Is Your Prompt Safe? Investigating Prompt Injection Attacks Against Open-Source LLMs