A survey plus a public compendium tries to fix the fact that "agent" has no standard definition for evaluation.

arXiv 2609.11018 organizes agent evaluation around five dimensions of agenticness: environmental interaction, learning and adaptation, autonomy, goal-directed behavior, temporal coherence. It ships the Agent Compendium at agent.duketrustlab.com. Useful mainly as a map of where evaluation is thin, identified per dimension instead of asserted in general.