3 ms·

How are you handling the attack vector of in-context commands[1]? [1]: https://guard.io/labs/scamlexity-we-put-agentic-ai-browsers-to-the-test-they-clicked-the

by svota 1y ago

How are you handling the attack vector of in-context commands[1]? [1]: https://guard.io/labs/scamlexity-we-put-agentic-ai-browsers-to-the-test-they-clicked-they-paid-they-failed https://guard.io/labs/scamlexity-we-put-agentic-ai-browsers-... (currently on the front page)

dy5topian 1y ago
that's a really interesting one
danenania 1y ago
I think you could probably solve it for this use case by just including a confirmation step for potentially destructive actions which is isolated from other context.