@veyhon: Compile a technical book into a Skill loadable by Claude Code / Amp: first extract frameworks, principles, and anti-patterns, then use chapter files to bring context on demand. https://github.com/virgiliojr94/book-to-skill… book-to…

Compile a technical book into a Skill loadable by Claude Code / Amp: first extract frameworks, principles, and anti-patterns, then use chapter files to bring context on demand. https://github.com/virgiliojr94/book-to-skill… book-to-skill first follows different extraction chains depending on the document type: for technical PDFs it uses Docling to preserve tables and code blocks, for text PDFs it falls back to pdftotext → PyPDF2 → pdfminer, and EPUB / DOCX / HTML / RTF / MOBI also have corresponding fallbacks. The extracted results are written to full_text.txt and metadata.json, then Claude identifies the book title, author, chapters, table of contents, and core topics. The output lands in the skills directory of Amp or Claude Code: SKILL.md contains the core mental models and topic index, chapters/ keeps each chapter within 800–1,200 tokens, and glossary.md, patterns.md, cheatsheet.md store terminology, patterns, and quick reference respectively. For large books, use grep / sed to read by chapter slices, avoiding repeatedly stuffing the entire book into the context.