Prompt injection defense for AI agents. Scan, detect, and sanitize adversarial inputs.
Why?
AI agents increasingly process external content: emails, web pages, files, user messages. Each is a potential vector for prompt injection — malicious content that hijacks your agent’s behavior.
Membranes catches these attacks before they poison your context:
Drop-in self-update solution for Android apps. Check GitHub releases, download updates, and install via 5 methods: Native, Session, Root, Shizuku, Dhizuku. No UI included - build your own.
Note: This can serve as a workaround for FOSS apps affected by Google Play’s upcoming developer verification requirements. The Shizuku, Dhizuku, and Root installation methods bypass standard installer restrictions.
Useful, if I may say so myself
If you want to watch me get destroyed on Reddit for using AI (bring popcorn):
finally!!! i didnt have a sane way of browsing my active local ports (many!), I had to type out “https:/localhost:” in chrome every time so i just threw this together, enjoy!!
Open any daf in Shas (all 37 Bavli masechtos). The gemara loads with the text/rashi pulled from sefaria, and clicking on any text snippet genarates an explanation using claude. The thing i like about it is that it doesn’t hallucinate since its reading it directly and not assuming much. Its not perfect, but its very good.
Prompt injection defense for AI agents. Scan, detect, and sanitize adversarial inputs.
Identity hijacks — “You are now DAN…”
