maddurvignesh/phishguard-ai ? reverse-engineered prompt

Build me a phishing URL detector web app that feels like a polished security tool. I want to paste in a website link and get back a clear answer like safe or suspicious, plus a short explanation of why it thinks that, using things like strange words, too many dots, short links, IP addresses, missing HTTPS, and other URL patterns. It should be fast, work from text only, and never open or crawl the site.

Make the app have a professional dark cyber style frontend with charts and a nice results page, and a backend API that serves the model quickly. Include a model comparison or playground view if that fits, and let me see simple stats or recent checks history. Use a real trained machine learning model, not hardcoded rules, and make the whole thing work end to end so I can run it locally right away. If you need to, look up current docs online for anything that has changed.