ERA OF MINDS

What is book-to-skill? Turn technical books into Claude Code agent skills

By Era of Minds ·

Short answer

book-to-skill is an open-source MIT tool that turns a technical book, PDF, EPUB, DOCX, folder or glob into a structured Agent Skill your coding agent can query by chapter. It writes a SKILL.md plus on-demand chapter files, a glossary, patterns and a cheatsheet. The README's measured claim: 24×–51× fewer tokens than dumping the book into context for one question. Works with Claude Code, Copilot CLI, Amp, Codex, Hermes Agent, OpenCode and OpenClaw. About 35,009 GitHub stars and 3,679 forks as of October 16, 2026.

Era of Minds cover: 'Turn books into agent skills.' book-to-skill open-source tool by virgiliojr94.
Figure: banner art from the virgiliojr94/book-to-skill README (MIT). Stars as of Oct 16, 2026.

What is the book-to-skill Claude Code tool?

You buy a technical book, read it once, then three months later you cannot remember which chapter covered replication. Searching the PDF gives page hits, not answers. Asking an agent about the book without the text usually ends in a confident guess.

book-to-skill (by virgiliojr94, created May 1, 2026) solves that by converting the book into an Agent Skill: structure, not a summary. Named frameworks, decision rules, anti-patterns and per-chapter files. Your agent loads the right chapter when you ask.

On October 3, 2026, @tom_doerr posted about it on X. When we checked on October 16 that post had about 37,254 impressions, 710 likes, 894 bookmarks and 74 retweets, and it was still circulating mid-month. Docs and install guides also live at booktoskill.is-a.dev.

How do you install book-to-skill?

Fastest path (agent skill CLI):

npx skills add virgiliojr94/book-to-skill

Claude Code via git clone (pulls the modular engine files):

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

Or paste this into a Claude Code session so it installs from the raw skill file:

Install book-to-skill: https://raw.githubusercontent.com/virgiliojr94/book-to-skill/master/SKILL.md

Then convert a book:

/book-to-skill ./my-book.pdf
# or with a name
/book-to-skill ./my-book.pdf my-book-slug

After it finishes, query the skill like any other:

/my-book-slug replication
/my-book-slug ch05
/my-book-slug "what chapters do you have?"

Docs at booktoskill.is-a.dev also list a standalone pip CLI for extraction only. That path installs the text extractor; it does not register the /book-to-skill agent command. Use the git clone or npx skills add path for the slash command.

What does book-to-skill generate?

A run writes a skill folder under the cross-agent skills root (typically ~/.agents/skills/<slug>/). Under Claude Code it also tries a verified symlink at ~/.claude/skills/<slug>/. Typical outputs from the README:

  • SKILL.md — core mental models and chapter index (~4,000 tokens)
  • chapters/ch01-*.md … — one file per chapter, loaded on demand (~1,000 tokens each)
  • glossary.md — key terms with chapter refs (~1,500 tokens)
  • patterns.md — techniques, algorithms, design patterns (~2,000 tokens)
  • cheatsheet.md — decision tables and quick rules (~1,000 tokens)

Chapter files do not count against the skill budget until you ask about that topic. You can also fold a whole docs/ folder, brand guidelines, research papers or RFCs into one skill, then update it later by pointing /book-to-skill at new material and an existing skill folder.

How does the 24×–51× token claim work?

The README calls this the Discovery Loop Tax. A PDF-reading agent does not just read; it re-fetches the table of contents, backtracks and re-processes pages on every turn. book-to-skill pays the structuring cost once at conversion.

From docs/performance.md, tokens to answer one targeted question:

  • Think Python 2: context dump 119,264 vs book-to-skill ~5,000 (24×)
  • Working Backwards: 175,253 vs ~5,000 (35×)
  • AI Engineering: 256,287 vs ~5,000 (51×)

The ~5,000 figure is a resident core (~4K) plus one compiled chapter (~1K). Full conversion cost in their Sonnet 4.5 estimate table is roughly $1 per book, paid once.

Which agents does it work with?

Any host that reads the open Agent Skills SKILL.md format. The README and site list Claude Code, GitHub Copilot CLI, Amp, Codex, Hermes Agent, OpenCode and OpenClaw. Formats supported: PDF, EPUB, DOCX, Markdown, HTML, RTF, MOBI/AZW and more.

If you already run Claude Code skills for other jobs, this sits next to Taste-Skill (landing-page design), Anatomy (isometric machine explainers) and tools like ManimGX or our roundup of AI tools for coding.

What are the limits?

Generated chapter files are synthesized from the source. Skills built from third-party copyrighted books should stay private; the converter's optional GitHub publish path defaults to a private repo. Some books without clear "Chapter N" headings need help with segmentation (the README notes Pro Git as an example).

Extraction mode matters for technical PDFs: fast pdftotext flattens tables and code; Docling technical mode is slower but keeps them. And like any skill, you still review the output. The tool structures the book. It does not replace reading the parts that matter.

If Claude Code token spend is already high, the Claude Code cost calculator helps compare API spend with Pro and Max. Want prompt ideas for other agent jobs? Grab our 50 free AI prompts that actually work.

Key facts

  • Product: book-to-skill, an open-source converter that turns books and docs into Agent Skills for Claude Code and other hosts.
  • Author / repo: virgiliojr94/book-to-skill, created May 1, 2026. MIT. About 35,009 stars / 3,679 forks as of October 16, 2026.
  • X buzz: @tom_doerr, Oct 3, 2026 (~37.3k impressions / 710 likes / 894 bookmarks / 74 RTs when checked Oct 16).
  • Install: npx skills add virgiliojr94/book-to-skill or git clone … ~/.claude/skills/book-to-skill. Usage: /book-to-skill ./my-book.pdf.
  • Token claim: 24×–51× fewer than context-dumping a book (docs/performance.md).
  • Docs: booktoskill.is-a.dev.

FAQ

What is book-to-skill?

book-to-skill is an open-source MIT tool by virgiliojr94 that turns technical books, PDFs, EPUBs, DOCX files, folders or globs into a structured Agent Skill (SKILL.md plus per-chapter files, glossary, patterns and cheatsheet). Your agent loads chapters on demand instead of dumping the whole book into context. It works with Claude Code, GitHub Copilot CLI, Amp, Codex, Hermes Agent, OpenCode and OpenClaw.

How do I install book-to-skill for Claude Code?

Quick path: npx skills add virgiliojr94/book-to-skill. Or clone: git clone https://github.com/virgiliojr94/book-to-skill.git ~/.claude/skills/book-to-skill. Or paste into a Claude Code session: Install book-to-skill: https://raw.githubusercontent.com/virgiliojr94/book-to-skill/master/SKILL.md. Then run /book-to-skill ./my-book.pdf.

How many fewer tokens does book-to-skill use?

The README and docs/performance.md claim 24x to 51x fewer tokens than dumping the book into context to answer one question. Measured examples: Think Python 2 about 24x, Working Backwards about 35x, AI Engineering about 51x. The skill loads a resident core of about 4,000 tokens plus one chapter of about 1,000 tokens, roughly 5,000 tokens per targeted question.

Is book-to-skill free and open source?

Yes. The virgiliojr94/book-to-skill repo is MIT-licensed. As of October 16, 2026 it has about 35,009 GitHub stars and 3,679 forks. Docs live at booktoskill.is-a.dev.

What formats does book-to-skill support?

PDF, EPUB, DOCX, TXT, Markdown, reStructuredText, AsciiDoc, HTML, RTF, and MOBI/AZW/AZW3. You can also point it at a folder, a glob, or a list of paths and fold them into one skill.

How is book-to-skill different from Taste-Skill or Anatomy?

All three are open-source Agent Skills you install once and call from Claude Code or similar hosts. Taste-Skill designs landing pages with art direction and a browser check. Anatomy builds interactive isometric machine explainers. book-to-skill turns books and docs into structured skills your agent queries by chapter. Different jobs: frontends, diagrams, and book knowledge.

Sources

Keep reading