POST a URL; get clean, LLM-ready content (Markdown/text/JSON) plus a prompt-injection risk verdict (clean | suspicious | flagged). A defense layer for agents and RAG pipelines, not a guarantee.