Skip to main content
When Sentry reports a new issue, you can fire a Shipfox workflow that hands the issue to an AI agent, which reads your repository and reports the likely cause — turning an alert into a first-pass investigation automatically. This guide sets up a Sentry-triggered workflow with a single agent step that receives the issue details through template interpolation.

​
What you’ll build

A workflow triggered by the Sentry issue.created event. A triage job runs one agent step whose prompt embeds the specific issue — its title, culprit, and level — using ${{ event.* }} interpolation. The agent has your repository checked out, so it can investigate the code behind the error and summarize what a developer should check.

​
The workflow file

Create the following file at .shipfox/workflows/triage-sentry.yml in your repository: 
name: Triage Sentry issues
runner: ubuntu-latest
triggers:
  on_issue:
    source: sentry_acme
    event: issue.created
jobs:
  triage:
    steps:
      - model: claude-opus-4-8
        thinking: high
        prompt: |
          A new Sentry issue was reported for this project:
          "${{ event.title }}" — culprit ${{ event.culprit }} (level ${{ event.level }}).
          ${{ event.webUrl }}

          Investigate the repository for the likely cause and summarize what a
          developer should check first. Be concise.
Each matching Sentry issue starts a run. Shipfox resolves the ${{ event.* }} expressions when the run is created, so the agent receives the actual issue — its title, culprit, level, and a link back to Sentry (webUrl). The agent’s summary streams to the step log.

​
Deploy it

Follow these steps to get the workflow running:
  1. Create the file .shipfox/workflows/triage-sentry.yml in your repository.
  2. Paste the YAML above, replacing sentry_acme with your Sentry connection slug.
  3. Commit the file and push. The workflow appears in the dashboard and fires on the next matching issue.
source is your Sentry connection slug — shown in your integration settings after you connect Sentry (for example sentry_<org>), not the literal word sentry. Connecting Sentry requires the Sentry integration.

​
Scope to a project

Shipfox delivers issue.created for every project in your connected Sentry organization. The event exposes the issue’s project as event.projectUrl, so you can point the agent at only what you care about — or name the project in the prompt: 
      - model: claude-opus-4-8
        prompt: |
          Triage this Sentry issue in ${{ event.projectUrl }}:
          "${{ event.title }}". Summarize the likely cause.
Server-side filtering by project or team — filter: event.projectUrl.contains("...") — is on the roadmap; the filter field is parsed but not evaluated yet. Until then, scope inside the workflow. See Expressions.

​
Going further

Turn triage into action by combining it with the other guides:
  • Auto-fix and verify — follow the agent with a shell step and a gate so it retries until tests pass. See Gate & Retry.
  • Open a PR or notify — add a run step that calls gh pr create or posts to your chat tool.
Branching on a structured agent decision (auto-fix vs. escalate) and structured agent output are on the roadmap. Today, act with run steps and gates.

​
Next steps

  • Sentry integration — connect Sentry and see every issue event.
  • Agent Steps — configure models, thinking levels, and providers.
  • Expressions — the ${{ }} context available to prompts.