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 catch it with broader validation than the most local unit test. Adjacent behavior often breaks because the patch matched one failing test but missed the surrounding contract of the module or workflow.

So I want targeted regression tests, related integration...

Upgrade to view official answer

Related Questions