How best to use this thing

[jpcoder2]

I have been playing around with spec-kit a good bit. I love the general idea of it but I have yet to generate any fully working code with it. I have mainly been using the sample prompts from the readme to test this. I've tried several different models and combination of models but have not found anything to really work well yet. I have done all my testing in VS Code.

Questions I have:

1. What model or models works best with this workflow? I've tried GPT 4.1, Grok Code Fast 1, Claude 3.7 and GPT-5. I've tried a combo of GPT-5 for the planning and Grok to implement the tasks once as well.

2. After spec, plan and tasks are created, what is the best way to prompt for implementing the features? This does not seem to be clear to me in the docs. I've tried asking it to implement the plan doc and asking to implement sets of tasks. Neither has been completely effective.

3. I've seem models struggle to build tests and know when to run them during the implementation process. Is there a way to make this run smoothly?

4. Does anyone have a full walk thru using this system including model selections, that is proven to work well?

Thanks for your help. Again, I really like the direction this is going.

[fabiodouek]

Answering your question 2: There is a new command now: /implement