Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Browser automation via Playwright CLI. Open pages, interact with elements, take screenshots, and more. Ideal for coding agents and automated testing workflows.
.claude/skills/sundial-org-playwright-cli/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | -23% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 4% | 0% |
| case-01 | ✗→✓ | ▲ Improved | 22% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 10% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 14% | 0% |
Browser automation via Playwright. Token-efficient CLI for coding agents.
bashnpm install -g @playwright/mcp@latest playwright-cli --help
| Command | Description | |---------|-------------| | playwright-cli open <url> | Open URL in browser | | playwright-cli close | Close the page | | playwright-cli type <text> | Type text into editable element | | playwright-cli click <ref> [button] | Click on element | | playwright-cli dblclick <ref> [button] | Double click | | playwright-cli fill <ref> <text> | Fill text into field | | playwright-cli drag <startRef> <endRef> | Drag and drop | | playwright-cli hover <ref> | Hover over element | | playwright-cli check <ref> | Check checkbox/radio | | playwright-cli uncheck <ref> | Uncheck checkbox | | playwright-cli select <ref> <val> | Select dropdown option | | playwright-cli snapshot | Capture page snapshot for refs |
bashplaywright-cli go-back # Go back playwright-cli go-forward # Go forward playwright-cli reload # Reload page
bashplaywright-cli press <key> # Press key (a, arrowleft, enter...) playwright-cli keydown <key> # Key down playwright-cli keyup <key> # Key up playwright-cli mousemove <x> <y> # Move mouse playwright-cli mousedown [button] # Mouse down playwright-cli mouseup [button] # Mouse up playwright-cli mousewheel <dx> <dy> # Scroll
bashplaywright-cli screenshot [ref] # Screenshot page or element playwright-cli pdf # Save as PDF
bashplaywright-cli tab-list # List all tabs playwright-cli tab-new [url] # Open new tab playwright-cli tab-close [index] # Close tab playwright-cli tab-select <index> # Switch tab
bashplaywright-cli console [min-level] # View console messages playwright-cli network # View network requests playwright-cli run-code <code> # Run JS snippet playwright-cli tracing-start # Start trace playwright-cli tracing-stop # Stop trace
bashplaywright-cli session-list # List sessions playwright-cli session-stop [name] # Stop session playwright-cli session-stop-all # Stop all playwright-cli session-delete [name] # Delete session data
bashplaywright-cli open https://example.com --headed
bash# Open and interact playwright-cli open https://example.com playwright-cli type "search query" playwright-cli press Enter playwright-cli screenshot # Use sessions playwright-cli open https://site1.com playwright-cli --session=project-a open https://site2.com
| Variable | Description | |----------|-------------| | PLAYWRIGHT_MCP_BROWSER | Browser: chrome, firefox, webkit, msedge | | PLAYWRIGHT_MCP_HEADLESS | Run headless (default: headed) | | PLAYWRIGHT_MCP_ALLOWED_HOSTS | Comma-separated allowed hosts | | PLAYWRIGHT_MCP_CONFIG | Path to config file |
Create playwright-cli.json for persistent settings:
json{ "browser": { "browserName": "chromium", "headless": false }, "outputDir": "./playwright-output", "console": { "level": "info" } }
--session flag for isolated browser instanceshttps://github.com/microsoft/playwright-cli
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-03 | fail→pass | 13,602 | 4,997 | -63% | 1 | 1 | 0% | 2,587 | 1,982 | -23% | 0 | 0 | — |
case-02 | fail→pass | 12,721 | 5,616 | -56% | 1 | 1 | 0% | 2,059 | 2,144 | +4% | 0 | 0 | — |
case-01 | fail→pass | 8,430 | 5,291 | -37% | 1 | 1 | 0% | 1,452 | 1,774 | +22% | 0 | 0 | — |
case-04 | fail→pass | 7,118 | 1,988 | -72% | 1 | 1 | 0% | 1,112 | 1,220 | +10% | 0 | 0 | — |
case-05 | pass→pass | 9,272 | 2,554 | -72% | 1 | 1 | 0% | 1,606 | 1,435 | -11% | 0 | 0 | — |
case-06 | fail→pass | 7,068 | 2,426 | -66% | 1 | 1 | 0% | 1,282 | 1,457 | +14% | 0 | 0 | — |
case-07 | fail→pass | 9,078 | 1,787 | -80% | 1 | 1 | 0% | 1,533 | 1,276 | -17% | 0 | 0 | — |
case-08 | pass→pass | 11,383 | 2,996 | -74% | 1 | 1 | 0% | 1,915 | 1,528 | -20% | 0 | 0 | — |
case-09 | fail→pass | 11,005 | 2,476 | -78% | 1 | 1 | 0% | 2,081 | 1,386 | -33% | 0 | 0 | — |
case-10 | pass→pass | 13,918 | 2,445 | -82% | 1 | 1 | 0% | 2,389 | 1,483 | -38% | 0 | 0 | — |
case-11 | fail→pass | 8,401 | 1,442 | -83% | 1 | 1 | 0% | 1,369 | 1,216 | -11% | 0 | 0 | — |
case-12 | fail→pass | 8,188 | 3,275 | -60% | 1 | 1 | 0% | 1,376 | 1,583 | +15% | 0 | 0 | — |
case-13 | pass→pass | 7,269 | 1,680 | -77% | 1 | 1 | 0% | 1,086 | 1,278 | +18% | 0 | 0 | — |
case-14 | pass→pass | 5,050 | 1,865 | -63% | 1 | 1 | 0% | 798 | 1,310 | +64% | 0 | 0 | — |
case-15 | pass→pass | 6,466 | 1,298 | -80% | 1 | 1 | 0% | 1,143 | 1,191 | +4% | 0 | 0 | — |
case-16 | pass→pass | 10,341 | 1,958 | -81% | 1 | 1 | 0% | 1,651 | 1,306 | -21% | 0 | 0 | — |
case-17 | fail→pass | 16,316 | 1,567 | -90% | 1 | 1 | 0% | 2,360 | 1,280 | -46% | 0 | 0 | — |
case-18 | fail→pass | 12,146 | 2,585 | -79% | 1 | 1 | 0% | 1,935 | 1,286 | -34% | 0 | 0 | — |
case-19 | pass→pass | 4,574 | 2,815 | -38% | 1 | 1 | 0% | 828 | 1,294 | +56% | 0 | 0 | — |
case-20 | pass→pass | 4,943 | 1,781 | -64% | 1 | 1 | 0% | 799 | 1,308 | +64% | 0 | 0 | — |
case-21 | pass→pass | 4,333 | 3,511 | -19% | 1 | 1 | 0% | 815 | 1,630 | +100% | 0 | 0 | — |
case-22 | pass→pass | 7,732 | 6,687 | -14% | 1 | 1 | 0% | 1,443 | 2,370 | +64% | 0 | 0 | — |
case-23 | pass→pass | 3,658 | 3,545 | -3% | 1 | 1 | 0% | 705 | 1,733 | +146% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 23 cases were attempted. The headline lift of +48 percentage points is the difference between those two pass rates over the 23 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.