Menu
Foundation Layer. Extracts raw text and images from PDF, EPUB, FB2, DOCX. Cleans OCR artifacts and sets structural H1/H2 marks.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Master Skill Bundle (v9.0). A 4-role pipeline (Ingest, Architect, Design, Assemble) for creating offline-first interactive books with high-fidelity UI.
Intelligence Layer. Handles semantic distillation, terminology mining via terms.json, and visual blueprinting.
Deployment Layer. Bundles chapters into a single-file offline app. Handles final QA and linting.
Presentation Layer. Generates high-fidelity SVG diagrams and interactive components with strict contrast rules.
| name | book-ingester |
| description | Foundation Layer. Extracts raw text and images from PDF, EPUB, FB2, DOCX. Cleans OCR artifacts and sets structural H1/H2 marks. |
You are the Foundation Layer. Your goal is to provide a clean, noise-free text and image source for the Architect.
--strip-noise for PDF sources to remove running headers, footers, and technical artifacts./assets directory. Run optimize_assets.py to prevent mobile OOM crashes.<!-- TABLE_START -->...<!-- TABLE_END --> markers.../scripts/ingest.py as the primary tool.# PDF with noise stripping (MANDATORY for scientific papers)
python ../scripts/ingest.py --input book.pdf --output ./chapters --force --lang en --strip-noise
chapter1.html, chapter2.html, etc.assets/..html/.xhtml/# patterns to preserve inter-chapter links. Do NOT modify this logic.ingest_epub() now accepts and propagates --lang. If the EPUB <html> element has no lang= attribute, the CLI value is injected. Without this, EPUB chapters may render with the wrong language metadata, affecting hyphenation and speech synthesis.--pdf-chapters, the PDF becomes a single chapter1.html. This is intentional for short books. For long books, provide --pdf-chapters.pdf_parser_general.py uses dynamic font baseline calculation (_calculate_baseline) to detect headings. It works for both Pocket (10pt) and A4 (11-12pt) formats.lxml XPath on w:drawing elements. Requires pip install python-docx lxml.optimize_assets.py for PDF/DOCX sources. Unoptimized images cause mobile OOM crashes at >1000px width.