Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Control a Playwright browser via CLI - navigate, interact, and screenshot
.claude/skills/aiskillstore-playwright-browser/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | -5% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 71% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 32% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 19% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -11% | 0% |
CLI tools for controlling a Playwright browser. Uses Playwright's accessibility tree for lightweight, efficient page interaction.
Start a persistent browser session and issue commands against it:
bash# Start session (opens visible browser window) bun .opencode/skill/playwright-browser/session.js start # Navigate to a site bun .opencode/skill/playwright-browser/navigate.js https://duckduckgo.com # Get page structure bun .opencode/skill/playwright-browser/snapshot.js # Type in search box and submit bun .opencode/skill/playwright-browser/type.js "combobox:Search with DuckDuckGo" "github" --press Enter # Click on a result bun .opencode/skill/playwright-browser/click.js "link:GitHub" # Take a screenshot bun .opencode/skill/playwright-browser/screenshot.js --output result.png # Stop session when done bun .opencode/skill/playwright-browser/session.js stop
Manage a persistent browser session for interactive use.
bashbun .opencode/skill/playwright-browser/session.js start bun .opencode/skill/playwright-browser/session.js stop bun .opencode/skill/playwright-browser/session.js status
Options:
--headless - Run in headless mode (default: headed/visible)Examples:
bashbun .opencode/skill/playwright-browser/session.js start bun .opencode/skill/playwright-browser/session.js start --headless bun .opencode/skill/playwright-browser/session.js status bun .opencode/skill/playwright-browser/session.js stop
Navigate to a URL in the browser.
bashbun .opencode/skill/playwright-browser/navigate.js <url>
Examples:
bashbun .opencode/skill/playwright-browser/navigate.js https://example.com
Get the accessibility tree of the current page.
bashbun .opencode/skill/playwright-browser/snapshot.js
Shows the page structure in a text format. Use this to find elements to interact with.
Output format:
- navigation:
- link "About"
- search:
- combobox "Search"
- button "Submit"Type text into an input element on the page.
bashbun .opencode/skill/playwright-browser/type.js <ref> <text> [--press Enter]
Arguments:
ref - Element reference (role:name, text:content, or CSS selector)text - Text to typeOptions:
--press <key> - Press a key after typing (e.g., Enter, Tab)Examples:
bashbun .opencode/skill/playwright-browser/type.js "combobox:Search" "hello" --press Enter bun .opencode/skill/playwright-browser/type.js "input[name='q']" "search query"
Click an element on the page.
bashbun .opencode/skill/playwright-browser/click.js <ref>
Arguments:
ref - Element reference (role:name, text:content, or CSS selector)Examples:
bashbun .opencode/skill/playwright-browser/click.js "button:Submit" bun .opencode/skill/playwright-browser/click.js "link:Sign in"
Take a screenshot of the current page.
bashbun .opencode/skill/playwright-browser/screenshot.js [--output <path>] [--full]
Options:
--output <path> - Output file path (default: screenshot.png)--full - Capture full page (scrolls entire page)Examples:
bashbun .opencode/skill/playwright-browser/screenshot.js bun .opencode/skill/playwright-browser/screenshot.js --output page.png
Commands that interact with elements (type, click) accept flexible element references:
| Format | Example | Description | |--------|---------|-------------| | role:name | combobox:Search | Accessibility role and name | | text:content | text:Sign in | Element containing text | | CSS selector | #login-btn | Standard CSS selector | | CSS selector | input[name='q'] | Attribute selector |
Recommended workflow:
snapshot to see the accessibility treerole:name format for reliable interactionCommands provide helpful error messages with suggestions:
.playwright-data/ in the project rootnetworkidle wait for reliable page loading| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 26,877 | 18,441 | -31% | 1 | 1 | 0% | 1,860 | 1,710 | -8% | 0 | 0 | — |
case-02 | fail→fail | 19,525 | 5,344 | -73% | 1 | 1 | 0% | 570 | 1,541 | +170% | 0 | 0 | — |
case-03 | fail→fail | 12,780 | 17,250 | +35% | 1 | 1 | 0% | 1,221 | 1,603 | +31% | 0 | 0 | — |
case-04 | fail→pass | 18,183 | 5,583 | -69% | 1 | 1 | 0% | 2,547 | 2,410 | -5% | 0 | 0 | — |
case-05 | fail→pass | 5,781 | 7,147 | +24% | 1 | 1 | 0% | 891 | 1,524 | +71% | 0 | 0 | — |
case-06 | fail→pass | 11,588 | 8,288 | -28% | 1 | 1 | 0% | 1,093 | 1,448 | +32% | 0 | 0 | — |
case-07 | fail→pass | 13,442 | 1,907 | -86% | 1 | 1 | 0% | 1,281 | 1,518 | +19% | 0 | 0 | — |
case-08 | fail→pass | 15,365 | 6,553 | -57% | 1 | 1 | 0% | 1,635 | 1,451 | -11% | 0 | 0 | — |
case-09 | fail→pass | 8,416 | 6,945 | -17% | 1 | 1 | 0% | 1,333 | 1,533 | +15% | 0 | 0 | — |
case-10 | fail→pass | 17,241 | 1,831 | -89% | 1 | 1 | 0% | 1,519 | 1,493 | -2% | 0 | 0 | — |
case-11 | fail→pass | 11,001 | 2,688 | -76% | 1 | 1 | 0% | 992 | 1,568 | +58% | 0 | 0 | — |
case-12 | fail→pass | 15,945 | 2,584 | -84% | 1 | 1 | 0% | 2,664 | 1,646 | -38% | 0 | 0 | — |
case-13 | fail→pass | 10,131 | 7,013 | -31% | 1 | 1 | 0% | 891 | 1,508 | +69% | 0 | 0 | — |
case-14 | fail→pass | 8,673 | 1,730 | -80% | 1 | 1 | 0% | 625 | 1,499 | +140% | 0 | 0 | — |
case-15 | fail→pass | 10,792 | 6,603 | -39% | 1 | 1 | 0% | 914 | 1,465 | +60% | 0 | 0 | — |
case-16 | fail→pass | 12,386 | 7,018 | -43% | 1 | 1 | 0% | 1,215 | 1,559 | +28% | 0 | 0 | — |
case-17 | fail→pass | 14,826 | 11,121 | -25% | 1 | 1 | 0% | 2,336 | 2,174 | -7% | 0 | 0 | — |
case-18 | fail→pass | 16,816 | 6,647 | -60% | 1 | 1 | 0% | 1,931 | 1,436 | -26% | 0 | 0 | — |
case-19 | fail→pass | 14,917 | 8,916 | -40% | 1 | 1 | 0% | 1,666 | 1,869 | +12% | 0 | 0 | — |
case-20 | pass→pass | 9,611 | 2,985 | -69% | 1 | 1 | 0% | 774 | 1,736 | +124% | 0 | 0 | — |
case-21 | pass→pass | 11,562 | 7,124 | -38% | 1 | 1 | 0% | 2,215 | 2,656 | +20% | 0 | 0 | — |
case-22 | pass→pass | 6,178 | 3,984 | -36% | 1 | 1 | 0% | 1,242 | 2,077 | +67% | 0 | 0 | — |
case-23 | fail→pass | 10,301 | 2,378 | -77% | 1 | 1 | 0% | 1,757 | 1,582 | -10% | 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, and 20 counted toward the lift figure. The other 3 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +74 percentage points is the difference between those two pass rates over the 20 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.