All products

Review book arguments with AI

Examine a nonfiction manuscript’s claims and evidence, then keep verified findings as traceable review notes without rewriting the manuscript.

Review book arguments with AI through a process that checks an objection before treating it as a defect. The review-ideas skill offers an editor mode with project context and an outside-reader mode in a fresh review room. Serious findings must survive comparison with the manuscript and available source evidence.

The result is review notes, not rewritten chapters. Previous author decisions and open notes prevent repeated findings. Prose editing and answering those notes belong to separate workflows; this pass concentrates on ideas, consistency and evidence.

What it does

  • Map a nonfiction book or paper’s main claims
  • Identify definitions that need closer attention
  • Compare serious objections with source evidence
  • Flag inconsistencies across chapters
  • Attach findings to quoted passages
  • Arrange a fresh-session outside-reader review

Install and use

  1. Use the complete book plugin in Claude Code with Python3.9+. Configure book.yaml with the manuscript, source and work paths. A standalone copy of this leaf omits required shared libraries and note tools.
  2. Try: “Review this nonfiction book’s argument in editor mode. Leave the manuscript unchanged and write serious findings only after checking the evidence.”
  3. For outside readers, prepare the review room and start a new session. This route uses several separate agent passes and can consume many tokens; agree scope and host costs first.
  4. Assess surviving notes yourself. Glossary additions need separate approval. Keep reports and run history, and distinguish reviewing, revising the manuscript and removing a completed temporary room.

FAQ

Source

Made by Igor Zarvanskyi. Original skill: review-ideas, published on GitHub under the MIT licence.

View on GitHub