id: '151'
narrative_anchor_date: '2028-01-01'
test: Create a historical evidence scorecard for the CS auto-nudge replay. Reconcile matches, field-specific
  holds, and suppressions. Show each category total, each cause as a subtotal of its parent category,
  and verify that all subtotals and the grand total add correctly. Explain how the non-matches relate
  to the shared-state contract and whether any record with current, supported state changed outcome.
load_bearing_facts:
- 101
expected_tool_calls:
- create_doc
grade:
  type: tool_trace
  config:
    check_version: 2
    today: '2028-01-01'
    semantic_judge_version: 2
    assertions:
    - type: field_equals
      tool: create_doc
      action_id: create_replay_scorecard
      path: result.ok
      value: true
      check_id: riley_151_00
    - type: field_llm_judge
      tool: create_doc
      action_id: create_replay_scorecard
      path: args
      criterion: The scorecard reconciles 412 decisions into 386 matches and 26 cases prevented by the
        shared-state contract. The 26 comprise 19 field-specific holds, including 15 for stale qualifying-human-admin
        state and four for unresolved external-system authority, plus seven suppressions for active Customer
        Success adoption or renewal work. It says no record with all required state current and supported
        changed outcome.
      check_id: riley_151_01
mock_state: {}