AI Agent Detection Rule Refinement Plan

Analyze false-positive/negative patterns by segment, recommend prompt logic and confidence threshold changes, track precision and recall quarter-over-quarter; reference confidence floors, evidence capture, versioned prompts