Score Stability Only from Current Evidence

Operational scores should be earned by evidence from the current environment. A subsystem may have passed last week while a configuration change silently invalidated the supported path today.

Break the score into dimensions such as core loop reliability, retrieval integrity, contract stability, and operator confidence. For each dimension, cite the command, timestamp, and limitation that bound the result.

The score is not the product claim. It is a compact map of what is working, what is missing, and which proof step would increase confidence next.