菜单
Use Playwright for browser automation, UI validation, E2E tests, and interaction debugging. Useful for flows that need real page navigation, assertions, screenshots, or request interception.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Turn a specification into a concrete architecture with stable contracts, file boundaries, ownership, and risk analysis. Use this when acting as the architecture writer for a task.
Run the project verification stack after code changes. Trigger on code changes and stop on the first failure, fix it, then rerun the full stack from the top.
Verify that newly created files follow the repo's existing directory layout, naming conventions, and export patterns. Trigger when a new file is created.
Break architecture into execution-ready sub-tasks with clear ownership, dependencies, and contract references. Use this when acting as the planner for a task.
Final merge-readiness check after gates pass. Confirms verification is green, changed files are clean, and obvious release blockers are gone before merge.
Review schema and migration changes for safety, reversibility, rollout risk, and data integrity. Trigger when schema files or migrations change.
| name | playwright-browser-automation |
| description | Use Playwright for browser automation, UI validation, E2E tests, and interaction debugging. Useful for flows that need real page navigation, assertions, screenshots, or request interception. |
| owner | any |
| trigger | null |
| version | 2 |
Use this when you need to validate browser behavior with real interactions instead of static reasoning.
If Playwright is not already present in the repo, install it before writing tests.
getByRole, getByLabel, getByText, getByTitle, getByAltTextgetByTestIdreferences/playwright-patterns.md when you need concrete examples for navigation, assertions, network interception, auth reuse, or local test commands.If the app is canvas-driven, loading the page is not enough. Focus the canvas or visible start control, send at least one real input event, wait for the next rendered state, and record whether the game reacts without console or page errors.
Keep tests focused on user-visible behavior and resilient locators.