e2e-run

Files (1)

• SKILL.md

SKILL.md (first lines)

You're the **/e2e-run** skill. Turn the request into a delegation to the `e2e-runner` agent, which drives a live browser to verify the flow's **acceptance criteria** and reports pass/fail with evidence. You don't write files, and neither does the agent — this is live verification, not test authoring.

The user's request: **$ARGUMENTS**

> Writing a reusable Playwright `.spec` file is a different job — that's playwright's `/playwright`. This skill *exercises* a flow live against its acceptance criteria and reports.

## Step 1: Parse
- The **ticket / flow** to verify, its **acceptance criteria** (what "done" means — from the ticket, or ask the user to state them), the **starting URL/route**, and any **inputs** (credentials, search terms).

## Step 2: Resolve
- Confirm a start URL. If only a route is given, ask for or infer the base URL (e.g. the local dev server).
- The agent reads `.bench/e2e.yaml` for the driver (Chrome vs Playwright MCP), pre/post steps, and capture settings (screenshots / network / GIF). Make sure the configured browser MCP is connected.

## Step 3: Build context blob
```
- Ticket/flow: {name}
- Acceptance criteria: {the criteria to verify, one per line}
- Start URL: {url}
- Inputs: {credentials / data to type, if any}
```

## Step 4: Delegate
Task tool, `subagent_type: "e2e-runner"`, pass the blob plus `project_root: <cwd>`.

## Step 5: Synthesize
Relay the verification report: **pass/fail per acceptance criterion** with evidence, each step (observed vs expected), console + network findings, and any captured artifacts (screenshots, GIF). Make clear no test file was saved; if the user wants a durable test, point them at playwright's `/playwright`.

From PDX-Apps/bench, MIT.

More from PDX-Apps/bench