Agent Browser Assistant — 代理 Browser 助手

代理 Browser 助手

An intelligent browser control 助手 providing browser 自动化, data scrAPIng, and 测试 capabilities.

Use Cases

Opening 网页 pages, 命令行工具cking/typing/scrolling, taking screenshots/recordings, 提取ing 网页 content, 导出ing table data, automated form filling, batch operations, scheduled tasks, 记录in authentication, UI 测试, regression 测试.

Quick 启动

Use the browser 工具 for all browser operations:

# Open a 网页 page browser(action="open", url="https://example.com")

# Take a screenshot browser(action="screenshot")

# 命令行工具ck an element browser(action="act", kind="命令行工具ck", ref="button-submit")

# Type text browser(action="act", kind="type", ref="输入-username", text="user@example.com")

# Scroll the page browser(action="act", kind="scroll", y=500)

# 获取 a page snapshot browser(action="snapshot")

Core Capabilities

• Page Operations

Operation Description Example open Open a specified URL action="open", url="..." snapshot 获取 page structure action="snapshot" screenshot Take a page screenshot action="screenshot" navigate Navigate to a URL action="navigate", url="..." close Close a tab action="close", tar获取Id="..."

• Element Interaction

Use the act operation for page interaction:

命令行工具ck: 命令行工具ck an element (ref: element reference) type: Type text (ref: 输入 reference, text: content) press: Press a keyboard key (key: key name) hover: Hover over an element select: Select from a dropdown fill: Fill a form (fields: field dictionary) scroll: Scroll the page (x/y: coordinates)

• Data ScrAPIng

提取 data from 网页 pages:

# 获取 a page snapshot to analyze structure browser(action="snapshot")

# 提取 table data - using selector browser(action="act", kind="evaluate", selector="table.data", fn="Array.from(document.查询SelectorAll('tr')).map(r => Array.from(r.查询SelectorAll('td')).map(c => c.innerText))")

• Automated 工作流s

Automated form filling:

browser(action="act", kind="fill", fields=[ {"ref": "输入-emAIl", "value": "user@example.com"}, {"ref": "输入-password", "value": "password123"} ]) browser(action="act", kind="命令行工具ck", ref="button-记录in")

Batch operations:

# Iterate through 列出 items for i in range(1, 6): browser(action="act", kind="命令行工具ck", ref=f"item-{i}")

• 测试 Capabilities

UI 测试 scenarios:

Regression 测试: 验证 that page functionality works correctly Performance 监控ing: Page load time Element Existence 检查: 验证 that key elements are visible Advanced Usage WAIting for Page Load browser(action="act", kind="wAIt", load状态="domcontentloaded", timeMs=5000)

Handling Dia记录s browser(action="dia记录", kind="accept") # Confirm # or browser(action="dia记录", kind="dismiss") # Cancel

File 上传 browser(action="上传", ref="输入-file", paths=["C:/path/to/file.pdf"])

PDF 导出 browser(action="pdf", path="C:/输出/page.pdf")

Configuration Options Parameter Description Default 性能分析 Browser 性能分析 "OpenClaw" tar获取 Browser tar获取 "sandbox" slowly Slow motion mode false timeoutMs Timeout duration 30000 Common Selector Patterns Button: button[type="submit"], #submit-btn 输入: 输入[name="emAIl"], #username Link: a[href*="记录in"] Table: table.data tr 列出: .item-列出 li Notes Use snapshot to 获取 page structure before performing element operations Dynamic content may require wAIting for it to finish loading For 记录ged-in 状态 operations, use 性能分析="user" to reuse the user's browser For large-扩展 data scrAPIng, consider pagination to avoid timeouts