Mock article — Editorial demonstration, not a published research result.
Follow the evidence
Read a trace as a sequence of observations and decisions. At each step, ask what information was available and what the agent did with it. Avoid judging an earlier decision using information that arrived later.
Find the first meaningful deviation
Several visible errors may share one cause. Locate the point where the run first moved away from the task, then follow its effects. This is often more useful than counting every downstream symptom.
Separate diagnosis from speculation
A trace can show that a tool returned an error or that a requirement was skipped. It may not explain why. Label explanations as hypotheses until another observation or a targeted test supports them.