braintrust-size-eval-dataset
Installation
SKILL.md
Size the dataset
Contract: references/interaction-contract.md. Calibration, templates, provenance: references/power-analysis.md.
Trigger
- "How many eval cases do we need?" / "Can 200 examples detect a 3-point gain?"
- "How many clean trials certify a low failure rate?"
- A null result that needs distinguishing from an underpowered one.