Uncle Bob just put out a pretty cool viewer for proposed changes from agents here. I broke down what he has built and it's system for improving design and planning with agents. The first image describes his design and the system he is proposing. It's super cool and should make designing with agents better, especially when evaluating proposed changes. The one big quibble I have is the current implementation is coupled to Grok. I believe that something like this should be able to layer onto any harness, and should be extensible. The second image shows my proposal for how it could be made harness agnostic. I'm building out a similar implementation inspired by his on my harness today, which already has a custom plan card and flow and could really benefit from a dedicated visual flow. I would love to make an issue and propose a change to Bob's library to drive compatibility like this in the future. Then I could refactor my harness to use it instead of maintaining an inspired copy.