https://portswigger.net/web-security/llm-attacks/lab-exploiting-insecure-output-handling-in-llms