菜单
Programmatic web automation using Playwright Python — data extraction, form filling, multi-step workflows, auth session management, screenshot/PDF capture, and anti-detection patterns. Use for scripted automation tasks, not interactive MCP-based browser control.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Creates structured plans from requirements. Generates comprehensive plans with steps, dependencies, risks, and success criteria. Coordinates with specialist agents for planning input and validates plan completeness. Uses template-renderer for formatted output.
Create, validate, and convert skills for the agent ecosystem. Enforces standardized structure for consistency. Enables self-evolution by creating new skills on demand, converting MCP servers and codebases to skills.
Research-backed skill refresh workflow for updating existing skills with TDD checkpoints, memory-aware integration, and EVOLVE/reflection trigger handling.
Ensure accessibility in UI components including semantic HTML, ARIA attributes, keyboard navigation, and WCAG 2.2 AA compliance.
Use when you want to improve response quality through meta-cognitive reasoning. Applies 15+ reasoning methods to reconsider and refine initial outputs.
N-round opposing-stance debates for trade-off analysis. Assigns pro/con roles to agents, runs structured debate rounds with quality scoring, and produces a moderator synthesis with confidence-rated recommendation. Generalizable to architecture, technology, security, and design decisions.
| name | browser-automation |
| description | Programmatic web automation using Playwright Python — data extraction, form filling, multi-step workflows, auth session management, screenshot/PDF capture, and anti-detection patterns. Use for scripted automation tasks, not interactive MCP-based browser control. |
| version | 1.0.0 |
| model | sonnet |
| invoked_by | both |
| user_invocable | true |
| tools | ["Read","Write","Bash","Glob","Grep"] |
| args | <url> [--extract] [--fill-form] [--screenshot] [--pdf] [--auth-state <path>] [--headless] |
| agents | ["developer","qa","researcher"] |
| category | Automation |
| tags | ["playwright","browser","automation","scraping","data-extraction","forms","screenshots","pdf","auth"] |
| best_practices | ["Use storage state to persist authenticated sessions across runs","Prefer role-based locators (get_by_role, get_by_label) over CSS selectors","Always use context managers (with sync_playwright() as p) for resource cleanup","Set realistic user-agent and viewport to avoid bot detection","Use page.wait_for_selector() instead of time.sleep() for dynamic content"] |
| error_handling | graceful |
| streaming | supported |
| verified | true |
| lastVerifiedAt | 2026-03-13 |
| source | builtin |
| trust_score | 100 |
| provenance_sha | a1970631f58f8bdd |
pip install playwright && playwright install chromium
pip install playwright-stealth # optional, for anti-detection
| Skill | Use Case |
|---|---|
browser-automation | Scripted automation, data pipelines, form bots |
webapp-testing | QA assertions, visual regression, test suites |
chrome-browser | Interactive MCP DevTools control |
NEVER USE time.sleep() — use page.wait_for_selector() or wait_for_load_state()
with sync_playwright() as p:
context = p.chromium.launch(headless=True).new_context(
user_agent="Mozilla/5.0 ...", viewport={"width": 1280, "height": 720})
page = context.new_page()
page.goto(url); page.wait_for_load_state("networkidle")
items = [i.inner_text() for i in page.get_by_role("listitem").all()]
page.get_by_label("Email").fill("user@example.com")
page.get_by_label("Password").fill("secret")
with page.expect_navigation():
page.get_by_role("button", name="Submit").click()
page.screenshot(path="full.png", full_page=True)
page.pdf(path="page.pdf", format="A4", print_background=True)
# Save session
context.storage_state(path="auth-state.json")
# Reuse session
context = browser.new_context(storage_state="auth-state.json")
browser = p.chromium.launch(args=["--disable-blink-features=AutomationControlled"])
context = browser.new_context(user_agent="...", locale="en-US", timezone_id="America/New_York")
page.add_init_script("Object.defineProperty(navigator, 'webdriver', {get: () => undefined})")
.claude/context/tmp/ as structured JSON.claude/context/artifacts/auth-state.json — add to .gitignore| Anti-Pattern | Correct Approach |
|---|---|
time.sleep() for waits | page.wait_for_selector() / wait_for_load_state() |
| CSS selectors for forms | get_by_label(), get_by_role() |
| Re-login each run | context.storage_state() persistence |
| Committing auth state | Add auth-state.json to .gitignore |
Before starting: Read .claude/context/memory/learnings.md
After completing:
.claude/context/memory/learnings.md.claude/context/memory/issues.mdASSUME INTERRUPTION: If it's not in memory, it didn't happen.