Almost every example on this page involves concatenating trusted and untrusted text - which means they are susceptible to prompt injection attacks.
That's currently unavoidable: we still don't have a robust protection against this form of attack. But it's important when teaching people prompt engineering like this to at least mention the category of vulnerability, so that they can understand that there are situations where prompt concatenation cannot be used safely, and design their software accordingly.
That's currently unavoidable: we still don't have a robust protection against this form of attack. But it's important when teaching people prompt engineering like this to at least mention the category of vulnerability, so that they can understand that there are situations where prompt concatenation cannot be used safely, and design their software accordingly.