Playwright Skill (OpenAI) for Codex: What It Does and How to Install It

Makes the agent control a real browser from the terminal with Playwright's CLI: open pages, fill forms, take screenshots.

Installs: 8.9K; on skills.sh
Stars: 28.4K; on GitHub
Files: 8; besides SKILL.md
Updated: Oct 2026; last repo commit

The agent opens a page, takes a snapshot that gives each element a reference such as e15, and then clicks, fills, types or presses keys using those references. It takes a new snapshot after navigation or big page changes, because references go stale. SKILL.md says to treat this as CLI-first work and not to switch to @playwright/test unless you ask for test files.

It needs npx, which comes with Node.js. A bundled wrapper script runs the Playwright CLI through npx, so a global install is optional. The skill is written for Codex: SKILL.md points the wrapper at the Codex skills folder. It is based on the playwright-cli skill from Microsoft, adapted by OpenAI, and is not by Microsoft or the Playwright team. Its repo README says the openai/skills repository is deprecated. For the Playwright MCP server from Microsoft, see the Playwright MCP server page.

Good for

Codex users who want the agent to run a web flow in a real browser
Developers debugging a UI flow with snapshots, screenshots and traces
Anyone pulling data out of a page without writing Playwright test files

Install playwright

Codex

Copy this prompt.

Install the playwright skill in Codex. Run this command:

$skill-installer playwright

Then check that it loaded and tell me how to use it.

2
Paste it into Codex. It shows the change and asks before running it.

Requires Node.js.

The README for this repo installs curated skills in Codex with the skill installer, which is the first Codex route. The skill needs npx, which comes with Node.js. SKILL.md points the wrapper script at the Codex skills folder, and the repo documents installs for Codex only, so there is no Claude Code or Claude app route here.

From the project README. Source.

What's in the skill

When it loads

Use when the task requires automating a real browser from the terminal (navigation, form filling, snapshots, screenshots, data extraction, UI-flow debugging) via `playwright-cli` or the bundled wrapper script.

Sections in SKILL.md

Prerequisite check (required)
Skill path (set once)
Quick start
Core workflow
When to snapshot again
Recommended patterns
Form fill and submit
Debug a UI flow with traces
Multi-tab work
Wrapper script
References
Guardrails

Key facts

Author: OpenAI (openai/skills repo); adapted from the Microsoft playwright-cli skill
License: Apache-2.0
Installs on skills.sh: 8,915 (Nov 2, 2026)
GitHub stars: 28,447 for the whole repo (Nov 2, 2026)
Last repo commit: Oct 28, 2026
Audits on skills.sh: Gen Agent Trust Hub: pass, Socket: pass, Snyk: warn
What it includes: 9 files including SKILL.md: a wrapper script, 2 reference guides, an agent config, 2 icons, LICENSE.txt and NOTICE.txt
Needs: npx (Node.js); the wrapper downloads @playwright/cli
Works in: Codex (SKILL.md sets the CODEX_HOME skills path)

What it includes

Runs command -v npx and asks you to install Node.js if it is missing
Runs code · Prerequisite check
Runs npx --yes --package @playwright/cli playwright-cli with your arguments
Reads one optional variable, PLAYWRIGHT_CLI_SESSION, to name a browser session.
Runs code · Wrapper script: scripts/playwright_cli.sh
The core loop: open a page, snapshot for element refs, then act on a ref
Add --headed to watch the browser window.
Try: “Open the signup form, fill in the email field and submit it”
Runs code · open, snapshot, click, fill, type, press
Captures artifacts; tracing-start and tracing-stop record a trace of a flow
In a repo, SKILL.md says to put artifacts under output/playwright/.
Runs code · screenshot, pdf and tracing
tab-new, tab-list and tab-select for multi-tab work
Runs code · Tabs
Run JavaScript in the page; SKILL.md says to prefer explicit commands
references/workflows.md uses eval to read document.title or an element's text.
Runs code · eval and run-code
--session or PLAYWRIGHT_CLI_SESSION isolates work; playwright-cli.json sets options
Described in references/workflows.md.
Runs code · Sessions and config

From the official docs

Permissions

Medium risk

The agent can drive a real browser, so it can click and submit forms on any site you point it at. Run it on test accounts first and watch the first run with --headed.

What it can do

Read your data
Create and change things
Delete things
Read-only mode

Watch out for

It downloads and runs a package each time
The wrapper script runs npx --yes --package @playwright/cli, which fetches the package without a prompt. SKILL.md also allows a global install of @playwright/cli if you want a fixed copy. github.com
It can act on logged-in sites
The workflows show form fills and named sessions, so the agent can submit forms and click through a site in your name. Watch the first run with --headed, as SKILL.md suggests. github.com
eval and run-code run JavaScript in the page
SKILL.md says to prefer explicit commands over eval and run-code unless needed. github.com
Written for Codex paths
The wrapper path is $CODEX_HOME/skills/playwright/scripts/playwright_cli.sh, with ~/.codex as the default. github.com

Security advisories

No published security advisories as of Nov 2, 2026.

Technical details

Turning on read-only	No read-only switch exists.
Talks to	npm registry (@playwright/cli via npx). Any website the browser opens
Sign-in and keys	SKILL.md and the wrapper script name no key or token.
Scopes	Opens a real browser and visits any URL you or the agent choose Clicks, types, fills forms and takes screenshots, PDFs and traces Downloads and runs @playwright/cli through npx

Questions

What does the Playwright skill do?

It makes the agent drive a real browser through the Playwright command-line tool. It opens pages, clicks, fills forms, extracts data, takes screenshots and records traces for debugging a UI flow.

How do I install it?

Use the install box at the top of this page. The repo README installs curated skills in Codex with the skill installer, and the skills CLI route is listed too. The skill needs Node.js for npx.

Does it work in Claude Code or Claude?

SKILL.md points the wrapper script at the Codex skills folder, and the repo documents installs only for Codex. No Claude Code or Claude app route is documented, so this page lists only Codex and the skills CLI.

Does it need an API key?

No key is asked for in SKILL.md or in the wrapper script. The script reads one optional variable, PLAYWRIGHT_CLI_SESSION, which names a browser session.

Is it safe?

It lets the agent click, type and submit on any site you point it at, and the wrapper downloads and runs a package through npx with --yes. SKILL.md says to prefer explicit commands over eval and run-code, and suggests --headed when a visual check helps. Point it at test accounts when you can.

How is it different from Microsoft's Playwright tools?

It is OpenAI's adaptation of the playwright-cli skill from Microsoft's playwright-cli repo, made for Codex, and it is not published by Microsoft. For a connection-based option see the Playwright MCP server page.

Is the repo maintained?

The openai/skills README opens with a notice that the repository is deprecated and points to the OpenAI plugins repository for current examples.

Information

updated Nov 2, 2026

Repository: openai/skills
Folder: skills/.curated/playwright

openai/skills skills.sh

Sources

Docs, issue trackers and discussions this page is built from.

Show all 5

openai/skills README · github.com
playwright SKILL.md · github.com
playwright_cli.sh · github.com
playwright NOTICE.txt · github.com
playwright references/workflows.md · github.com