LLM email summarization

Invisible email spans reach the summarizer because text extraction never checks visibility

The hidden-keyword trick that once fooled Bayesian spam scoring now lands inside an LLM's context window. The one component that could strip it is deliberately built to leave styling alone.