id: '161'
narrative_anchor_date: '2028-01-01'
test: Create a concise incident case study for the 2028 reliability-training library about the first live
  pre-breach source-age alert. Cover what stalled, the warning and restoration timing, the affected pending
  decisions, the fresh reevaluation outcomes, and whether any decision became stale or required manual
  release, rollback, or customer correction. Use a clear title and brief, practical sections suitable
  for training.
load_bearing_facts:
- 123
- 124
- 125
- 126
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_reliability_case_study
      path: result.ok
      value: true
      check_id: riley_161_00
    - type: field_llm_judge
      tool: create_doc
      action_id: create_reliability_case_study
      path: args
      criterion: 'The case study accurately identifies a stalled qualifying-human-admin source refresh
        affecting 17 pending CS auto-nudge decisions, says the alert fired 43 minutes before the freshness
        limit, and says Daniela restored the refresh with 18 minutes remaining. Missing or contradictory
        timing, count, source, or restoration information must fail.


        The case study states that a fresh reevaluation occurred and accurately reports that 13 decisions
        proceeded while four were suppressed because current Customer Success adoption or renewal work
        was active. Missing, reversed, or contradictory dispositions must fail.


        The case study clearly states that none of the affected decisions entered stale state, nobody
        manually released a decision, and neither a customer correction nor a rollback was required. Any
        claim that one of these events occurred must fail.


        The document title clearly identifies this as an incident case study about the first live pre-breach
        source-age alert. A generic, irrelevant, or misleading title must fail.'
      check_id: riley_161_01
mock_state: {}