A question for those using #ai to code. Does your team review your prompts, and do you check them into source control?

#vibeCoding #claudecode #codex #codereview

@jongary This has been something we've discussed at work. We've recently committed some skills for some common tasks, but otherwise the team has a really cautious approach to it.

I personally think you absolutely should, it's useful to approach the problem(s) consistently no matter who is working.

@tmcarr seems like the ai tools are like compilers, so the prompts are part of the source.

@jongary I'm not sure that a compiler is the right metaphor. It lacks the deterministic output that a compiler should have.

But yeah, I agree, it's very related to the source that is in the repo if it's helping write it.... Plus, if someone else wants to work on the same codebase, they're at a huge disadvantage from the get-go if they dont have the context to get the LLM firing on all cylinders from the start.

@jongary no, nobody wants to see me swearing and insulting Claude when it does stupid shit lol. At most I’ll commit some markdown plan files though. Considering a prompt doesn’t reliably produce consistent results not sure it’s that valuable to commit them like source.