I've found myself having "Oh! Yeah!" moments where I need to tell Claude Code something important like (make sure you actually RUN the unit tests and ensure they pass before committing the code). I have to stop it, then tell it, then tell it to continue on and not to ignore the thing I told it "no" to in order to stop it.
I realized, there's no reason to do that. I can just open a NEW instance of Claude Code and tell it in that instance and to update it's rules or todo notes... You ARE having it create a todo list and/or rules or PRD document, aren't you? If not, get on that NOW!
This only works on AI tools that have external files that direct them on their work since 2 instances generally (at the time of this writing) can't communicate with each other.
Open TWO instances of Claude Code
By Mike
51 views
0