SWEny, YAML workflows for AI agents I'm running in prod (triage, E2E)

First release of swenyai/sweny/actions/e2e — an agentic end-to-end test runner for GitHub Actions.

Builds on actions/run and adds everything you typically need to run an LLM-driven E2E suite against a deployed app:

• Installs agent-browser (accessibility-tree-based browser automation that agents drive via shell)
• Exposes a BASE_URL input pointing at the environment under test
• Uploads the screenshots / results directory as a workflow artifact (always — even when the workflow fails)

You provide the workflow YAML; the action handles the install, execution, and artifact upload.

Usage

name: E2E UAT
on:
  workflow_run:
    workflows: ["Deploy to Staging"]
    types: [completed]

jobs:
  e2e:
    runs-on: macos-latest
    timeout-minutes: 20
    steps:
      - uses: actions/checkout@v4
      - uses: swenyai/sweny/actions/e2e@v4
        with:
          workflow: .sweny/e2e/uat.yml
          base-url: https://staging.example.com
          claude-oauth-token: ${{ secrets.CLAUDE_CODE_OAUTH_TOKEN }}
        env:
          AUTH0_M2M_CLIENT_ID: ${{ secrets.AUTH0_M2M_CLIENT_ID }}
          AUTH0_M2M_CLIENT_SECRET: ${{ secrets.AUTH0_M2M_CLIENT_SECRET }}

Inputs

| Input | Required | Default |
|---|---|---|
| workflow | yes | — |
| claude-oauth-token | one of | — |
| anthropic-api-key | one of | — |
| base-url | no | — |
| cli-version | no | latest |
| agent-browser-version | no | latest |
| node-version | no | 24 |
| working-directory | no | . |
| screenshots-path | no | results/ |
| artifact-name | no | e2e-screenshots |
| artifact-retention-days | no | 14 |

Versioning

This action ships out of the swenyai/sweny repo. The moving @v4 tag tracks the same release stream as the root swenyai/sweny@v4 action — pin to @v4 for the latest, or to this immutable tag (@actions/e2e/v1.0.0) for stability.

See the README for full docs.