pixiebrix/agent-browser-shield — reverse-engineered prompt

Build me a Chrome browser extension that helps keep AI browsing agents safer and less distracted when they visit web pages.

The extension should clean up pages before the agent sees them, hiding things like cookie banners, footers, chat widgets, sponsored content, engagement rails, dark patterns, hidden text, HTML comments, and risky user generated content. It should also mask private info and credentials so emails, personal details, passwords, keys, or similar sensitive text don’t get sent to the model. Include a simple options page where people can turn individual protection rules on or off and save their preferences.

Please also make a small demo shopping site called RiverMart that intentionally includes these annoying and risky page elements, so I can compare the page with and without the extension. Add clear build, test, local run, and packaging instructions, including how to load it unpacked in Chrome and how to create a zip for agent browser platforms. Look up current browser extension docs online if you need to.