Comment by jbvlkt

LLM are not designed for that. Unless prompt injection is really solved I would avoid showing LLM untrusted code.