I've seen a bunch of AI Agents + automated trading demos again these days, pretty fancy, but the security part is basically brushed over... My kind of paranoia just wants to draw the red line first: never put mnemonic phrases into web pages or plugin pop-ups, even if they look exactly like the official ones; don't blindly click "Confirm" on signature authorizations, if you don't understand it, stop. Especially those that ask you to give unlimited permissions right away, which is basically handing over a master key to the other party.

Should I hand over my assets to an Agent for automatic interactions?
I won't, anyway. If I really want to try, I’ll only use a small account with a small amount, and revoke the permissions immediately after use.

And about phishing sites, the most disgusting thing is "redirecting to a nearly identical domain." I now habitually bookmark my frequently used links, and I never click on links I find through temporary searches. I also avoid looking at cross-chain bridges, because if the patching process is slow, I definitely won't give large permissions... That's it for now. When it comes to security, I’d rather be a bit more troublesome.