Your approval flow blocks too many harmless tool calls and users stop trusting the assistant. What would you redesign?

Instruction: Describe how you would rebalance approvals when they are too broad.

Context: Tests how the candidate diagnoses the problem, chooses the safest next step, and reasons through recovery. Describe how you would rebalance approvals when they are too broad.

Official answer available

Preview the opening of the answer, then unlock the full walkthrough.

I would redesign approvals around action risk, not around tool category. If harmless tool calls are blocked constantly, the policy is probably too coarse and the assistant starts feeling timid rather than trustworthy.

I would identify which actions are truly sensitive,...

Upgrade to view official answer

Related Questions