Report failed Shipfox workflow runs

How it works

1. A workflow run failstriggerThe workflow watches every workflow in the project.
2. The workflow reads the failed runtoolIt gets the failed jobs and steps, the first error, and the last 15 log lines.
3. You get a report on SlackwriteThe report suggests a next step. Optional: an agent adds a diagnosis in the thread.

What it writes

• Slack
  • Posts a message for each failed run

Before you start

• Connect Slack and invite the Shipfox app to the report channel.

Choices you make

When you set up this workflow, your coding agent asks you these questions. The workflow file on this page uses the default answers.

• Which projects should the report cover?Default: This project1 other choice

  • This projectDefaultReports runs of the project that holds this workflow. Requires the project ID.
  • Every project in the workspaceReports runs of every project into one channel. Add this workflow to one project only, or each failure is reported once per copy.
• Which workflows should the report cover?Default: Every workflow1 other choice

  • Every workflowDefaultReports failed runs of every synced workflow except this one.
  • Named workflow filesReports only the workflow files you list. A moved or renamed file stops being reported until you update the list.
• Should the report include cancelled runs?Default: Failed runs only1 other choice

  • Failed runs onlyDefaultCancelled runs are not reported.
  • Failed and cancelled runsAlso reports runs a person cancelled or that timed out. Runs replaced by a newer run in their concurrency group are never reported.
• Should an agent add a diagnosis from the failed step logs?Default: Report only1 other choice

  • Report onlyDefaultNo model or inference cost. The report lists the failed jobs and steps, a log excerpt, and a next step.
  • Add an agent diagnosisAfter each failed-run report, an agent reads the run's failed logs and replies in the report's thread. Each failed run uses inference.

Models

• diagnosetested with gpt-6-luna · low thinkingReads one failed run and its step logs, then writes a short diagnosis that cites the log lines.

When you set up this workflow, your coding agent suggests models that your workspace can use. You choose the model for each step.

Related examples

• Investigate and repair default-branch CI failures