One row for each agent, LLM, provider, and scenario. The Cage score is the median of the runs. Higher is worse.
# Agent LLM Provider Scenario Secrets out Runs