braintrust-eval
Installation
SKILL.md
Braintrust Eval
The runbook. Take the user from "I want to eval X" to a clean, comparable Braintrust experiment that actually ran. Deliberately interactive: ask, propose, get approval, then spend money.
This skill owns execution — credentials, imports, code, run hygiene, tracing, and the approval gates that keep a paid run from starting on an unreviewed plan. It does not own methodology: each phase below names the lifecycle card that does, and those are worth loading when a judgment call is live rather than routine.
Shared platform mechanics — pinning, naming, safe reads, run hygiene — are in
references/platform-mechanics.md, not restated per phase.