# Target Confirmation Rerun Instruction

<!-- auto-oncall-optimization-record
{
  "schemaVersion": "auto-oncall-optimization-record-v1",
  "optimizationId": "target-confirmation-rerun-instruction-2026-08-13",
  "problems": [
    {
      "problemId": "target-only-internal-note-overexplains-future-diagnostics",
      "summary": "APID-29340 Internal Only correctly translated the target-resolution failure but still listed a full future diagnostic and escalation checklist, obscuring the only immediate Support action: confirm the actual cluster and invoke Auto Oncall again with that cluster name.",
      "status": "optimized_verified",
      "applicability": "Human-required diagnostic responses where no Provider evidence was acquired and unresolved mapping from a ticket cluster alias to the actual cluster is the sole blocker.",
      "issueKeys": ["APID-29340"],
      "sourceRefs": [
        "https://tidb.atlassian.net/browse/APID-29340#problem:target-confirmation-rerun"
      ],
      "signalTextIncludesAny": [
        "confirm the actual cluster then rerun",
        "target-only Internal Only is too verbose",
        "明确集群后 @ bot 重新生成"
      ],
      "observedThrough": "2026-08-13",
      "effectiveAt": "2026-08-13T08:50:08Z",
      "verifiedAt": "2026-08-13T09:02:11Z",
      "attemptedAction": "Make a target-only Internal Only message one compact action paragraph that tells Support to confirm the actual cluster and then create an internal Jira comment with a real ADF @Auto Oncall mention plus the cluster name; omit the future query and escalation inventory.",
      "evidenceUrl": "https://tidb.atlassian.net/browse/APID-29340?focusedCommentId=23548869"
    }
  ]
}
-->

- Optimization ID: `target-confirmation-rerun-instruction-2026-08-13`
- Owner: Scenario Router and Final Response audience projection
- Source reference: `APID-29340`, internal comment `23548869`
- Implementation status: merged in `dce55d8e` by PR #876 and deployed to
  staging in the combined `b57324bb` image
- Effective status: `optimized_verified`

## Bounded problem

The support-readable Internal Only note still listed every future metric,
query, and escalation branch even though no diagnostic evidence had been
acquired. The only immediately executable Support action was to confirm which
actual cluster “亚太01” referred to and then invoke Auto Oncall again with that
cluster name.

## Candidate change

- Use one short action paragraph when target identity is the sole blocker.
- Tell Support to confirm the actual cluster, then create an internal Jira
  comment with a real ADF `@Auto Oncall` mention and the cluster name.
- Do not emit a future diagnostic checklist or escalation tree that the rerun
  will select after it has a valid target.
- Preserve the blocked technical conclusion in the customer message rather
  than repeating it internally.

## Verification evidence

- Focused prompt and formatting tests passed; the complete local suite passed
  791/791 tests. The unrelated Wetech concurrency timing check passed on CI
  rerun before PR #876 merged.
- Staging read replay `faece2b7-b910-4bd3-91ba-a14116066d8c` produced one
  compact Internal Only paragraph that asks Support to confirm the actual
  cluster, select and mention `@Auto Oncall` in an internal Jira comment, add
  the cluster name, and regenerate the analysis.
- Authorized write replay `42a2c129-4b91-4ae0-ad89-01f7e29cb052` updated the
  existing Jira internal comment `23548869` in place. Independent Jira
  readback found exactly that one internal comment, with `public=false` and the
  expected compact text.

## Next validation action

Preserve the compact target-only instruction for future cases. If Support
supplies the actual cluster in a real ADF `@Auto Oncall` continuation, verify
the next analysis resolves the target and performs the evidence-backed
diagnosis.