A coding agent writes a plausible fix and breaks adjacent behavior. How would you catch that?

Instruction: Describe how you would protect against locally correct but globally harmful patches.

Context: Tests how the candidate diagnoses the problem, chooses the safest next step, and reasons through recovery. Describe how you would protect against locally correct but globally harmful patches.

Official answer available

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

I would not trust a local fix just because it looks reasonable. I want the agent to inspect adjacent behavior...

Related Questions