Retrospective as architectural quality checkpoint

[mikehentges]

As mentioned on Discord, the retrospective after the first epic is usually useful to me for identifying things that should have been clearer in the architecture documentation. Frequently, I've asked bmad "how can we prevent these findings in future epics/stories", and had it make updates to its documentation. One example update was to document that "all SQL needs to validate against actual table definitions in directory ./XXXX" - where I had put SQL definitions of existing tables - the dev agent kept guessing at column names and churning on resulting errors. After that first quality pass/improvement, the need to run other retrospectives goes down for me. For my corporate work, I've built out a reference architecture document to guide new development - but greenfield new architecture creation can leave some gaps that development of the first epic is good at fleshing out.