design-evaluator
Installation
SKILL.md
Design an evaluator
Create a new evaluator — a versioned LLM/audio judge — and its first live
version. An evaluator scores agent output against a rubric you write. create
makes v1 and sets it live. Ask at each step; if $ARGUMENTS already carries a
name, pre-fill it and skip that question.
See ../../references/agent-mode.md for how to
read command output (TOON/JSON) and
../../references/config-shapes.md for the
version-param shape. Keep what you say to the user plain — see
../../references/voice.md.
An evaluator is a versioned judge: the object holds a history of versions,
one of which is live. Here you author v1. Later tuning is a separate flow
(/iterate-evaluator).