Evaluate agent thread checkpoints, collaboration trace, human follow-through, AI self-assessment, handoff quality, and thread closeout.
