Sounds great, what I wonder is how you treat edge cases that your Ai requirement phase identifies. Do you let Ai decide how to handle them or do you jump in and review/adjust manually. I'm currently getting my hands dirty with spec kit and feel like I write very detailed specifications. Still edge cases occur regularly where I then adjust the spec or plan.

In my requirement review and design review phase, I specific that the agent need to review edge cases and come up with 3 suggestions. So that I can choose. In case it is not match my expectation I will tell it to remember the case and provide answer for it to remember.

Nice, maybe I'll give devkit a try to compare it to the spec kit approach.

Appreciate it, I know there are so much room to improve, looking forward to hearing feedback from you.

I'm building an alpaca breeding and show management SaaS which is the first one to combine animal register with a strong approval workflow (think like a pull request), herd management and show management. No Ai involved but might be something I'll put on the roadmanp to help with breeding decisions. (alpacuna.com)