Skip to content
OBSERVE. QUESTION. VERIFY.
01CASE FILE
Reproduction

First, make the failure repeatable.

A convincing explanation cannot rescue an unstable reproduction. Before asking an AI coding assistant for a fix, establish an input and sequence of actions that expose the failure. Capture what happened, what should have happened, and the environment in which you observed it.

Preserve the conditions

Record the relevant runtime, configuration, and repository state. Keep the smallest input that still triggers the problem, including empty fields or ordering when they matter. Remove private data without changing the property that causes the failure. Separate a user-visible symptom from the underlying error output.

Reduce one variable at a time

Try a nearby valid input as a control. Remove unnecessary steps and check that the failure remains. If the problem is intermittent, record the conditions of each attempt instead of pretending it is deterministic. A reduced reproduction should be simpler to run and still preserve the original behavior.

What if the failure will not repeat?

Keep the original report intact and mark the reproduction as unresolved. Gather environment differences and timestamps before changing code based on an unconfirmed explanation.

A question for your assistant

Read only. Help reduce this report to a reproducible case. Separate exact observations from suspected causes. Identify the smallest input, environment details, and comparison case needed. Do not propose a patch until the failure is sufficiently established.
CONTINUE THE INVESTIGATIONTrace the data