Clarification on the "Single Agent Implementation" requirement

Hello, I have a question regarding the rule stated on the homepage: “A single agent implementation must connect to all five and play each game in the same session without manual intervention.”

Could you clarify what constitutes a “single agent” in this context?

  1. Does this imply that we need to design one unified agent architecture (a general-purpose model) that handles all games simultaneously, rather than creating separate architectures for each game?
  2. Or, does this rule prohibit a “multi-agent system” where multiple agents collaborate to achieve the results?

I want to make sure I understand the constraints correctly. Thanks in advance.