role-selection-scorecard:
  version: 1
  owner: OPS-ORCH
  purpose: Decide which role agents should be called now, later, or skipped.
  scoring-fields:
    relevance:
      range: 0-3
      meaning: How directly the role owns the requested problem.
    risk-coverage:
      range: 0-3
      meaning: How much unique risk coverage this role adds.
    evidence-need:
      range: 0-3
      meaning: Whether the role can collect or interpret required evidence.
    decision-authority:
      range: 0-3
      meaning: Whether the role has DRAI decision or recommendation authority.
    implementation-impact:
      range: 0-3
      meaning: Whether execution will affect this role's domain.
    duplicate-penalty:
      range: 0-3
      meaning: How much the existing reports already cover this role's likely contribution.
  rubric-anchors:
    note: 각 필드 0-3 앵커. 아래는 relevance/risk-coverage 예시이며 나머지 필드도 같은 척도(0=없음, 1=약함, 2=상당, 3=핵심)를 따른다.
    relevance:
      s0: 이 family 도메인과 무관.
      s1: 간접 연관(참고만).
      s2: 문제의 상당 부분을 이 family가 소유.
      s3: 이 family가 문제의 1차 소유자.
    risk-coverage:
      s0: 추가 리스크 커버 없음.
      s1: 경미한 리스크 일부.
      s2: 고유 리스크를 상당 부분 커버.
      s3: 이 family 없이는 커버 불가한 중대 리스크(보안/신뢰성/법무/매출).
  topology-modifiers:
    source: org-os/00-role-registry/team-topology-map.yaml
    rules:
      - Stream-aligned customer-domain work increases relevance for PM, PO, product design, product engineering, and data analyst roles.
      - Platform or golden-path work increases relevance for Platform PO, platform design, platform engineering, DevOps, SRE, and DevSecOps roles.
      - Complicated subsystem work increases relevance for Technical PO and architecture specialist roles.
      - Enabling work increases relevance for business architecture, strategy, QA, security, UX research, and productivity roles.
      - Repeated cross-team duplication increases platform-role relevance and duplicate-penalty for isolated stream-only execution.
      - Product-speed-vs-technical-stability conflict increases C-Level and CPTO relevance.
      - Revenue, GTM motion, pricing, demand generation, lead-to-cash, PLS handoff, churn, expansion, pipeline, or partnership work increases relevance for gtm-revenue roles.
      - Product-led sales handoff increases relevance for Growth PM, PMM, RevOps, Sales, Customer Success, and Legal/Compliance when enterprise claims or contracts are involved.
  total-score-formula: relevance + risk-coverage + evidence-need + decision-authority + implementation-impact - duplicate-penalty
  decision-thresholds:
    skip:
      score-range: 0-4
    call-later:
      score-range: 5-8
    call-now:
      score-range: 9-12
    call-now-priority:
      score-range: 13-15
  hard-rules:
    - wave-size must not exceed 5 role agents.
    - High or Critical security, reliability, financial, legal, privacy, or user-harm risk must include the relevant auditor or decider role.
    - duplicate-penalty lowers priority; it must never increase total score.
    - skipped roles require skipped-role-reason in workflow-state-registry.
    - A workflow must declare primary-team-topology before the first execution wave.
    - If primary-team-topology is platform, release acceptance must include golden-path or paved-road impact.
    - If primary-team-topology is stream-aligned, output must include customer value, product metric, or learning objective.
    - If primary-team-topology is complicated-subsystem, output must include interface simplification for consuming teams.
    - If primary-team-topology is enabling, output must include the target team friction or capability gap being removed.
    - If primary-team-topology is gtm-revenue, output must include revenue metric, GTM motion, handoff owner, and lead-to-cash impact.
    - A workflow must declare mode and tier before the first execution wave.
    - Divergent mode must cover at least the tier's min-distinct-lenses in parallel (governance-tiers).
    - Converge heavy must expose lens tradeoffs and must not merge distinct lenses into one recommendation.
    - Scoring and wave-size are counted per capability-family, not per individual role (execution-policy family-collapse).
    - Tie-break when call-now candidates exceed wave-size, order by, total-score, then risk-coverage, then relevance; remaining ties prefer families carrying an auditor or decider for the document type.
  output-template:
    workflow-id:
    candidate-family:
    candidate-role:
    primary-team-topology:
    relevance:
    risk-coverage:
    evidence-need:
    decision-authority:
    implementation-impact:
    duplicate-penalty:
    total-score:
    decision: call-now / call-later / skip
    reason:
    mode: divergent / converge
    tier: light / standard / heavy
    assigned-lens:
    lens-coverage: