Keeping your backlog flush means always having work queued so your agents are never idle — especially before you step away. This article covers how to read what needs attention and how to queue work so nothing sits still.
Read the indicators
Keep half an eye on the status area. It tells you what's happening without your having to dig:
- A plan waiting, a pipeline running, a finished pipeline you haven't opened, and a task that's done and needs attention each show a distinct indicator.
- The Overview page collects what's unread — notifications you may have missed, and pipelines that are running or have finished but you haven't looked at.
- You get a toast when something needs your attention, so you can step away and come back without losing the thread.
Queue work before you step away
Before you log off — an evening, a weekend, or just a break — fire off a few things so work continues while you're gone. Pipelines are the usual choice because they run the longest, but a chat works too. When you come back, you check how each one went and queue the next round.
Don't let anything sit idle
Keep every item moving to its next step — an implementation running, or a code review running on what's already built. If something is just sitting with nothing happening to it, that's backlog going stale. The goal is simple: there should always be something in flight.
Mind the bottleneck: testing
Once everything is moving, the bottleneck that tends to remain is testing. Build a strong, automated verification framework so agents and CI can confirm their own work — otherwise the queue stalls at "is this actually correct?" See Build a verification framework.
Next steps
- Working in parallel — the workflow this habit supports.
- Build a verification framework — clear the testing bottleneck so the queue keeps flowing.
- Watchers — get pinged when queued work finishes.