fivetaku/insane-search — reverse-engineered prompt

Reverse engineered prompt

GitHub

Build me a Claude Code plugin that helps Claude read websites that normally block simple scraping. I want to install it with no setup, no API keys, no accounts, and then just ask normal questions like “summarize this YouTube video”, “search X for posts about this topic”, “read this Naver blog”, or “find product info on Coupang”.

It should try easy methods first, then automatically get more stubborn when a site returns 403, shows a login wall, loads empty JavaScript pages, or uses bot protection. Try reader services, alternate URLs, mobile pages, RSS feeds, public JSON endpoints, browser like headers, TLS impersonation, and finally a real browser when needed. If useful helper tools are missing, install them quietly.

Please include special handling for common sources like X, Reddit, YouTube, Naver, GitHub, Wikipedia, arXiv, Stack Overflow, and news feeds, but keep the main behavior generic so it can work on unknown sites too. Return clean text, titles, summaries, metadata, links, and a clear message when real authentication is required.

Want more depth? Deep Reverse