What problem does it solve? Reloading a full book or manual into context every session wastes tokens and buries the useful knowledge. This Skill distills a long document (PDF, EPUB, DOCX, HTML, Markdown, and more) into a reusable Claude Code skill organized by chapter, so an agent loads only the relevant chapter instead of the entire source. ## Core Features & Use Cases - Chapter-level distillation: Splits the source into one file per chapter using a fixed template covering core ideas, named frameworks, key concepts, mental models, anti-patterns, and decision rules. - Silent mojibake detection and fix: Documents a real bug where pdftotext corrupts accented characters while reporting success, and provides a pypdf-based extraction script plus verification checks (accent counts, chapter counts) before proceeding. - Skills 2.0 adaptation: Generates the supporting files (glossary, patterns catalog, decision cheatsheet) and guides adding skill.yaml and schema.json so the output works as a native skill in this repo. - Use Case: You have the 413-page SWEBOK v4 in Spanish. Run this Skill to produce 18 distilled chapter files plus a question-to-chapter topic index, so future queries about requirements elicitation load only the relevant chapter. ## Quick Start Convert the attached PDF manual into a Claude Code skill with one distilled file per chapter, a glossary, and a decision cheatsheet.
