Skip to main content
This guide walks you through creating a Shipfox workflow that runs your test suite and an AI code review on every push to GitHub. You’ll set up a two-job pipeline: one job for automated checks, one for an agent review that runs after tests pass.

​
What you’ll build

You’ll create a workflow with two jobs: checks and review. The checks job runs npm ci && npm test to validate the codebase on every push. The review job declares needs: checks, meaning it only starts once checks succeeds, and uses an agent step that reads the repository and flags likely bugs and missing tests — mechanical validation and an agent review, in one workflow that lives with your code.

​
The workflow file

Create the following file at .shipfox/workflows/checks-on-push.yml in your repository: 
name: Checks on push
runner: ubuntu-latest
triggers:
  on_push:
    source: github_acme
    event: push
jobs:
  checks:
    steps:
      - run: npm ci
      - run: npm test
  review:
    needs: checks
    steps:
      - model: claude-opus-4-8
        thinking: high
        prompt: >
          Review this repository for likely bugs and missing test coverage.
          Be concise — summarize the top 3 issues only.
The triggers block tells Shipfox to fire this workflow on every GitHub push. The review job’s needs: checks declaration creates a dependency edge — if checks fails, review is skipped automatically.

​
Deploy it

Follow these steps to get the workflow running:
  1. Create the directory and file in your repository: .shipfox/workflows/checks-on-push.yml.
  2. Paste the YAML above into the file.
  3. Commit the file and push to GitHub.
The workflow appears in the Shipfox dashboard immediately after the push. The next push to any branch fires it.
Push triggers require a GitHub repository connected to your Shipfox project. GitLab is on the roadmap — use source: manual, event: fire if your repo is on GitLab.

​
Parallel checks

When lint and tests are independent, running them in parallel cuts wall-clock time. Split checks into independent lint and test jobs that both start immediately, then fan in to the review job once both finish: 
jobs:
  lint:
    steps:
      - run: npm ci
      - run: npm run lint
  test:
    steps:
      - run: npm ci
      - run: npm test
  review:
    needs: [lint, test]
    steps:
      - model: claude-opus-4-8
        prompt: Summarize any issues found in the test or lint output.
Because each job runs in its own isolated environment, both lint and test install dependencies independently. The review job waits for both to finish before the agent step executes.

​
Next steps