id: 096
narrative_anchor_date: '2028-01-01'
test: I’m gathering concrete input for Owen’s 2028 performance discussion. Create one short evidence note
  based on the earlier first-pass packet reviews, identifying the relevant packet examples, which source-completeness
  fields were present, and how much hands-on structural correction I had to provide. Keep the conclusion
  strictly limited to demonstrated source-structure performance rather than recommending any expansion
  of authority. Then email Priya Devarajan a concise summary with the same bounded conclusion.
load_bearing_facts:
- 223
expected_tool_calls:
- create_doc
- send_email
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_evidence_note
      path: result.ok
      value: true
      check_id: riley_096_00
    - type: field_equals
      tool: send_email
      action_id: email_priya_summary
      path: result.ok
      value: true
      check_id: riley_096_01
    - check_id: riley_096_02
      type: field_equals
      tool: send_email
      action_id: email_priya_summary
      path: args.to
      value: priya@helio.example
    - type: field_llm_judge
      tool: create_doc
      action_id: create_evidence_note
      path: args
      criterion: 'The document must identify all three reviewed examples: Owen’s July monthly sanitized
        campaign snapshot, the Q2 source lock, and the complete August Helio Start paid-pilot packet.
        It must state that their first passes contained the required denominator or coverage dates, outcome
        window, measurement limits, exclusions, source owner, reliability evidence, and revenue and forecast
        treatment, and that Riley did not have to correct the source structure line by line. Missing or
        contradictory information fails.


        The document’s conclusion must remain limited to Owen’s demonstrated first-pass source-structure
        or source-completeness performance and must not recommend expanding his authority or decision
        rights. Contradictory or overreaching conclusions fail.'
      check_id: riley_096_03
    - type: field_llm_judge
      tool: send_email
      action_id: email_priya_summary
      path: args
      criterion: 'The email must concisely summarize all three reviewed examples—Owen’s July monthly sanitized
        campaign snapshot, the Q2 source lock, and the complete August Helio Start paid-pilot packet—and
        explain that on first pass they included the required denominator or coverage dates, outcome window,
        measurement limits, exclusions, source owner, reliability evidence, and revenue and forecast treatment
        without Riley having to correct the source structure line by line. Missing or contradictory information
        fails.


        The email must keep its conclusion limited to demonstrated source-structure or source-completeness
        performance and must not recommend expanding Owen’s authority or decision rights. Contradictory
        or overreaching language fails.'
      check_id: riley_096_04
mock_state: {}