Assessing Automated Prompt Injection Attacks in Agentic Environments