An Agent's Work Isn't "Done Later" Until You Can Find It Later
I have a habit of asking agents to pause instead of pushing through. Wait for a review. Wait for an external rate limit to clear. Wait for me to make a call I'm not ready to make yet. Pausing is not the problem. The problem is what happens to the paused work afterward. A few weeks ago I went looking for a task I knew an agent had finished setting up, and then couldn't find it. The output existed…
When delegating work to an agent, it's crucial to ensure that the task remains accessible later. If a task is paused, it often happens for a valid reason such as waiting for external input or human review. However, the challenge lies in how this paused state is recorded. Frequently, it isn't recorded at all, and instead resides within the memory of the person monitoring it.
Memory is not a queue, and it degrades quickly when attention shifts to other tasks. Consequently, a few days later, questions arise about the task's status, whether it has passed its checks or is still blocked.
The solution proposed is to record three key elements when a task is paused: the actual artifact produced by the agent, its current status, and the next action to be taken. The artifact should be the direct output or a pointer to it, not just a description. The current status should not only be "in progress" or "done," but include additional details like whether it's been reviewed or if there are any external constraints.
The next action should be specific, detailing the precise steps to resume the task. When all three elements are recorded together, a paused task can be quickly resumed without the need to reconstruct its state from memory. This method transforms the task into one that can be resumed with minimal effort, as opposed to relying on someone's memory from a specific moment in time.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
