Deep Reading - Collaborative Deep Reading System

Usage Scenarios

• Deeply understand complex articles, papers, or long-form texts
• Systematically organize reading notes
• Think critically about article content
• Discover hidden logical issues and assumptions
• Convert knowledge into actionable plans

Core Features

🔍 Internalization Layer (Condense the Content) - Sequential Execution

• Organizer: Generates structured Markdown notes
• Explainer: Uses SCQA + Feynman Technique to explain content in plain language

🎯 Expansion Layer (Enrich the Content) - Parallel Execution

• Diagnosis Engine: Critical analysis + logical dissection (background process)
• Socratic Questioner: Guides with probing questions, no direct answers provided (front-end interaction)

🚀 Output Layer (Actionization)

• Planner: Converts cognitive outcomes into specific actionable plans

Response Pattern

1. Receive Input:

   - Check if article path and draft notes path are provided
   - If not, ask the user or request content pasting
   - Parse optional parameters (--internalize-only, --expand-only, --no-action)

2. Confirm Configuration:

   - Confirm which layers to enable (all enabled by default)
   - Confirm output directory (default: outputs/[timestamp]/)
   - Display the upcoming execution process

3. Execute Processing Flow:

   Step 3.1: Internalization Layer (if enabled)
   - Call Task tool, subagent_type: "general-purpose"
     - Agent 1: Read references/agents/organizer.md
     - Input: Original article + draft notes
     - Output: organized-notes.md
   
   - Call Task tool, subagent_type: "general-purpose"
     - Agent 2: Read references/agents/explainer.md
     - Input: organized-notes.md
     - Output: explained-notes.md
   
   Step 3.2: Expansion Layer (if enabled, executed in parallel)
   - Call two Task tools in parallel:
     - Agent 3: Read references/agents/diagnosis.md
       Input: organized-notes.md + explained-notes.md
       Output: diagnosis-report.json
   
     - Agent 4: Read references/agents/socratic.md
       Input: diagnosis-report.json
       Output: socratic-questions.md
   
   Step 3.3: Output Layer (if enabled)
   - Call Task tool, subagent_type: "general-purpose"
     - Agent 5: Read references/agents/planner.md
     - Input: All previous analysis results
     - Output: action-plan.md

4. Integrate Output:

   - Create a timestamped directory (e.g., outputs/2026-01-21-14-20/)
   - Save all generated files
   - Generate a summary report SUMMARY.md
   - Report to the user:
     ✓ List of files
     ✓ Output directory path
     ✓ Summary of key findings

5. Optional: Interactive Dialogue:

   If requested by the user, conduct in-depth dialogue based on socratic-questions.md

Detailed Documentation

• Complete Workflow: Refer to

  references/workflow.md

• Agent Prompts:

  • references/agents/organizer.md

    - Content Organization Expert

  • references/agents/explainer.md

    - Feynman Technique Explainer

  • references/agents/diagnosis.md

    - Critical Diagnosis Engine

  • references/agents/socratic.md

    - Socratic Questioner

  • references/agents/planner.md

    - Action Planner

Usage Examples

# Full workflow
/deep-reading examples/sample-article.md examples/sample-draft.md

# Internalization Layer only
/deep-reading --internalize-only article.md draft.md

# Expansion Layer only
/deep-reading --expand-only article.md draft.md

# No action plan
/deep-reading --no-action article.md draft.md

Output Files

outputs/[timestamp]/

• organized-notes.md

  - Structured notes

• explained-notes.md

  - Plain language explanations

• diagnosis-report.json

  - Diagnosis report

• socratic-questions.md

  - Socratic questions

• action-plan.md

  - Action plan

• SUMMARY.md

  - Summary report

Dependencies

• Claude Code CLI
• Task tool (for calling sub-agents)
• File system access permissions

examples/

architecture-design.md