playwright-interactive
- Use it when:
- The agent is building a local web or Electron app in Codex and should check its own work in a real browser before calling it done.
- It does:
- Keeps Playwright's browser, page and Electron handles alive in a js_repl kernel, reloading for renderer changes and relaunching for main-process ones.
- Checks against:
- A QA inventory written first: every requirement, control, state and claim it means to sign off on, plus at least two off-happy-path scenarios.
- You get:
- Signoff only after functional QA with real input, a separate visual pass backed by screenshots, viewport-fit checks and a short exploratory pass.
- Needs:
- Codex with js_repl enabled, run with --sandbox danger-full-access for now, and Playwright installed with npm in the project.
- Unlike playwright:
- That one, OpenAI's CLI-based skill, drives any site command by command; this one is a QA loop for your own app with handles kept alive.
Codex only: needs its js_repl feature enabled and, for now, Codex started with sandboxing off (--sandbox danger-full-access).
TL;DR
A QA workflow from OpenAI's catalog of Codex skills, built on Codex's js_repl tool, a persistent JavaScript kernel. Setup installs Playwright in the project; a bootstrap cell loads it once, and top-level handles for the browser, its pages and the Electron app are reused across iterations. Web apps run at an explicit viewport by default (1600×900, or 390×844 with touch for mobile), with a separate native-window pass when window size or DPI matters; Electron always runs as a native window. Before testing, the agent writes a QA inventory from your requirements, what it actually built and the claims it plans to make, and maps every item to a check. Functional QA uses real clicks, keys and taps, takes each toggle through its full cycle and ends with 30 to 90 seconds of exploration. Visual QA is a separate pass over the initial viewport, the densest and post-interaction states and a smaller window, looking for clipping, contrast, spacing and weak aesthetics, with screenshots as evidence that metrics can't overrule. Cleanup is manual: leaving Codex doesn't close the browser or Electron app. The openai/skills repository is marked deprecated; its README points to openai/plugins for current Codex skills.
Summarizes the skill at 49f948f.
npx skills add https://github.com/openai/skills --skill playwright-interactive skills.sh pageSource on GitHubAll skills from openai/skills
- Total Downloads
- 5,700 on skills.sh, Nov 9, 2026
- Repo stars
- 28,602 openai/skills
- First seen
- Mar 5, 2026 on skills.sh
Weekly values
| Week to | Installs |
|---|---|
| Sep 21 | 181 |
| Sep 28 | 149 |
| Oct 5 | 199 |
| Oct 12 | 98 |
| Oct 19 | 112 |
| Oct 26 | 101 |
| Nov 2 | 121 |
| Nov 9 | 82 |
| Category |
Find root causes, review code and test what you built. Drive browsers, scrape pages and search the web. Everything for building software: planning, coding, debugging, design, frameworks, cloud, agents and skills. |
|---|---|
| Type |
A multi-step process the agent follows: interview, debug, run a TDD loop, review. |
| Effect |
Its own steps create or change files, git state or local tools. |
| Needs |
A specific command-line tool or desktop app must be installed, such as gh, agent-browser, LibreOffice, Obsidian or a vendor CLI. Drives a real Chromium or Chrome browser. |
| Publisher |
A verified publisher on skills.sh, or the company or maintainer behind the product the skill operates. |
| Topics | |
| Stack |
SKILL.md
No copy of its SKILL.md yet.