I love AI studio, but precision can be tricky. So whilst adding instructions for specific cases/branches in rules in the general Guidance for AI, they tend to bleed over into branches where they are not applicable.
If it would be possible to connect more detailed instructions to one or more specific branches (that would only be included/considered when a specific branch is chosen) this could really help making AI responses more reliable/precise.
Specifically, they would apply to the set of actions that is being executed after a branch is chosen.