Book Summarizer Skill
Transforms a .md book into a rich, referenceable knowledge artifact.
When to Use
- •User provides a markdown file of a book (converted from epub/pdf)
- •User says "summarize this book" or "extract insights from..."
- •Processing books for the Knowledge/Books/ folder
Architecture
Pass 1 (Sonnet 1M): Structure & Theme Extraction Pass 2 (Sonnet 1M): Quote Mining & Evidence Collection Pass 3 (Opus): Synthesis, Connections & Actionable Insights
Execution
Pass 1: Structure & Themes
Use Sonnet with this prompt:
code
You are analyzing a book to extract its deep structure. Read the entire text carefully. TASK: Identify the book's architecture and core arguments. OUTPUT FORMAT: ## Core Thesis [1-2 sentences: What is the author's main claim?] ## Structural Overview [How is the book organized? What's the logical flow?] ## Key Themes (max 7) For each theme: - **Theme Name**: [descriptive label] - **Summary**: [2-3 sentences] - **Chapter Coverage**: [which sections address this] ## Author's Methodology [How does the author build their case? Evidence types, rhetorical strategies] ## Intended Audience & Purpose [Who is this for? What action/belief change does the author want?] ## Tensions & Nuances [Where does the author acknowledge complexity? Counterarguments addressed?]
Pass 2: Quote Mining
Use Sonnet with this prompt:
code
You are a scholarly reader extracting the most valuable passages from this book. TASK: Find quotes that are quotable, insightful, or essential to understanding. CATEGORIES TO EXTRACT: ## Thesis Statements (3-5) [Passages where the author states their core argument most clearly] ## Memorable Lines (5-10) [Beautifully written, surprising, or highly quotable passages] ## Key Evidence (5-10) [Specific data, stories, or examples that support major claims] ## Counterintuitive Insights (3-5) [Ideas that challenge conventional wisdom] ## Actionable Wisdom (3-5) [Practical guidance or principles that can be applied] ## Controversial Claims (2-3) [Bold statements that might provoke debate] FORMAT FOR EACH QUOTE: > "Exact quote here..." **Context:** [1 sentence explaining where this fits] **Why Notable:** [Why this quote matters]
Pass 3: Synthesis (Opus)
Use Opus for deep synthesis:
code
You are synthesizing a book analysis into a permanent knowledge artifact.
You have:
1. Structural analysis (themes, methodology, organization)
2. Curated quotes (thesis statements, memorable lines, evidence)
TASK: Create the definitive summary that makes re-reading unnecessary for reference purposes.
OUTPUT THE FINAL SUMMARY:
---
title: "{BOOK_TITLE}"
author: "{AUTHOR}"
date_read: "{DATE}"
rating: [1-5]
tags: [books, {genre}, {themes}]
status: summarized
---
# {BOOK_TITLE}
*by {AUTHOR}*
## One-Sentence Summary
[The book in a single, memorable sentence]
## The Core Argument
[3-5 sentences: What is this book really saying? Not a description—a distillation.]
## Why This Book Matters
[2-3 sentences: What's the significance? Why should someone care?]
## Key Insights
### 1. [Insight Title]
[2-3 sentences explaining the insight]
> "Supporting quote..."
### 2. [Insight Title]
[Continue for 5-7 major insights...]
## The Best Quotes
> "Quote 1..."
> "Quote 2..."
[Include 5-7 of the absolute best, most referenceable quotes]
## Practical Applications
- [How can this be applied?]
- [What should change based on this?]
- [What decisions does this inform?]
## Connections
- **Agrees with:** [Other books/ideas that align]
- **Disagrees with:** [Contrasting perspectives]
- **Builds on:** [Foundational ideas this extends]
- **Reminds me of:** [Personal connections, projects, situations]
## Critical Assessment
**Strengths:** [What the book does well]
**Weaknesses:** [Gaps, blind spots, overreaches]
**Who Should Read:** [Specific recommendations]
**Who Should Skip:** [Who won't benefit]
## If You Only Remember One Thing
[The single most important takeaway—make it stick]
---
*Summarized by Claude on {DATE}*
Output Location
Save to: Knowledge/Books/Summaries/{Book Title} ({Author}).md
Agent Orchestration
code
1. Read the source .md file (full content) 2. Launch Sonnet agent (Pass 1) → save to temp 3. Launch Sonnet agent (Pass 2) → save to temp 4. Read both temp outputs 5. Launch Opus agent (Pass 3) with combined context 6. Write final summary to Knowledge/Books/Summaries/ 7. Offer to delete source .md if desired
Example Invocation
code
User: Summarize ~/Downloads/Thinking-Fast-and-Slow.md Claude: I'll process this book in three passes: 1. Structure & themes (Sonnet) 2. Quote extraction (Sonnet) 3. Deep synthesis (Opus) [Executes passes...] Created: Knowledge/Books/Summaries/Thinking Fast and Slow (Daniel Kahneman).md Key insight: [preview of main takeaway]
Tips for Best Results
- •Source markdown should be clean (not OCR garbage)
- •Books under 200k tokens work best for single-pass
- •For very long books (500k+ tokens), consider chapter-by-chapter processing
- •Non-fiction works better than fiction for this format
- •If the book is technical, adjust quotes toward "key definitions" and "core principles"