1natsu-document-harness-model

How to organize harness documentation starts with applying a three-layer model: instruction, index, and knowledge. You map content to CLAUDE.md, .claude/rules/, docs/, and feature READMEs using explicit placement rules that separate repository-wide guidance from per-package knowledge.

The three-layer documentation model for Claude harness references consists of instruction, index, and knowledge layers. It organizes internal references by defining clear placement responsibilities, preventing cross-talk between operational rules and background knowledge across the repository.

You need explicit placement rules for repository documentation when managing multiple reference points like CLAUDE.md and per-package docs. Without standardized placement responsibilities, cross-talk occurs between rules and background knowledge, making references ambiguous and unusable across skills.

Per-package documentation belongs in feature READMEs and specific docs/ directories, while repository-wide rules go in CLAUDE.md and .claude/rules/. Separating these placement responsibilities keeps internal references predictable and usable across the entire repository structure.

Yes, the harness documentation model works without external dependencies. It functions as a standalone reference framework that standardizes how content is organized across CLAUDE.md, .claude/rules/, docs/, and feature READMEs within your repository.

Cross-talk in harness documentation happens when placement responsibilities are ambiguous. Without a standardized three-layer model separating instruction, index, and knowledge across CLAUDE.md and docs/, rules and background knowledge overlap, making internal references confusing and inconsistent.