Most days I have three coding agents working on the same repo, each on its own branch. Starting them takes a minute. Reviewing what they did takes the rest of the afternoon.
The summaries they write at the end are confident and usually a little wrong. "Added tests" sometimes means one test. "Refactored the handler" sometimes means it also renamed a function three other files call.
What works for me so far: I read the diff before the summary, I run the checks myself before merging, and I merge one branch at a time, so a conflict shows up while I still remember what each agent was asked to do.
What still doesn't work: when two agents finish at the same time, I start skimming. And skimming is exactly when the renamed function gets through.
How do you review agent output? One branch at a time, a checklist, a second agent reviewing the first, or something else? I'd especially like to hear from anyone who has stopped reading the summaries altogether.
Top comments (1)
When you have three agents on three branches, how do you scope their initial tasks to minimize the blast radius and prevent them from stepping on each other's toes in the first place?
I'm new to dev.to! Any feedback or thoughts on my posts would be greatly appreciated.