Hey PH! 👋 Britt from PixieBrix here. We've been building browser tooling for enterprise teams for years, and when we started seeing AI agents get deployed at scale to browse the web, we noticed a gap: agents have zero protection from the stuff (most) humans have learned to watch out for. Prompt injection is OWASP's #1 AI security threat — and it's trivially easy to embed hidden instructions in a webpage that your agent will follow without question. On top of that, most pages are full of junk (cookie banners, footers, chat widgets) that your agent reads and pays for in tokens. Agent Browser Shield is our answer: a free, source-available browser extension that strips all of that before the model sees the page. We built it in the open because this is an evolving problem — new dark patterns, new injection techniques — and no single team can keep up alone. Would love your feedback on what to build next. And if you're running browser agents in production, let us know what failure modes you've hit because we want to keep building and write better rules to make this even better. GitHub: https://github.com/pixiebrix/age...

Agent Browser Shield

Block prompt inject & cut token costs for AI browser agents

Block prompt inject & cut token costs for AI browser agents