braintrust-design-eval-metric-bundle
Installation
SKILL.md
Design the metric bundle
Contract: references/interaction-contract.md. Calibration, templates, provenance: references/metric-bundle.md.
Trigger
- Requests to choose metrics, define "good," or set targets.
- A single headline KPI or composite score driving a ship decision.
- A targeted change being judged by a whole-response average.