book-to-skill
Turns a technical book into a chunked Claude Code skill: one main file holding the concepts, one file per chapter, a glossary and a cheat sheet. When you ask a question, Claude loads only the files it needs. The book stays available from one session to the next without ever filling the whole context window. Two same-named repositories exist under `book2skills`, they are not this project.
Strengths
- The book loads in useful chunks rather than whole, which leaves the context window room to breathe
- The generated glossary and cheat sheet also serve outside the agent, for human reading
- The knowledge outlives the session: the book is reusable without being re-pasted every time
Limitations
- Chunking is automatic, so a book whose thesis builds chapter by chapter loses its thread
- The announced token gain comes from the creator and compares against the worst possible case, not normal usage
- Two same-named repositories circulate under a close name, you have to check which project you land on
Best for
- A reference work you reopen monthly: a standard, an API manual, a language guide
- Giving an agent a stable theoretical base that will not move between two projects
- Onboarding onto a domain whose vocabulary you lack: the generated glossary earns its place from session one