Agent is midway through a long migration and you already know the two follow-up steps you want after it. What can you do with the chat input right now?

The principle — Typing a next instruction while Agent works adds it to a queue rather than interrupting the run; queued items hold their order and start once the current task ends.

Why the key is correct — Typing the next steps while the current one runs is the intended workflow, and it is what makes a long autonomous run worth starting. The queue preserves order and the same conversation context.

Why the others are wrong — The input is not disabled during a run, so there is no reason to sit and watch. An ordinary submission queues rather than interrupting, though an immediate send exists separately for genuine redirects. A second chat would throw away the context the first has built up.

Remember this — Stack the next instructions while the current one runs; the queue holds them in order.

Sources — Cursor's Agent overview documentation.