Hacker News new | past | comments | ask | show | jobs | submit
LLM are not designed for that. Unless prompt injection is really solved I would avoid showing LLM untrusted code.