Adversarial prompt-injection benchmark for LLMs - View it on GitHub