Skip to Content
ReferenceActions Reference

Actions Reference

Every step names an action in uses. Each action has its own page below, with its parameters, the res object it returns, and worked examples. The rules that hold for all of them are at the end of this page.

Built-in actions

  • http - Make HTTP/HTTPS requests and validate responses
  • db - Execute database queries on MySQL, PostgreSQL, and SQLite
  • shell - Execute shell commands and scripts securely
  • ssh - Run commands on a remote host over SSH
  • smtp - Send email notifications and alerts
  • imap - Connect to IMAP servers and manage email operations
  • grpc - Call gRPC services by reflection
  • dns - Ask a DNS server for the records of a name
  • embedded - Run another workflow as a step
  • hello - Simple test action for development and debugging

External actions

These are published alongside Probe, each in a repository of its own. A step names one by repository and commit, and Probe downloads it the first time a workflow uses it. See External Actions.

  • browser - Drive a real Chrome with chromedp
  • container - Start a container for the steps of a run, and remove it once the run is over
  • graphql - Send a GraphQL query over HTTP
  • jmap - Call JMAP methods
  • mail-latency - Measure delivery latency from a Maildir
  • redis - Run commands on a Redis or Valkey server
  • s3 - Read and write objects in S3 and S3-compatible storage
  • websocket - Send and receive messages over a WebSocket connection

Action Error Handling

A step fails when its test is false or when the action itself returns an error. The remaining steps of the job still run, the job is marked failed, and jobs that list it in needs are skipped.

There is no switch to ignore a failure. When a check should not fail the workflow, record its result as an output instead of asserting it.

steps:
  - name: Required check
    uses: http
    with:
      method: GET
      url: "{{vars.api_url}}/health"
    test: res.code == 200
 
  - name: Optional check
    id: optional
    uses: http
    with:
      method: GET
      url: "{{vars.api_url}}/experimental"
    outputs:
      available: res.code == 200
      detail: 'res.code >= 400 ? res.status : ""'
 
  - name: Report
    uses: hello
    echo: "Experimental endpoint: {{outputs.optional.available ? \"available\" : outputs.optional.detail}}"

Use retry for a transient failure and timeout for a step that may hang:

  - name: Flaky endpoint
    uses: http
    timeout: 10s
    retry:
      max_attempts: 3
      interval: 2s
    with:
      method: GET
      url: "{{vars.api_url}}/flaky"
    test: res.code == 200

Performance Considerations

  • Jobs without a needs relation run in parallel, so independent checks do not queue behind each other.
  • rt.sec and rt.duration measure the action’s round trip, not the whole step. Every step runs its action in a process of its own, so the round trip of an http step includes connecting.
  • A large response body is held in memory; a binary body is written to a file and reported as res.filepath.
  • repeat with async: true starts a run of a job every interval, without waiting for the one before, which sends requests at a steady rate. It is for taking samples, not for load tests; see Performance Testing.
jobs:
  - name: Ping at 10 requests a second
    repeat:
      count: 50
      interval: 100ms
      async: true
    steps:
      - name: Ping
        uses: http
        timeout: 2s
        with:
          method: GET
          url: "{{vars.api_url}}/ping"
        test: res.code == 200 && rt.sec < 0.5

See Also

Updated at
CLI ReferenceHTTP