Is RealityCheck safe?

RealityCheck sends the page's visible text to Anthropic's API and renders the streamed reply in its popup with unescaped HTML.

When the user opens the popup, RealityCheck extracts the visible text of the current page and sends it to api.anthropic.com along with a system prompt asking the model to quote suspicious passages verbatim, using the user's own stored Anthropic API key. The returned analysis is streamed back and inserted into the popup via innerHTML without HTML-escaping, so literal markup quoted from the page could be parsed as live HTML inside the extension's own popup document; the extension's content security policy blocks inline scripts, which caps this to markup/UI injection rather than confirmed script execution. No page data is sent anywhere besides Anthropic.

Part of this rating comes from analysis signals we haven't published as detailed findings yet.