Open Source
Open Source

book-to-skill: Turn Any Technical Book Into an AI Agent Skill (16.2K Stars)

book-to-skill (virgiliojr94/book-to-skill, 16.2K GitHub stars, Python, MIT) distills technical books/PDFs/EPUBs/doc folders into structured agent skills following the open Agent Skills standard - install once, works across GitHub Copilot CLI, Amp, and Claude Code. Generates SKILL.md + per-chapter files + glossary + patterns + cheatsheet, with chapters loaded on-demand so they don't count against your token budget. Ships a benchmark tool measuring 24-51x fewer tokens than dumping the full book into context (tested on 3 real books). Beyond books: internal docs, brand systems, research clusters, specs. Includes a copyright-compliance note.

Published August 4, 20267 min read
<!-- book-to-skill-resource | resource | book-to-skill: Turn Any Technical Book Into an AI Agent Skill You Can Call On Demand -->

You buy a great technical book, read it once, and three months later you can't remember that chapter 7 covered the exact pattern you need right now. It's a universal developer pain point, and the usual fixes all fail: searching the PDF gives you a list of pages, not answers; asking your AI agent directly means it either hallucinates or says it doesn't have the content; taking notes by hand leaves you with a 200-line doc you never open again. book-to-skill offers a fourth path: turn the whole book into a structured agent skill, and load whichever chapter you need on demand.

What it is

book-to-skill (github.com/virgiliojr94/book-to-skill) is an open-source project by developer Virgilio Jr with 16.2K GitHub stars, 1,727 forks, primary language Python, MIT license, created 2026-05-01, last pushed in late July. One-line pitch: turn any technical book, document folder, or collection of sources into a unified agent skill that GitHub Copilot CLI, Amp, and Claude Code can study, reference, and invoke while you work.

The key is "structured skill" rather than "summary" or "dump the full text into context." It follows the open Agent Skills standard (github.com/agentskills/agentskills) - install a skill once and all three host agents read the same SKILL.md.

The 3-step workflow

text
/book-to-skill ./my-book.pdf

Step one, point it at a file, folder, or glob. Step two, the tool distills the book into a skill - frameworks, decision rules, anti-patterns, per-chapter files; structure, not summary. Step three, your agent loads it on demand:

text
/my-book replication

The agent reads the right chapter and answers from the real content, no hallucination. The book becomes part of your workflow.

What the generated skill looks like

Running /book-to-skill your-book.pdf produces a full set of files in your agent's skills directory (~/.copilot/skills/<slug>/ for Copilot CLI, ~/.agents/skills/<slug>/ for Amp, ~/.claude/skills/<slug>/ for Claude Code):

FilePurposeSize
SKILL.mdCore mental models + chapter index~4,000 tokens
chapters/ch01-*.mdOne file per chapter, loaded on-demand~1,000 tokens each
glossary.mdEvery key term with chapter refs~1,500 tokens
patterns.mdTechniques, algorithms, design patterns~2,000 tokens
cheatsheet.mdDecision tables and quick-reference rules~1,000 tokens

The core mechanism is on-demand loading: chapter files don't count against your token budget until you ask about that topic. SKILL.md holds only the skeleton and index, not the body.

Why it saves tokens: 24x-51x

This is the project's killer feature. book-to-skill ships a benchmarking tool, tools/discovery_tax.py, that measures the real token cost of answering one targeted question from a book, measured on real books:

BookContext dumpDiscovery loopbook-to-skillSavings
Think Python 2 (119K tokens)119,26412,152~5,00024x / 2.4x
Working Backwards (175K)175,25333,444~5,00035x / 6.7x
AI Engineering (256K)256,28777,866~5,00051x / 15.6x

Against a context dump it's a consistent 24-51x, and that cost recurs every turn. Against a one-time discovery loop it's 2.4-15.6x. The advantage scales with chapter size.

Why the gap is so large: a PDF-reading agent doesn't just read, it navigates - fetches the table of contents, notices a term it can't define, pulls more pages, backtracks. Every navigation hop lands in the conversation history and gets re-processed on every subsequent turn. book-to-skill pays that navigation cost once, at compile time; queries after that only read the relevant chapter.

Beyond books

The name says book, but the input is any structured prose. If you re-open a document often enough to wish you'd memorized it, it's a candidate:

  • Internal docs: architecture decision records, runbooks, onboarding guides. Fold a whole docs/ folder into one skill and query it while you code.
  • Brand & design systems: voice guidelines, component principles. Turn a brand book into a skill your team queries instead of skimming a 60-page PDF.
  • Research clusters: a stack of papers plus your notes, merged into a single skill, updated as new material lands (fold-in).
  • Specs & standards: RFCs, API contracts, compliance docs you reference but never memorize.

How to use it

text
/book-to-skill <file/folder/glob>... [skill-name-slug]

Supported formats: PDF, EPUB, DOCX, TXT, Markdown, reStructuredText, AsciiDoc, HTML, RTF, MOBI/AZW/AZW3. A few typical invocations:

bash
# Merge several files into one skill
/book-to-skill ~/papers/paper1.pdf ~/notes/export.txt unified-research

# Process all supported files in a folder together
/book-to-skill ~/workspace/project-docs/ project-knowledge

# Glob match
/book-to-skill "~/books/*.epub" my-library

# Fold new material into an existing skill
/book-to-skill ~/articles/new-paper.pdf ~/.claude/skills/project-knowledge

Takeaway

book-to-skill hits a real, high-frequency pain point: technical books and long docs that are "read once, then unfindable." Rather than duking it out with RAG vector retrieval, it uses a lighter structured-distillation-plus-on-demand-loading approach, moving navigation cost from every query to a one-time compile. The idea is elegant.

It suits two crowds: people who buy technical books, read docs, and do research and want that knowledge precisely callable inside an AI agent; and engineering teams with a pile of internal docs that need unified querying. The bar to entry is using a host that supports the Agent Skills standard (Copilot CLI / Amp / Claude Code) and spending one compile pass to convert a book into a skill. The risk is copyright - the tool itself is fine, but converting a copyrighted book into a skill for personal use is one thing, redistributing it depends on the book's license. The README is explicit: when in doubt, follow the source document's license; the project is a tool, how you use it is on you.

In the trend of AI agents treating "skills" as first-class citizens, book-to-skill is an open-source template for plugging existing knowledge (books, docs) into the agent skill system. 16.2K stars says developers genuinely need it.


References

This article is AI-assisted and human-edited. Last updated: 2026-08-04

Related

Open Source

video-shotcraft: Turn Claude Code Into a Cinematic Product Video Studio (3.5K Stars)

video-shotcraft (Vincentwei1021/video-shotcraft, 3.5K GitHub stars, TypeScript, Apache-2.0) is an agent skill that turns Claude Code or Codex into a motion-design studio - point it at your product and it storyboards, animates, and sound-designs a cinematic promo with Remotion. Ships 104 shot recipe cards, 161 motion previews, a validated 36.2s Ink Press template, 2.5D camera moves, beat-synced cuts, and film-grade SFX. Output is reproducible TSX, not black-box generation. Includes a China AtomGit mirror and 3 headless-rendering CI flags.

Aug 4, 20268 min read
Open Source

worldmonitor: The Open-Source AI Global Intelligence Dashboard at 79K Stars

koala73/worldmonitor is an open-source real-time global intelligence dashboard written in TypeScript under AGPL v3, with 79,487 GitHub stars. It uses AI to aggregate 500+ news feeds, geopolitical signals, market data, and infrastructure status; a dual map engine (globe.gl + deck.gl) with 56 layers, cross-stream correlation of military/economic/disaster/escalation signals, a Country Instability Index scoring 31 Tier-1 countries, local Ollama inference with no API key, six site variants from one codebase, a Tauri 2 desktop app, and 26 languages with RTL support.

Aug 7, 20269 min read