Indirect prompt injection lets attackers bypass LLM supervisor agents by hiding malicious instructions in profile fields and contextual data. Learn how this attack works and how to defend against it.
Results that may be inaccessible to you are currently showing.
Hide inaccessible results