Test 103

Jan 1, 2028 / 1 fact

Request

Please post a decision comment on proposal document `doc_activation_recovery_banner_2028`. The historical replay is complete and found no semantic disagreement. Record the furthest stage the proposal may currently reach, the production approvals that are still required, and the intake fields that must remain explicit under our standing shared-state consumer admission gate.

Required memory

Fact 200

Before replay work begins for any future shared-state consumer, the requester must identify the customer action, every governed field used with its unchanged meaning, each source owner and freshness limit, fail-closed and suppression behavior, customer-correction and incident-stop ownership, and engineering rollback ownership. A candidate may move from replay to shadow only if no semantic disagreement appears; production requires Riley Tanaka’s approval of the customer-action scope and Daniela’s approval of release and rollback controls.

Source evidence (1)

003604Jun 9, 2026 / 16:15 UTC-05:00

Daniela and I completed the renewal-prioritization review and declined the proposed direct connection, so Lifecycle Systems remains at four production consumers. For any future shared-state consumer, the requester must name the customer action; every governed field used with its unchanged meaning; each source owner and freshness limit; fail-closed and suppression behavior; customer-correction and incident-stop ownership; and engineering rollback ownership before replay work begins. A candidate can move from replay to shadow only if no semantic disagreement appears. Production requires my approval of the customer-action scope and Daniela’s approval of release and rollback controls. For the immediate renewal question, Owen gave Customer Success a one-time aggregate sizing cut rather than an account-level feed. Update `Shared-state consumer admission review — renewal prioritization` with the final decline, the replay-to-shadow-to-production admission gate, the bounded aggregate alternative, and the final four-production-consumer boundary.

Message 003604 in history

Expected tool calls

post_doc_comment

Grading

1. field_equals / post_doc_comment

{
  "type": "field_equals",
  "tool": "post_doc_comment",
  "action_id": "post_admission_decision",
  "path": "result.ok",
  "value": true,
  "check_id": "riley_103_00"
}

2. field_equals / post_doc_comment

{
  "check_id": "riley_103_01",
  "type": "field_equals",
  "tool": "post_doc_comment",
  "action_id": "post_admission_decision",
  "path": "args.doc_id",
  "value": "doc_activation_recovery_banner_2028"
}

3. field_llm_judge / post_doc_comment

{
  "type": "field_llm_judge",
  "tool": "post_doc_comment",
  "action_id": "post_admission_decision",
  "path": "args",
  "criterion": "The comment clearly states that, because replay found no semantic disagreement, the proposal may advance to shadow but may not yet advance to production.\n\nThe comment clearly identifies both remaining production approvals: Riley must approve the customer-action scope, and Daniela must approve the release and rollback controls. Missing either approval or implying production is already approved must fail.\n\nThe comment requires all of these intake items to remain explicit: the customer action; every governed field used with its established meaning unchanged; each source owner and freshness limit; fail-closed behavior and suppression behavior; ownership for customer corrections and incident stops; and engineering rollback ownership. Missing any category must fail.",
  "check_id": "riley_103_02"
}

Complete grading specification

{
  "type": "tool_trace",
  "config": {
    "check_version": 2,
    "today": "2028-01-01",
    "semantic_judge_version": 2,
    "assertions": [
      {
        "type": "field_equals",
        "tool": "post_doc_comment",
        "action_id": "post_admission_decision",
        "path": "result.ok",
        "value": true,
        "check_id": "riley_103_00"
      },
      {
        "check_id": "riley_103_01",
        "type": "field_equals",
        "tool": "post_doc_comment",
        "action_id": "post_admission_decision",
        "path": "args.doc_id",
        "value": "doc_activation_recovery_banner_2028"
      },
      {
        "type": "field_llm_judge",
        "tool": "post_doc_comment",
        "action_id": "post_admission_decision",
        "path": "args",
        "criterion": "The comment clearly states that, because replay found no semantic disagreement, the proposal may advance to shadow but may not yet advance to production.\n\nThe comment clearly identifies both remaining production approvals: Riley must approve the customer-action scope, and Daniela must approve the release and rollback controls. Missing either approval or implying production is already approved must fail.\n\nThe comment requires all of these intake items to remain explicit: the customer action; every governed field used with its established meaning unchanged; each source owner and freshness limit; fail-closed behavior and suppression behavior; ownership for customer corrections and incident stops; and engineering rollback ownership. Missing any category must fail.",
        "check_id": "riley_103_02"
      }
    ]
  }
}

App state

Download JSON

Source file

tests/riley/103.yaml

SHA-256: d5d2fffb45bcc543ea83cc867e243652f466d9435ee1547a5f3819ce656cb1d5