@theredsix and you should collaborate.

Your tool's method of returning element references is clever and should greatly improve llm handling of the page components (and greatly reduce token cost).