A project manager designs an agent that auto-labels incoming GitHub issues as bug, enhancement, or question and assigns priority. Teams disagree on urgency and mislabeled security reports slip through. Which specification of decision-making criteria will make the agent behave consistently?

The correct specification succeeds because consistent agent behavior requires all three decision elements together: clear label definitions with examples resolve ambiguity between bug, enhancement, and question, explicit priority thresholds resolve team disagreement on urgency by mapping severity to action, and security override rules prevent mislabeled safety-critical reports from being deprioritized. The specification that defines labels and severity-based priority thresholds but omits security handling fails on precedence, since severity alone cannot guarantee that a security report labeled as question still escalates correctly. The specification that defines labels with escalation workflows for sensitive reports but omits measurable priority thresholds fails on calibration, since teams will still disagree on urgency without shared cutoffs linking severity and impact to priority levels. The specification that sets severity-based priority thresholds and urgent security overrides but omits label definitions fails on classification, since priority logic cannot fix systematic confusion at the initial bug-enhancement-question decision point. An exam caveat is to select the answer that fully satisfies every stated pain point rather than the most sophisticated partial workflow, as tempting industry-standard triage patterns often look complete while dropping one requirement. A practical check is to test the prompt with edge cases such as a low-severity-worded security issue and a high-urgency feature request to verify that labeling, priority assignment, and override each trigger as intended.