[jlongo78]

That sounds solid. The structured outputs per task are useful, especially the token tracking.

One thing worth considering: a lightweight timeline view that correlates those per-task outputs across parallel agents. When something goes wrong mid-workflow, jumping between container logs to reconstruct what happened concurrently gets tedious fast. Even a simple ordered event stream tying the branches and PR links together would save a lot of debugging time.

[gjkim042]

Get the idea.

I was wondering what’s the real scenario that needs agent chaining?

I implemented the task chaining using “dependsOn” field, but there’s no real usage now. Can we just write a pipeline prompt using the numbering? 1. Do this, 2. And do this?…

When do you think it’s useful to chain coding agents step by step?

[jlongo78]

Chaining really shines when steps have conditional branching based on previous output. A numbered prompt works fine for linear tasks, but falls apart when step 3 needs to decide between two different paths depending on what step 2 returned.

Also useful when steps have wildly different context needs - you don't want to stuff everything into one massive prompt. Keeping agents focused tends to produce cleaner outputs than one giant instruction list.