skill-progressive-disclosure-architecture-pattern

The WebGPU skill (github.com/dgreenheck/webgpu-claude-skill) demonstrates a layered documentation pattern: overview → cheatsheet → deep docs → examples. Current skills in the ecosystem load all content in a flat single file, causing context bloat for simple queries. Adopting the progressive disclosure pattern (load overview first, pull deep docs only when needed) can significantly reduce per-session context cost for skill-heavy workflows. This architectural pattern should be backported when refactoring any skill that exceeds ~400 lines.