book-research-assistant — Claude Code skill: research for nonfiction

What book-research-assistant does

The skill runs in two phases. First it takes the Research Gaps Document from book-architect and turns each gap into a self-contained deep research prompt, building a Chapter Research Tracker with gap IDs, priorities and the kind of evidence needed. Later, when you bring results back, it goes gap by gap and renders a verdict: Complete, Needs More or Problematic. Each finished chapter gets a summary, and the skill certifies whether you are ready to move on.

It does not run the research itself. You execute the deep search outside, with Claude and Gemini, while the skill handles planning, prompt writing, validation and organization. Under the hood it leans on reference files for question formulation, evidence types, source evaluation and output format.

Upstream documents are required: Research Gaps Document, Master Architecture Document and Section Blueprints from book-architect, plus the Book Concept Document from book-ideation. Without them planning will not start. The skill also treats LLM research as something that can invent sources, so checking citations and confidence flags is part of the workflow.