Hacker News Reader

Can’t you just ask another LLM to analyse the text of the input to determine if it’s an attempted prompt injection?

That's a possible mitigation mentioned in the article.

Maybe I'll read it next time :D