Build a skill from PDF(s) via vision

This skill is the playbook for converting one or more PDF reference documents into a high-quality Claude Code skill. The output is a skill folder where:

• Source PDFs are bundled (the skill is self-contained and portable)
• Each chapter/section is a structured markdown file (real headings, tables, blockquotes, inline diagram references)
• Diagrams are extracted as PNG and named by content
• SKILL.md has frontmatter triggers, navigation, quick-lookup tables, and "common questions → file" pointers

The defining technique is vision extraction via a general-purpose subagent per chapter. Reading PDF pages as images and writing structured markdown gives dramatically higher quality than pdftotext for any document with tables, multi-column layouts, callouts, or diagrams.

Target skill structure