Not all data involved in development belongs in the codebase.

Some data is purely contextual:

debug output

partial transformations

one-off examples

Checking this into version control creates more noise than value.

When I need to move this kind of data across tools, I treat it as disposable by default.

That means no scripts, no automation, no permanence.

In one instance, encoding the content was enough to make it transferable.
I used a small encoder (mmtocm.net), shared the result, and discarded the process entirely.

The goal wasn’t elegance.
It was containment.

Keeping temporary work out of the system keeps the system easier to reason about.