book-to-skill: Turn Documents into Agent Skills

Overview

book-to-skill is a Python tool that extracts content from books and other documents, then guides an AI agent to turn it into a structured skill. The result organizes key ideas and chapter references so compatible agents can retrieve relevant material when asked, rather than requiring users to search a source file or include all of it in every prompt.

It is most useful for material you revisit during study or work, such as technical books, internal documentation, research papers, and specifications. Extraction happens locally, while any text sent to a cloud-hosted model follows that provider's data terms.

Key Features

• Accepts individual files, folders, globs, or lists of sources.
• Supports PDF, EPUB, DOCX, HTML, RTF, MOBI, TXT, Markdown, reStructuredText, and AsciiDoc.
• Creates a skill with a chapter index and separate chapter files for on-demand reference.
• Organizes extracted material into reference files such as a glossary, patterns, and a cheatsheet.
• Targets hosts that support the Agent Skills format, including Claude Code, GitHub Copilot CLI, Amp, Hermes Agent, OpenCode, and OpenClaw.
• Includes analyze-only, generate-from-analysis, and update or fold-in workflows.
• Offers a standalone extraction CLI and a command to check available format extractors.

Use Cases

• Developers can turn a technical book into a reference skill they query while implementing a related system.
• Teams can combine internal runbooks, architecture decisions, or onboarding documents into a skill for day-to-day work.
• Researchers can organize a collection of papers and notes into a unified reference, then fold in new sources over time.
• Practitioners can make specifications, standards, or design guidelines easier to consult without repeatedly searching long documents.

Getting Started

Install the skill through the skills CLI:

npx skills add virgiliojr94/book-to-skill

For installation options, extractor dependencies, and usage examples, see the README.

Alternatives

• karpathy-llm-wiki: It builds a cited, structured wiki from sources for LLMs, rather than converting books into skills for coding agents.
• Article-Assistant--RAG-Telegram-Bot: It makes PDFs and other sources searchable through a Telegram RAG bot, rather than packaging their content as agent skills.

Considerations

• PDF extraction depends on installed tools. Technical books with code, tables, or formulas may benefit from Docling; simpler text-heavy PDFs can use tools such as pdftotext. Other formats may also require optional extractors.
• Scanned PDFs need OCR before extraction. The project says it detects PDFs without a text layer and stops rather than producing an empty result.
• The generated skill is synthesized reference material, not a verbatim copy. The repository advises keeping skills derived from copyrighted books private and respecting source licenses.
• Processing is local, but text provided to an agent may be sent to its model provider according to that provider's terms.
• The reported token savings are project measurements, not a guarantee for every document or query. Consult the project's performance documentation for its methodology.