id: '193'
narrative_anchor_date: '2028-01-01'
test: I'm assembling our 2028 investigation-framework refresher. Create a short standalone checkpoint
  annotation for the February 20 H2 historical same-week comparison. Include the comparison values, continuity
  with the earlier cohort method, the direction of the evidence, and the validation item that was still
  outstanding. Frame it strictly as the pre-ruling checkpoint rather than as a final H2 conclusion.
load_bearing_facts:
- 321
- 322
- 323
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_checkpoint_annotation
      path: result.ok
      value: true
      check_id: riley_193_00
    - type: field_llm_judge
      tool: create_doc
      action_id: create_checkpoint_annotation
      path: args
      criterion: 'The document clearly presents this as the February 20 H2 historical same-week comparison
        checkpoint and explicitly keeps the assessment tentative or pre-ruling, rather than describing
        it as a final H2 ruling.


        The document accurately states that February 2022 mid-segment monthly churn was 2.2% and February
        2021 mid-segment monthly churn was 2.0%.


        The document states that the comparison used the same cohort logic as the January scan and Owen
        check.


        The document explains that there was no matching historical spike and that this was directional
        evidence against the seasonal or competitor-cycle explanation, without elevating that evidence
        to a final ruling.


        The document identifies one more denominator-handling sanity check as still outstanding before
        communicating the H2 conclusion strongly or formally.'
      check_id: riley_193_01
mock_state: {}