book-to-skill 🧩 Skill Open source

What it does

book-to-skill turns a technical book, or any folder of long documents, into an Agent Skill your coding agent can consult while you work. You run /book-to-skill ./my-book.pdf. A Python extractor pulls clean text and chapter structure from the file, and your agent then follows the skill's instructions to write a structured skill from it. The result is not a summary. It captures the book's frameworks, decision rules, anti-patterns and chapter contents, so you can later ask something like /my-book replication and get an answer grounded in the relevant chapter.

What is inside

Each generated skill contains:

• a SKILL.md with the core mental models and a chapter index
• one file per chapter, loaded only when needed
• a glossary, a patterns file and a cheatsheet of decision tables

The converter reads PDF, EPUB, DOCX, HTML, RTF, MOBI (through Calibre), Markdown and plain text. It can merge several sources into one skill and add new material later. It also has an analyse-only mode and can publish a finished skill to a private GitHub repository. The repository includes a validator for generated skills, a token-cost measurement tool and a pytest suite.

Works with

The README names GitHub Copilot CLI, Amp, Claude Code and Hermes Agent. Generated skills go to the shared ~/.agents/skills/ folder, which Codex also reads, with a symlink for Claude Code. Any agent that follows the Agent Skills standard can use the output.

How to install

npx skills add virgiliojr94/book-to-skill

Or clone it into your skills folder:

git clone https://github.com/virgiliojr94/book-to-skill.git ~/.claude/skills/book-to-skill

Maintenance and safety

MIT licensed, with frequent tagged releases and a changelog. Extraction runs locally with Python, and optional packages such as docling or pypdf can improve results. Text you convert is still sent to your agent's model provider. The tool ships no book content. The README warns against sharing skills made from copyrighted books, so keep those private.

Who should use it

Developers who own reference books, internal docs, runbooks or specs and want their agent to apply that knowledge accurately instead of guessing or reading whole PDFs into context.