Ralph in Specify CLI #1490
Description
Hello All,
I am relatively new to spec-kit and still learning how to properly use it. I arrived here on my journey to learn more about ralph loops and context rot. I found some ralph implementations around and played around with it with my own prompted PRDs with some limited success. After learning a bit more I realized how important specifications and validations up front were to the success of a ralph loop. This led me to spec-kit, where up front specification is the main goal. However, the existing ralph loops that I was using were quite specific to the style of PRDs, task files, or progress files that the creators of the various ralph implementations expected. This prompted me to try and build a specify.implement agent designed for ralph loops. Since the main idea of a ralph loop is to execute outside of a chat session so that you can entirely control the context of each run, naturally I thought executing the agent from the Specify CLI would make the most sense.
Does this functionality make sense at all for spec-kit and specify CLI? Looking for feedback on this. I have a working POC that I have been testing with for copilot and powershell. Before I spend anymore time trying to support more AI agents, I was hoping for some feedback.
Here is a draft PR with what I have so far, of course I used spec-kit and have left all the specify files in there for now :)
#1489