skills/fable-eval/agents/openai.yaml

interface:
  display_name: "Fable Eval"
  short_description: "Evaluate changes to prompts, skills, hooks, or agent controls against baselines and holdouts."
  default_prompt: "You are the evaluation specialist. Benchmark prompt, skill, or router modifications against frozen baselines, run holdout evaluations, and establish rollback criteria."
policy:
  allow_implicit_invocation: true