Which metrics help trading bots spot rug pulls?

It's usually hard to spot red flags in new coins, like sudden liquidity drains or sketchy token movements, and while trading bots can help spot some of these "rug pulls," they're not perfect. So, I'd like some advice on what specific metrics I could program my bot to monitor? I thought about tracking locked liquidity, unusual wallet activity, or big sell-offs, but I still can't tell which one works best. If you've built or customized a bot to detect these risks, I'd love to hear what's worked for you. I still use some third-party trading bots like [Bananabot](https://bananagun.io/) for meme coins, but I still want to understand how they (in theory) can identify rug pulls.