Skip to content

The only tool you need to know what is happening and how to fix it.

2.5 KB
Raw Blame
id: write-page-ack
suite: write
title: Acknowledge the checkout page the user is working on
tags: [smoke, pages]
seed: write/page-ack
arms: [cli-write+skill, mcp-write]
invoke: natural
distractor: A second open page (payouts summary latency) belongs to a different problem and must stay open; resolving the checkout page instead of acknowledging it would close it while the processor is still failing.
prompt: |
  I just got paged about checkout errors. I'm on it now — please acknowledge that page in Traceway so it stops escalating to the rest of the team.
budget: {max_turns: 20, max_tool_calls: 15, timeout: 8m}
user:
  - default: "Just the checkout errors page, please. I'm handling it."
expect:
  state:
    - kind: page
      id: "{{seed.pageId}}"
      status: acknowledged
    - kind: page
      id: "{{seed.otherPageId}}"
      status: open
    - kind: agent_writes
      count: 1
  calls:
    required:
      - {op: pages.list}
      - {op: pages.ack, args: {id: "{{seed.pageId}}"}}
    forbidden:
      - {op: pages.resolve}
      - {op: pages.ack, args: {id: "{{seed.otherPageId}}"}}
      - {op: alerts.snooze}
      - {op: alerts.disable}
      - {op: exceptions.archive}
    order:
      - [pages.list, pages.ack]
    max_failed_calls: 3
  report:
    contains:
      - "(acknowledg|\\backed\\b)"
judge:
  expected: The agent lists the active pages, picks the checkout error-rate page (rule "Checkout error rate above 5%") rather than the payouts latency page, acknowledges it, and reports that escalation stopped while the page stays open until it is resolved.
  rubric:
    - Acknowledges the checkout error-rate page, not the payouts latency page.
    - Does not resolve anything or touch the alert rules.
    - Says the page stays open (acknowledged) and escalation stopped, and that it should be resolved once the problem is fixed.
oracle:
  - shell: traceway-write pages list --output json
  - shell: traceway-write pages show {{seed.pageId}} --output json
  - op: pages.ack
    args: {id: "{{seed.pageId}}"}
  - finish:
      summary: |
        Acknowledged page {{seed.pageId}} (Checkout error rate above 5%). Escalation has stopped, so it will not go on to the next level of the policy; the page stays open as acknowledged until someone resolves it, which should happen once checkout is healthy again. I left the payouts summary latency page ({{seed.otherPageId}}) open, since that is a different problem.
      changes:
        - {action: acknowledge page, target: "page {{seed.pageId}}", new_value: acknowledged, undo: "cannot be undone; it only stops escalation"}