Book Methodology Distiller

Problem

A book can contain reusable decision frameworks, principles, checklists, and terms, but those pieces are often scattered across chapters and not in a form an agent can invoke. book2skill is not a summarizer. It distills the methodology in a book into a set of executable skills.

How It Works

It runs a staged pipeline:
- Stage 0: Whole-book understanding: Read the supplied text, apply the Adler steps for structure, interpretation, criticism, and application, then write BOOK_OVERVIEW.md.
- Stage 1: Parallel extraction: Spawn sub-agents to extract decision frameworks, principles and rules, in-book cases, failure modes, and glossary terms.
- Stage 1.5: Triple verification: A candidate must be supported by multiple passages, able to answer a new question, and non-obvious; rejected items are kept in rejected/ for auditability.
- Stage 2: Skill construction: Build each skill with R, I, A1, A2, E, and B, where A2 defines the trigger situation, E gives concrete 1-2-3 steps, and B states where the skill should not apply.
- Stages 3 and 4: Linking and stress testing: Add related-skill links, generate INDEX.md, and write test-prompts.json covering positive, negative, and ambiguous cases; failures send the item back for rebuilding rather than superficial patching.

Boundaries

This skill is for turning methodology, frameworks, and principles into callable skills, not for book excerpts, reflections, or author-style roleplay. It requires accessible book text, title, author, and publication year before running, and it should not infer the book from memory.