playwright-cli Browser Automation Skill
Drives Playwright from the command line so an AI agent can navigate sites, fill forms, capture screenshots, and extract data without you writing browser scripts.
A browser-automation example skill (open browser, click, fill forms, screenshot) — same risk category as other Playwright-family skills (id 758): fundamentally arbitrary browser-interaction capability; here it's a framework example, not an independently published production skill.
What does this skill do, and when should you use it?
playwright-cli is an Agent Skill bundled in the VoltAgent repository's examples directory that packages browser automation into a directly callable CLI. It works through a snapshot loop: a snapshot returns element refs (like e15), which the agent then uses for clicks, typing, and uploads. Beyond core interaction, it covers multi-tab handling, cookies and web storage, request mocking, console/network inspection, tracing, and video recording. Its frontmatter uses Claude Code's `allowed-tools: Bash(playwright-cli:*)` syntax, so it is a Claude Code-style skill file that needs minor edits to port elsewhere.
Runs the playwright-cli command-line tool: opens browser sessions (choosing chrome/firefox/webkit/msedge or a persistent profile), navigates pages, uses snapshot to obtain element refs, then performs click/type/fill/select/upload/hover/drag actions, captures page or element screenshots, exports PDFs, evaluates JavaScript (eval / run-code), manages tabs, saves and loads cookies/localStorage/sessionStorage, mocks network responses with route, inspects console and network logs, and runs tracing or records WebM video. Seven references/ documents cover request mocking, test generation, session management, and more.
- A frontend developer wants an agent to run end-to-end checks on a local or live web app: open pages, interact, and verify via snapshots and screenshots
- A QA engineer needs to batch-fill forms, submit them, and capture result screenshots instead of manual regression passes
- A data analyst extracting information from pages where a saved storage state can restore an authenticated session
- A debugger investigating frontend issues by reading console errors and network requests, replaying steps under tracing
- A QA team generating test code or recording operation videos to attach to bug reports
What are this skill's strengths and limitations?
- Extremely broad command coverage: interaction, tabs, storage, network mocking, DevTools, tracing, and video in one tool
- Snapshot-ref interaction model is LLM-friendly and more stable than coordinates or CSS selectors
- Named sessions and persistent profiles suit multi-task workflows that need login state
- Strong documentation with ready-made form, multi-tab, and debugging examples plus seven reference guides
- Frontmatter uses Claude Code's `allowed-tools: Bash(...)` syntax and needs manual adjustment on other platforms
- Standalone installation of the playwright-cli binary is not fully documented in the source
- Requires a real browser environment and network access; sandboxed or offline setups cannot run it
- The repo README describes the whole VoltAgent platform and says nothing about this skill's capabilities — evaluation must rely on SKILL.md alone
How do you install this skill?
The skill lives at examples/with-workspace/workspace/skills/playwright-cli/ in the VoltAgent repo, containing SKILL.md and a references/ folder. Copy that folder into your Agent Skills client's skills directory. SKILL.md mentions playwright-cli install --skills and playwright-cli install-browser, but the source does not document a standalone installation channel for the playwright-cli binary itself — consult its own docs for that.
How do you use this skill?
After installing a browser, trigger it with natural language, e.g.: 'Open https://example.com, fill in [email protected], submit the form, then take a screenshot.' The agent follows the SKILL.md flow: playwright-cli open <url> → snapshot to get refs → fill/click interactions → screenshot → close. Use --browser=chrome to pick a browser and -s=<name> to manage multiple concurrent named sessions.
How does this skill compare with similar options?
The source names no direct competitor. A sponsor blurb in the README claims Ego Lite is 3.45x faster than agent-browser (Vercel), but that is sponsored content about the VoltAgent platform overall and cannot serve as an objective comparison for this skill.