metric/evaluate-cslc Kernel

Context

This skill implements FPF Pattern A.18 (Metrics & Scales). To avoid "averaging nonsense", every metric must have a strictly defined Scale Type: Nominal, Ordinal, Interval, or Ratio.

Instructions

Given a value and a scale_definition, validate the value and return its normalized Coordinate.

1. Scale Types

Handle the following types strictly:

•Nominal: Categorical only (e.g., {pass, fail}). No ordering.
•Ordinal: Ranked (e.g., 1-5). Order matters, but distance is undefined. You CANNOT average these (use median).
•Interval: Distance matters, no true zero (e.g., Temperature C). You can average.
•Ratio: True zero exists (e.g., Time in seconds). You can average and divide.

2. Output

If the value is valid for the scale, return:

json

{
  "valid": true,
  "coordinate": number | string,
  "scale_type": string
}

If invalid (e.g., "7" on a 1-5 scale), return valid=false.