What you want
Right now all the skills/commands relies only on Claude models, will be great to support other models (like opencode free ones) for operations like caveman-compress where there's not the need of a lot of "brain juice"
Before/after example
Before: compress.py script relies on `call_claude` function
After: all the skills read the provider/model to use from ENV (or similar)
Why good
Not all operations (like compressing) need high end models to be efficient/accurate, we can save $ using cheaper/free/opensource models, especially in third parties Harnesses (opencode/py/hermes...)
What you want
Right now all the skills/commands relies only on Claude models, will be great to support other models (like opencode free ones) for operations like
caveman-compresswhere there's not the need of a lot of "brain juice"Before/after example
Why good
Not all operations (like compressing) need high end models to be efficient/accurate, we can save $ using cheaper/free/opensource models, especially in third parties Harnesses (opencode/py/hermes...)