id: '104'
narrative_anchor_date: '2028-01-01'
test: Create a concise metric card for the certified CS auto-nudge coverage in the closed four-lane source
  packet. Give the complete coverage dates including the year, total decisions, both disposition counts
  and rates, arithmetic reconciliation, source-age alert outcome, control failures, and the narrow system-control
  interpretation.
load_bearing_facts:
- 228
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_metric_card
      path: result.ok
      value: true
      check_id: riley_104_00
    - type: field_llm_judge
      tool: create_doc
      action_id: create_metric_card
      path: args
      criterion: 'The document states that the coverage period was July 1 through October 31, 2026 and
        reports exactly 6,704 unique decisions, comprising exactly 249 visible field-specific holds and
        exactly 6,455 decisions that proceeded or were suppressed using current state. It must make clear
        that the two disposition counts reconcile to the total (249 + 6,455 = 6,704). Missing or contradictory
        figures fail.


        The document gives both correctly calculated disposition rates using 6,704 as the denominator:
        the visible-hold rate is approximately 3.71% and the proceeded-or-suppressed-using-current-state
        rate is approximately 96.29%. Reasonable rounding such as 3.7% and 96.3% is acceptable, and the
        rates should total approximately 100%.


        The document records that four source-age alerts fired and all resolved before breach, and that
        there was no governed-field disagreement, no customer correction, and no rollback. Missing or
        contradicted outcomes fail.


        The interpretation is explicitly limited to system-control evidence or control performance. It
        must not characterize the figures as evidence of customer success, engagement, conversion, revenue,
        pipeline, forecast, or other commercial performance, and must not make causal claims.'
      check_id: riley_104_01
mock_state: {}