Oh, this is interesting! I assume most of the heavy lifting to identify and filter content is being performed by the content script. In practice have you see any issues with models accessing page content before the extension has been able to sanitize the page?