Skip to content

Timeout and agent-run alerts: agent_request kind fires for any error after the deadline, most kinds carry no requestId, and the runbook delivery line and watcher fixture are stale #945

Open

Description

@alex-mextner

What

Found by the single-seat Opus review of #941 (work of #935), finding listed in #935 (comment). Finding 8. The agent_request timeout is reported for any error after the deadline, not only for aborts; most timeout kinds in timeout-reporter.ts carry no requestId, so an investigation cannot join them to the AI request metric; the delivery-line format in deploy-runbook.md and the fixture in hcb-alert-watcher.test.ts do not match what the code writes.

Why (motivation)

An alert has to say what really timed out and which request it belongs to.

User impact

Investigations chase non-timeouts and cannot correlate timeouts with a user request.

Cost of inaction

The repair flow works from imprecise alerts.

Acceptance criteria

  • Test: a non-abort error after the deadline is not reported as an agent_request timeout
  • Test: the runbook example and the watcher fixture parse from real reporter output

Screenshots

(none)

Deploy

Deployed: no

Links

  • (none)

Skipped gates

  • duplicate: searched incident, alert, watcher, validator, timeout, receipt; no ticket covers this
  • screenshots: no UI change
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions