AgentSkillsCN

writing-browser-app-e2e-tests

在 browser-app-e2e-tests 包中编写 Playwright 端到端测试。

SKILL.md
--- frontmatter
name: writing-browser-app-e2e-tests
description: Write a Playwright e2e test in the browser-app-e2e-tests package.

Writing browser-app e2e tests

When to use this skill

When you need to write a Playwright e2e test for browser-app package.

Base instructions

  • •Directory: packages/tests/browser-app-e2e-tests/src/scenarios
  • •Test title: three-digit identifier + dot + short description of the test
  • •File name: three-digit sequential identifier + - + snake-cased-short-description + .test.ts

For assertions, ONLY use visual comparisons of screenshots via VisualEvaluator.expectToSee. DO NOT use other assertions.

Test structure

Tests are structured as multi-step scenarios in which the simulated user tries to reach a specific goal. Each test starts from the empty state of the application, so the first steps of the scenarios should be about configuring and creating everything needed for the later steps.

The Bazaar section of the app contains ready-made "packs" of collection categories, collections, and apps, that can be installed to create a useful starting point.

Technically, a test is a sequence of Playwright test steps. Each step has two sections (visually separated by // $SectionName comments):

  • •Exercise: takes actions and waits for them to have effect.
  • •Verify: calls VisualEvaluator.expectToSee exactly once, which:
    • •Takes a screenshot of either a specific element (preferred) or the whole page.
    • •Describes the current state it expects to see in the screenshot. (No references to what happened before.)

Additional requirements:

  • •Step title: two-digit-index + dot + title. Example 00. Go to Homepage.
  • •Steps must be ordered correctly. I.e., 00. $title must go before 01. $otherTitle.
  • •Snapshot names must match the step index. Example: 00.png for step 00. $title.

Run test and evaluate screenshots

Run the test with yarn workspace @superego/browser-app-e2e-tests test:playwright.

When the test runs the first time there are no "golden reference snapshots" yet, so the test will fail.

However, the run generates the yet-to-be-evaluated snapshots in the packages/tests/browser-app-e2e-tests/src/scenarios/$TEST_FILE_NAME-snapshots directory.

One by one, visually review the snapshots and ensure they match the corresponding described expectation.

Reusable test logic

Reusable test logic is shared code that makes scenarios simpler and more reliable by extracting common selectors and interactions into locators, actions, and routines.

When writing a new locator, action, or routine:

  • •Check existing ones and follow the same style and patterns.
  • •Follow the specific guidelines written below.

Locators

The packages/tests/browser-app-e2e-tests/src/locators directory contains reusable locator builders.

Guidelines:

  • •Locators must not contain steps, assertions, or waits.
  • •Return Locators (or small grouped locator objects) and scope within containers when possible.
  • •Name by intent + target (e.g., bazaarPackCard, installPackButton).

Actions

The packages/tests/browser-app-e2e-tests/src/actions directory contains reusable functions that each perform a small single action shared across multiple tests.

Guidelines:

  • •Actions must not contain steps.
  • •Do one thing: a single user interaction (click, type, select, etc.).
  • •Be specific and UI-focused: name the exact intent + target (e.g., clickInstallPackButton), not generic helpers.
  • •Include only the minimal waits/guards needed to make the action reliable (e.g., ensure the element is visible/enabled before interacting).
  • •Do not chain multiple actions; if you need a sequence, create a routine.

Routines

The packages/tests/browser-app-e2e-tests/src/routines directory contains reusable functions that compose multiple actions into a single higher-level interaction. A routine performs a specific goal (e.g., "installing a pack", "filling-in a form") by calling actions in sequence and handling any expected intermediate UI transitions (waiting for navigation, toasts, modals, etc.). Check existing routines for examples.

Guidelines:

  • •Routines can contain actions, but must not contain steps.
  • •Keep routines goal-oriented and deterministic: given the same starting state, they should do the same thing.
  • •Prefer a small, stable API: accept explicit inputs and avoid hidden defaults.
  • •Use assertions/validation only if the routine needs a minimal guard to ensure it can proceed (e.g., wait for a page to be ready).
  • •Avoid branching logic when possible; if you need variants, create separate routines.

After writing a test

After writing a test, check other tests and see if there are duplicated locators, actions, or routines. If there are, suggest a refactor to the user.