How to start using part-way through a project?

[philled]

I'm pretty new to coding with AI. I'm creating a web site for my wife's business and am using Opencode to build it. But I have a long background in IT and understand the importance of a structured approach with approval gates, so the OAC concept appeals to me.

But I'm worried about starting to use it on an existing project that is over half way complete. My current workflow is to define a PRD with the agent, ask it to create issues in GitHub, then ask it to code those issues. After that I test, ask it to create ADRs and update relevant docs, and then I close the issues in GitHub. My concerns are:

1. Will it alter code and break things that are already working?
2. Will it start doing things differently and I'll end up with two different types of code in my project (pre-OAC and post-OAC code) that will make future maintenance difficult?
3. What would be a good way to try it out without risk? Maybe install it and try adding a new feature to my project? Should I use a different branch from main?
4. If I try it and don't like it, how do I remove it from my project and revert to how things were operating before?
5. I notice that the agent definition MDs state which model the agent should use. Can this be blank so that it uses whichever model I've selected with Opencode /models?