What problem does it solve? AI coding agents waste tokens and make errors when repository documentation is bloated, duplicated, or disorganized. This Skill applies the Interpretable Context Methodology (ICM) to structure AGENTS.md, CLAUDE.md, and docs/ trees so agents load only the context a task requires. ## Core Features & Use Cases - Three operating rules: Enforces pointers over payload, one authoritative home per fact, and section-scoped references so routing files stay thin and facts never drift. - Layered context model: Maps ICM's five context layers to concrete repo artifacts like AGENTS.md, index.md catalogues, design docs, and exec-plans. - Honest limits guidance: Explains what CI freshness checks can and cannot enforce, and why stale canonical docs mislead agents more than no docs. - Use Case: When setting up a new repository, write AGENTS.md as a thin routing catalogue with a pointer table, add index.md files to directories with 3+ docs, and back load-bearing claims with tests instead of prose. ## Quick Start Restructure this repository's AGENTS.md and docs folder using the agent context layer skill so an AI agent can navigate the project through pointers instead of reading everything.
