Converting Books to Agent Skills

Bash
# Invoke the CLI to extract and prepare a book for skill generation
book-to-skill convert path/to/book.epub --output ./extracted/

# Pipeline runs automatically:
# 1. Dependency check (format-specific libraries present?)
# 2. Format dispatch (pdf/epub/html/docx/rtf/calibre fallback)
# 3. Text decoding + sanitization
# 4. Structure/metadata detection (chapters, headings, TOC)
# 5. Extracted text + metadata handed to agent generator

The agent then reads SKILL.md (generation spec) + extracted text/metadata and produces chapter files, glossary, patterns, and cheatsheets for the target agent host.

Progress:

• Step 1: Parse CLI args, load run configuration (formats, paths, output dir)
• Step 2: Run dependency checks for the detected input format; fail fast with actionable install instructions
• Step 3: Dispatch to format-specific extractor (PDF, EPUB, HTML, DOCX, RTF); fall back to Calibre conversion for unsupported/corrupt formats
• Step 4: For PDF, run a lightweight inspection/preflight pass before full extraction to catch malformed files early
• Step 5: Decode raw text (handle encoding edge cases, especially RTF/HTML)
• Step 6: Sanitize text — strip control characters, normalize whitespace, remove boilerplate (headers/footers/page numbers)
• Step 7: Detect structure — chapter boundaries, headings, metadata (title, author, TOC)
• Step 8: Package extracted text + structural metadata as input to the skill-generation step
• Step 9: Agent generator follows the generation specification (SKILL.md) to produce chapter files, glossary, pattern docs, cheatsheets
• Step 10: Optionally publish generated skill to GitHub or install directly for a target agent host

Format dispatch logic

if format == pdf:
    run pdf_inspector preflight -> if pass: parsers/pdf.py; else: fallback to calibre
elif format == epub:
    parsers/epub.py -> delegates HTML parsing to parsers/html.py for content.opf chapters
elif format == html:
    parsers/html.py -> text decoding
elif format == docx:
    parsers/docx.py
elif format == rtf:
    parsers/rtf.py -> text decoding
else:
    calibre conversion (universal fallback) -> re-dispatch on converted format