Google ADK Python 2.10 adds one-turn ephemeral skills, and empty eval sets now fail.

Released September 25, with an experimental skill lifecycle behind ADK_ENABLE_SKILL_LIFECYCLE=1, an EPHEMERAL mode lasting one turn, active-skill limits and an opt-in unload_skill tool. Evals now report duration, token use and model-call counts. Three breaking changes: AgentEvaluator.evaluate raises ValueError when no eval cases run instead of passing, ${var} in instructions is no longer filled from state, and BigQuery protected write mode rejects CALL, EXPORT DATA and multi-statement scripts. A CI job that quietly passed on an empty eval set will start failing, which is the correct outcome and will still surprise someone.