first-draft-experiment-spec:
  version: 1
  purpose: "동일 모델·동일 요청의 수정 전 첫 결과를 current harness(A)와 experience-foundation treatment(B)로 비교한다."
  control-invariants:
    same-model: true
    same-request-sha256: true
    first-result-only: true
    generation-attempts-per-arm: 1
    revision-count-at-capture: 0
    evaluator-blinded: true
  arms:
    A:
      treatment: current-harness
      forbidden-input-kinds: [competitive-experience-benchmark, experience-blueprint, wireframe-set, design-md, component-registry]
    B:
      treatment: experience-foundation
      required-input-kinds: [competitive-experience-benchmark, experience-blueprint, wireframe-set, design-md, component-registry]
      generation-scope: representative-section-or-core-screen
  metrics:
    score-1-to-5:
      - product-purpose-understood-in-5s
      - primary-action-found
      - information-hierarchy
      - category-quality-parity
      - visual-craft
      - mobile-clarity
      - design-system-consistency
      - human-preference
    lower-is-better:
      - revision-count-to-acceptance
      - tokens-to-acceptance
  decision-rule:
    treatment-supported-when:
      - "B human-preference > A"
      - "B quality-score mean > A"
      - "B revision-count-to-acceptance < A 또는 B tokens-to-acceptance < A"
    no-claim-before-completed: true