Agent Use Patterns

It's a tricky thing, managing so many agents. Too many things can go wrong! But it's also clear that there are a lot of ways things can go right. work queue + orchestrator polling agent/"cron job" message queue + responder what are the patterns that make using these things useful? memory + context management CLI TOOLS ARE A MUST CLI is ultimate ergonomics for text-based beings; if you can wrap a CLI over your workflow, the agents become so much better