htbusers/HITL-Loop ? reverse-engineered prompt

Build me a Python library called LORE that helps catch unclear or logically shaky parts of a user’s question before an AI answers. I want it to look at an input, find the exact phrase that is causing the problem, keep the original wording, and decide whether it can safely continue, should abstain, or needs human review.

Please include a simple analyze flow, a small command line tool, and a basic HTTP API if that fits naturally. It should support a few deterministic checks for things like unsupported assumptions, contradictions, ambiguity, and obvious time related confusion. When it flags something, give me a review package with the original text, the highlighted span, some context, a confidence score, and a clear reason for escalation.

Make it work offline if possible, keep the core lightweight, and add tests plus a couple of examples so I can see it in action. If you need to look up current docs online while building, go ahead.