Definition
Behavioral stability of a codified skill across model versions, measured on a frozen input set. Only adjudicated regressions count against stability; accepted improvements reset the baseline.
The Measurement Gap
There is no metric for output drift of a codified skill across model versions (same skill, same input class, changed behavior after upgrade). The operator's procedure library silently re-interprets itself with every platform release.
Formula
SOS_k = 1 - ( regression_diff_points_k / total_diff_points_k ), per skill k, at each release
Proposed formula, pending validation pilots.
At a Glance
Decision Supported
When a codified asset needs maintenance, and how much trust to place in it between maintenance passes.