Shipfox

Getting Started with Shipfox: Run Your First Workflow

Add a workflow and verify a command and agent step in your first Shipfox run.

In this lesson, you will add a complete workflow to a test repository and run it. The final run prints a command result and contains an agent summary based on the repository's files.

Before you begin

This lesson starts with a Shipfox project connected to a repository where you can push a test workflow. If your workspace needs more setup, the in-product Get started guide lists the next step. Open it from the workspace home or the Get started button in the top bar.

No project yet? Open your workspace and follow its setup flow. No workspace yet? See Installation.

Add your first workflow

A workflow is a YAML file that tells Shipfox when to run (triggers), where to run (runner), and what to do (jobs and steps). Create .shipfox/workflows/hello.yaml. This is a complete workflow:

# yaml-language-server: $schema=https://www.shipfox.io/docs/workflow.schema.json
name: Inspect the repository
runner: shipfox      # label of the runner that executes the jobs
triggers:
  on_demand:
    source: manual         # fires when you click Run in the dashboard
jobs:
  inspect:
    steps:
      - run: echo "hello world"
      - prompt: |
          Read this repository and summarize its purpose in two sentences.
          Name one top-level file that supports your answer.
          Do not change any files.

The manual trigger works in every workspace with no further setup. Every field is described in the Workflow Schema reference.

Commit and push the file to the repository's default branch. Shipfox detects the workflow automatically. No manual import is required.

Run it

A run records one use of a workflow. Open the project's Workflows page, find Inspect the repository, and select Run.

Open the run to stream its output. The command step should show:

inspect · echo "hello world"
hello world

Shipfox run detail page showing a succeeded run with its job graph and step list

A run detail page: status, the job graph, and every step with its duration.

A run stuck in pending means no online runner matches the runner: value in your workflow. See Resolve pending jobs.

Verify the result

Open the agent step after it finishes.

  1. Confirm that the response has two sentences.
  2. Open the file named by the agent and check that it supports the summary.
  3. Confirm that the agent did not change any repository file.

The agent uses the checkout for the commit that started the run. Its messages, tool calls, token use, cost, and final response remain in the step log.

What you built

You added a complete workflow and started it with a manual trigger. One job ran a fixed command and an agent in the same checkout. The run now provides a visible record of both steps and their results.

Next steps

Was this page helpful?
Edit this page on GitHub

On this page