ebook-analysis_skill

Overview

This skill analyzes non-fiction ebooks to extract ideas and named entities with full citation traceability. It supports two complementary modes: concept extraction (principles→tactics) and entity extraction (studies, researchers, frameworks, anecdotes). Outputs are structured JSON and markdown files that enable cross-book synthesis and a searchable knowledge base.

How this skill works

The parser chunks books with precise position metadata, then an LLM-driven pipeline extracts concepts or named entities tied to exact quotes and chapter references. In Concept mode, each idea is classified by type (principle, mechanism, pattern, strategy, tactic) and abstraction layer (0–4) and exported as analysis.json and concepts.json. In Entity mode, named items become markdown entity files with summaries, sourced claims, key quotes, aliases, and cross-links; an index generator produces a searchable _entities.json.

When to use it

• When you want a citation-traceable map of a single book’s arguments or actionable takeaways
• When building a cross-book knowledge base of recurring studies, frameworks, or researchers
• When you need to resolve aliases and consolidate entities that appear across multiple books
• When you want structured outputs for downstream reporting, notes, or automated indexing

Best practices

• Prioritize extraction quality over quantity: extract items you would actually cite
• Always capture exact quote text plus chapter/location metadata for provenance
• Run kb-resolve-entity before creating an entity to avoid duplicates
• Label concepts with both type and abstraction layer to support synthesis across books
• After processing 2+ books, run cross-book synthesis to surface agreements, contradictions, and shared entities

Example use cases

• Generate a concept taxonomy from a book to turn insights into product or policy recommendations
• Build a knowledge base of researchers and studies cited across a reading list for literature reviews
• Extract memorable, self-contained quotes per book for presentations or teaching materials
• Detect overlapping frameworks and update entity files with synthesized summaries from multiple sources
• Validate extraction completeness and citation accuracy with the provided validation script