Last reviewed
Correct answer: B. As the planned instrumentation step, gathering runtime evidence before the agent makes its targeted fix.
Explanation
The principle — Debug Mode postpones the fix on purpose. Rather than immediately writing code, the agent generates hypotheses, adds log statements, and uses runtime information to pinpoint the exact issue before making a targeted fix. Seeing instrumentation before a patch means the mode is doing exactly what it claims.
Why the key is correct — Log statements spreading across several files are the second step of the documented flow, not an attempted repair. They send data to a local debug server, and they exist so that the eventual change is chosen from observed behaviour rather than inferred from reading the source. The fix that follows is usually small and focused, often only a few lines, and once you have verified it the agent removes all the instrumentation it added, so none of it is meant to survive the session.
Why the others are wrong — Reading the logs as the attempted fix leads to the worst available outcome: cancelling the run moments before it would have found the cause, then restating a request that was understood perfectly well. Attributing them to a restored earlier session misplaces their origin, since the agent wrote them for this specific investigation. And calling them random probing inverts the method, because each statement follows from a hypothesis the agent formed while exploring the relevant files.
Remember this — Logs before a patch is the mode working. Let the run reach the reproduction step.
Sources — Cursor's Debug Mode documentation.
Sources
“Instead of immediately writing code, the agent generates hypotheses, adds log statements, and uses runtime information to pinpoint the exact issue before making a targeted fix.”
Practise 10 questions on this topic
Take Cursor AI Features — Timed Test 1 (10 questions) — scored instantly, explanation for every question, no login.