Writing Books for AI Agents
Add to your library first to use in Claude Code
About
Load when authoring a new reference book for an AI agent, auditing a book for agent-readability, converting a human book to agent-readable form, or running a gap-to-book authoring pipeline. Meta-guide grounding 'a good book' in knowledge-management theory + agent-consumption mechanics; serves both marketplace books and internal company knowledge layers.
Preview
Writing Books for AI Agents
Load this book when:
- You are authoring a new reference book intended for an AI agent to read.
- You are auditing an existing book for agent-readability (is it navigable, findable, current?).
- You are converting a human-oriented book into an agent-readable one.
- You are running a gap → source → book authoring pipeline and need the style guide + acceptance test.
This is a meta-reference: a book about how to write books that AI agents can consume. It has two theory pillars and it obeys its own rules — it is its own worked example.
Pillar 1 — Agent-consumption mechanics. How an LLM actually reads, searches, and navigates a document, and how to structure pages so it can answer from one page without reading the whole book. Drawn from deployed agent-readable books: SEO for AI Agents, Code Review for AI Agents v4, Building Your Agent Team, Replicating Anthropic's Self-Service Analytics, the Anthropic Engineering Playbook.
Pillar 2 — Knowledge-management theory. What makes knowledge a valuable, auditable asset — independent of the reader being human or machine. Drawn from Knowledge Assets and Knowledge Audits (Handa, Pagani & Bedford), CandleKeep ID cmpvjpwpe0btool0z8qkysfak.
Every quality rule in this book serves two audiences: a public marketplace reference book (value accrues through circulation) and an internal company knowledge layer (value accrues through governed reuse). Where the audiences need different handling, the chapter says so explicitly.
How to use this book
- Read this page (the loading triggers and the matrix on the next page).
- Go to Chapter 2 — Decision Matrix. Find your task. Jump to the listed chapters. Do not read linearly.
- Each chapter is one page and self-contained. It restates what triggers it and cross-references other chapters by number.
- Before you ship a book, run Chapter 17 — The Self-Audit Checklist against it. The book is not done until it passes.
Each content chapter has a Narrative section (for a human deciding whether to adopt the practice) and an Agent rules section in RULE N / WHY / APPLY form (for an agent executing the practice). If you only have budget for one, the agent rules are load-bearing; the narrative is context.