You hand off three unrelated tasks by entering three prompts, one after another, and go and do something else. An hour later two have made progress and the third has not moved at all. What is the situation?

The principle — a session you dispatched is not a message in a queue and not a colleague who will chase you. It is running, it costs, and if it needs you it waits.

Why the key is correct — three documented facts, one for each part of the picture. On identity: every prompt you enter here starts its own new session, so three prompts is three sessions rather than one with a backlog. On cost: each session uses your subscription quota independently, which is the price of the arrangement and the reason to dispatch what is genuinely independent — use agent view when you have several independent tasks Claude can work on without you watching every step. And on the stall: Claude is waiting on something only you can provide: an answer to a question, a permission decision, or another prompt only you can answer. The task runs autonomously, so it can't ask clarifying questions — which is why an hour passed with nothing happening and nothing arriving to tell you so.

Why the others are wrong — the prompts did not queue in one session. The quota is not shared out between them. And a stalled session will not come and find you.

Remember this — one prompt, one session, one draw on your quota — and a stall is a thing you have to go and look for.

Sources — Anthropic's Claude Code documentation on managing multiple agents, and on common workflows.