The problem lies in surveillance and others understanding what you did. Say your security department records every shell interaction with prod services: how does one then review and understand what happened? This is a fairly tricky problem. Perhaps through it at an LLM, but it'd have to be well trained to look for malicious actions.