Google's official agent skill for writing production Maps Platform code — grounded in freshly fetched docs, with a demo key path that needs no billing account.
Playwright CLISkill
Summary
Microsoft's official Playwright skill — drives a real browser from the command line using accessibility snapshots and element refs, and plans, generates and heals Playwright tests.
Features
- Drives Chrome, Edge, Firefox and WebKit from the command line
- Accessibility-snapshot loop with element refs instead of screenshots
- Role, test-id and CSS locators, plus find/--regex to grep large snapshots
- Network routing and response mocking; console and request inspection
- Traces, video recording with chapters, and run-code for arbitrary Playwright script
- state-save / state-load to reuse an authenticated browser session
- Test generation: plan, generate and heal Playwright tests
- show --annotate opens a human UI-review loop with annotated screenshots
- Experimental WebMCP support, with page-supplied tools treated as untrusted
- Emulates colour scheme, reduced motion, forced colors, contrast and print media
Install This Skill
Add this skill to your favorite AI agent in a few steps.
Skill Content
Usage Instructions
Learn how to use this skill with different AI agents.
Example Usage
playwright-cli open https://example.com/form
playwright-cli snapshot
playwright-cli fill e1 "user@example.com"
playwright-cli fill e2 "password123"
playwright-cli click e3
playwright-cli snapshot
playwright-cli closeDescription
The Playwright CLI skill is Microsoft's official Agent Skill for playwright-cli, the command-line front end to Playwright. It is aimed at coding agents that need to actually drive a browser — fill a form, check that a fix renders, generate a test — rather than reason about a page they have never loaded.
The design choice that matters is that the agent works from accessibility snapshots rather than screenshots. After every command the CLI prints the page's state as a YAML tree in which each element carries a short ref (e15), and the agent clicks, fills and hovers by ref. That keeps the loop cheap in tokens compared with sending images, and it makes the interaction deterministic. CSS selectors and Playwright locators such as getByRole('button', { name: 'Submit' }) work too, and find greps a large snapshot with surrounding context instead of capturing the whole thing.
The command surface is wide: navigation and tabs, keyboard and mouse primitives, file upload and drag-and-drop, dialogs, cookie/localStorage/sessionStorage inspection and state-save/state-load for reusing an authenticated session, network routing for mocking responses, and emulation of colour scheme, reduced motion, forced colors, contrast and print media. On the debugging side it exposes console output, a request log, Playwright traces, video recording with chapter markers, and run-code for arbitrary Playwright script. show --annotate opens a dashboard where a human draws boxes on the live page and types comments, and the agent receives the annotated screenshot plus the user's notes — a genuine review loop rather than a one-way capture.
Two details are worth calling out. It supports the experimental WebMCP API, where a page registers its own tools for agents; the skill instructs the agent to prefer those calls over driving the UI, while treating the page-supplied names, schemas and results as untrusted input. And it documents attaching screenshots and short videos to pull requests through gh --attach, which is the difference between a reviewer trusting a UI fix and having to check it out.
Bundled reference files cover running and debugging Playwright test suites, request mocking, session management, storage state, test generation (plan / generate / heal), tracing, video recording, PR attachments and element attributes. Installation is playwright-cli install --skills for the Claude Code layout, --skills=agents for .agents/skills, or -g to install into the home directory; it works with Claude Code, GitHub Copilot, Cursor and any agent that reads locally installed skills. Apache-2.0.
Related Skills
Railway's official agent skill: create projects, provision databases and buckets, deploy, manage variables and domains, and read build failures back — from the CLI, API or MCP server.
Diagnoses wrong gradients in differentiable NVIDIA Warp programs by measuring first — comparing autodiff against finite differences on a shrunk reproduction before proposing any fix.
Google's official skill for the gws CLI — drive Gmail, Drive, Calendar, Sheets, Docs, Chat and Admin APIs from an agent, with Model Armor screening.