Negative Conditioning

Is there any plan to implement negative conditioning? I use it differently than others across most models and find them to be extremely useful for consistency and winrates. The lack of it here imo prevents it from being easy to evaluate.

The issue names no files, tests, or entry points. Start by locating the conditioning path and defining how negative conditioning should behave across models; done requires an agreed implementation scope and validation criteria.

Written by the indexing model from the issue text.