps

Output Directory Behavior

By default, the output directory is dynamic — derived from the topic name:

| Topic | Output Directory | |-------|-------| | "machine learning" | ./machine-learning/ | | "debug memory leak" | ./debug-memory-leak/ | | "explain quantum" | ./explain-quantum/ |

Override with config.output.dir for a fixed directory.

File Naming

Files use numbered prefix by default:

| Naming Mode | Example | |-------------|------| | numbered (default) | 1think.md, 2think.md, 3think.md | | named | think-context.md, think-analysis.md | | datetime | think-2026-05-11T10-00-00-context.md |

Plug & Play Extension

One-command setup for Claude Code, Claude Desktop, Cursor, Windsurf, and any MCP-compatible LLM.

# Auto-detect your IDE and configure
npx tp-setup auto

# Or specify your IDE
npx tp-setup claude-desktop
npx tp-setup cursor
npx tp-setup windsurf

See .nakprc/INSTALL.md for full installation guides for each platform.

MCP Server (for LLMs)

Connect this to any MCP-compatible client (Claude Desktop, Cursor, Windsurf, etc.) so LLMs can automatically generate thinking patterns.

Add to your MCP config:

{
  "mcpServers": {
    "thinking-patterns": {
      "command": "node",
      "args": ["/path/to/thinking-pattern/mcp-server.mjs"]
    }
  }
}

Tools exposed to connected LLMs:

| Tool | Description | |------|--| | generate_thinking_patterns | Generate sequential files (1think.md, 2think.md...) for any topic | | analyze_ai_response | Reverse-engineer thinking from any AI response text | | list_patterns | List all available patterns | | view_thinking_files | List all generated thinking files | | get_thinking_file | View content of a specific thinking file | | get_config | Show current configuration |

Research

This tool is based on research into AI reasoning patterns. Read the full paper: research/thinking-patterns-paper.md

"Thinking Patterns of AI: Making AI Reasoning Visible Through Sequential Analysis"

License

MIT