A shell command the agent wants to run could be useful or destructive. How would you decide?

Instruction: Explain how you would judge whether to allow a risky command.

Context: Tests how the candidate diagnoses the problem, chooses the safest next step, and reasons through recovery. Explain how you would judge whether to allow a risky command.