- Introduce task provenance tracking: agent_proposal, llm_backfill, controlled_fallback - Add quality computation function to classify workflows as 'AI-personalized' or 'guided baseline' - Calculate agent origin ratio, fallback ratio, and per-pillar coverage metrics - Implement configurable agent personalization threshold (default 35%) - Enhance plan metadata with comprehensive quality dimensions: - agentOriginRatio, agentOriginPercent, agentOriginTaskCount - agentOriginPillars, fallbackRatio, fallbackPercent, fallbackTaskCount - totalTaskCount and configurable thresholds - Simplify task provenance metadata handling in sanitization - Add backfill logic for existing plans to populate missing quality metrics - Maintain backward compatibility with existing plan storage
11 KiB
11 KiB