This lesson's interactive features are locked, please to keep using them

Queueing

Steering blends your message into the task the agent is already working on, but sometimes you don't want that.

Say the agent is fixing a bug and you want the formatter run when it's done. Not a correction, but a follow-up. That's what queueing is for.

To queue a prompt, type it in, then press Ctrl+X before Enter to send it. The agent will finish its entire current task, then pick up the queued prompt and start working on that.

Steer: changes the work in flight (default)
Queue: waits until the current work is done, then sends the next prompt (Ctrl+X before Enter)

Assignment

Queue a follow-up while OpenCode fixes Sloperator's division-by-zero error. This SlopLang code:

hallucinate(1 / 0)✨

It should produce a nice SlopLang runtime error message, not a Python traceback.

Tell the model to fix the Python traceback from hallucinate(1 / 0)✨. It should report a SlopLang runtime error saying division by zero. Be sure it adds a test for this case.
While it's working, type another prompt: "Then run the error message tests and Ruff formatter." Press Ctrl+X, then Enter, to queue it.
Watch the follow-up get handled after the fix instead of during it.

Tell OpenCode to run and submit the CLI tests from the project root.