| name | playwright-cli |
| description | Browser-automation skill for terminal-driven Playwright work. Use when Codex needs to navigate websites, inspect pages, interact with forms, capture screenshots or traces, manage browser sessions, mock requests, or generate test code through a local `playwright-cli` workflow instead of writing a full test first. |
| metadata | {"author":"jovd83","version":"2.0.0","dispatcher-category":"testing","dispatcher-capabilities":"browser-automation-cli, playwright-cli-automation","dispatcher-accepted-intents":"drive_playwright_browser_cli","dispatcher-input-artifacts":"browser_task, target_url, automation_constraints","dispatcher-output-artifacts":"browser_run_artifact, cli_workflow, generated_steps","dispatcher-stack-tags":"playwright, cli, browser-automation","dispatcher-risk":"medium","dispatcher-writes-files":true} |
Playwright CLI
Use this skill when terminal-driven browser automation is the fastest path.
Assumptions
- A local
playwright-cli installation is available.
- If command names differ by version, inspect local help output before assuming an exact subcommand shape.
Operating Rules
- Snapshot or inspect the page before interacting with referenced elements.
- Prefer deterministic session naming for parallel or long-running work.
- Save and restore auth state deliberately when login is expensive.
- Capture traces before reproducing a failure, not after.
- Drop into custom code only when the built-in command surface is not enough.
Command Families
