Benchmark models

Run an agent's linked tests against multiple models and rank the models on a leaderboard. Drives the calibrate agent-tests benchmark / get-benchmark commands. Ask at each step; if $ARGUMENTS already carries an agent UUID or name, pre-fill it and skip that question.

See ../../references/agent-mode.md for how to read command output (TOON/JSON) and for the launch-and-poll pattern used in Phase 5. Keep what you say to the user plain — see ../../references/voice.md.

Get the latest instructions

These instructions are updated often, and the copy installed on this machine can be months old. Before doing anything else, pull the latest copy of this skill: