id: '036'
narrative_anchor_date: '2028-01-01'
test: Create a short document titled `April 2023 onboarding and mid-segment analysis — data-quality erratum`
  so future readers do not reuse the stale cohort logic; include the corrected cohort boundary and both
  the classification and timing issues that made the original read noisier.
load_bearing_facts:
- 369
- 370
- 371
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_erratum
      path: result.ok
      value: true
      check_id: riley_036_00
    - type: field_llm_judge
      tool: create_doc
      action_id: create_erratum
      path: args
      criterion: 'The document is clearly an erratum for the April 2023 onboarding and mid-segment analysis
        and states that the corrected/default mid-segment boundary is 75–200 seats rather than the stale
        logic that included 50–75-seat accounts. It records both data-quality issues: the segment job
        misclassified 50–75-seat accounts as mid-segment, muddying or corrupting the denominator, and
        the onboarding send read stale segment membership from a cache window, adding timing noise. Equivalent
        wording is acceptable, but the corrected boundary, both issues, and their relationship to the
        noisy read must all be present.'
      check_id: riley_036_01
mock_state: {}