bash

AI Wrote Bad Code, Ran rm -rf: Time to Reckon with Agent Permission Safety

A dev approved an LLM's rm -rf "fix" for its own bad bash commands. When AI has execution rights, its self-repair can be deadlier than the initial err