What should your agent do? Plain language is fine.

I'll ask follow-up questions — inputs, output, which of your connectors it should read, its rules and a name — until it is clear, then show you the finished agent to create. Nothing is created until you press Create agent.