Token导航 LogoToken导航TokenDH.com
研究检索敏感数据github未标认证来源可访问许可证需确认审计提醒

kernel-agent-browser内核 Agent 浏览器

Agent Skill

kernel-agent-browser 用于处理浏览器自动化、网页检查和页面信息提取,适合在 Codex、Claude、Cursor、Gemini CLI 中需要让 Agent 打开页面、读取网页或验证前端流程时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

总安装

8,323

周安装

340

GitHub Stars

3

下载量

2,666
CodexClaudeCursorGemini CLI

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

unknown

最后核验

2026-05-01

来源状态

来源可访问

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:kernel-agent-browser(内核 Agent 浏览器)
来源仓库:https://github.com/kernel/skills
仓库路径:skills/kernel-agent-browser
安装命令:
npx skills add https://github.com/kernel/skills --skill kernel-agent-browser
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/kernel/skills --skill kernel-agent-browser

简介

kernel-agent-browser 用于处理浏览器自动化、网页检查和页面信息提取。

  • 适合让 Agent 打开页面、读取网页或验证前端流程时使用。
  • 通过 npx skills add 命令从指定 GitHub 仓库安装,需结合原始 README 确认具体用法。
  • 安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写操作。
  • 适用宿主包括 Codex、Claude、Cursor、Gemini CLI,接入前应确认版本、权限和运行环境要求。

SKILL.md

Agent-Browser with Kernel Cloud Browsers

This skill documents best practices for using agent-browser's built-in Kernel provider (-p kernel) for cloud browser automation.

When to Use This Skill

Use this skill when you need to:

  • Automate websites using agent-browser -p kernel commands
  • Handle bot detection on sites with aggressive anti-bot measures
  • Persist login sessions across automation runs using profiles
  • Work with iframes including cross-origin payment forms
  • Get live view URLs for debugging or manual intervention
  • Find the underlying Kernel session ID for advanced Playwright scripting
  • Create site-specific automation skills for new websites

References

Prerequisites

Load the kernel-cli skill for Kernel CLI installation and authentication.

Environment Variables

Set these before your first agent-browser -p kernel call. The CLI holds state between invocations.

VariableDescriptionDefault
KERNEL_API_KEYRequired. Your Kernel API key for authentication(none)
KERNEL_HEADLESSRun browser in headless mode (true/false)false
KERNEL_STEALTHEnable stealth mode to avoid bot detection (true/false)true
KERNEL_TIMEOUT_SECONDSSession timeout in seconds300
KERNEL_PROFILE_NAMEBrowser profile name for persistent cookies/logins(none)

Recommended Configuration

export KERNEL_API_KEY="your-api-key"
export KERNEL_TIMEOUT_SECONDS=600     # 10-minute timeout for complex workflows
export KERNEL_STEALTH=true            # Avoid bot detection (default)
export KERNEL_PROFILE_NAME=mysite     # Persist login sessions across runs

Profile Persistence

When KERNEL_PROFILE_NAME is set:

  • The profile is created if it doesn't exist
  • Cookies, logins, and session data are automatically saved when the browser session ends
  • Future sessions with the same profile name restore the saved state

This is especially useful for sites requiring login—authenticate once, reuse across sessions.

Basic Usage

agent-browser -p kernel open <url>        # Navigate to page
agent-browser -p kernel snapshot -i       # Get interactive elements with refs
agent-browser -p kernel click @e1         # Click element by ref
agent-browser -p kernel fill @e2 "text"   # Fill input by ref
agent-browser -p kernel close             # Close browser and save profile

Always use the -p kernel flag with each command.

Semantic Selectors (Recommended)

Instead of ephemeral @e refs that change on every page load, use semantic selectors via the find command for more stable, readable automation:

# By ARIA role + accessible name (most stable)
agent-browser -p kernel find role button click --name "Log In"
agent-browser -p kernel find role textbox fill "user@email.com" --name "Email"

# By visible text content
agent-browser -p kernel find text "View Menus" click
agent-browser -p kernel find text "Submit Order" click

# By form label (great for inputs)
agent-browser -p kernel find label "Username" fill "myuser"
agent-browser -p kernel find label "Password" fill "secret123"

# By placeholder text
agent-browser -p kernel find placeholder "Search..." type "query"

# By data-testid (if the site uses them)
agent-browser -p kernel find testid "submit-btn" click

# By position (when needed)
agent-browser -p kernel find first "li.item" click
agent-browser -p kernel find nth 2 ".card" hover

When to Use Which Selector

Selector TypeBest ForStability
find role --nameButtons, links, navigation⭐⭐⭐ Most stable
find labelForm inputs with labels⭐⭐⭐ Most stable
find textClickable text elements⭐⭐ Stable
find testidSites with test attributes⭐⭐⭐ Most stable
find placeholderSearch boxes, inputs⭐⭐ Stable
@e refsUnknown sites, quick iteration⭐ Ephemeral

Recommendation: Use find for production automation. Use @e refs for exploration and quick prototyping, then convert to semantic selectors.

Finding Session ID and Live View URL

agent-browser creates a Kernel browser session under the hood. To get the session ID or live view URL:

# List all Kernel browsers (find yours by profile name or creation time)
kernel browsers list

# Get live view URL for a specific session
kernel browsers view <session-id>

This is useful when:

  • You need to execute Playwright scripts directly against the session
  • You want to share a live view URL with the user for manual intervention
  • You're debugging and want to watch the browser in real-time

Handling Bot Detection

Stealth Mode

Stealth mode (KERNEL_STEALTH=true) is enabled by default and helps avoid detection. However, some sites have aggressive bot detection that still triggers.

Manual Login Fallback

If login automation fails due to bot detection:

  1. Get the live view URL: kernel browsers list # Find your session by profile name kernel browsers view <session-id>
  2. Share the live view URL with the user and ask them to complete the login manually
  3. Once logged in, continue automation—the profile will save the authenticated state

JavaScript Fallback for Tricky Elements

Some elements (especially on bot-protected sites) don't respond to standard commands:

# Click by CSS selector
agent-browser -p kernel eval "document.querySelector('.submit-btn').click()"

# Fill by selector (with event dispatch)
agent-browser -p kernel eval "
  const el = document.querySelector('#email');
  el.value = 'user@example.com';
  el.dispatchEvent(new Event('input', {bubbles: true}));
  el.dispatchEvent(new Event('change', {bubbles: true}));
"

# Click by test ID
agent-browser -p kernel eval "document.querySelector('[data-testid=\"submit\"]').click()"

Anti-Bot Form Fields

Some payment processors (e.g., Point and Pay) use decoy form fields. Only fill fields matching specific patterns:

agent-browser -p kernel eval "
  const realInputs = Array.from(document.querySelectorAll('input'))
    .filter(el => el.name && el.name.startsWith('xeiinput'));
  // Fill only these inputs
"

Handling Iframes

Same-Origin Iframes

Use the frame command to switch context:

agent-browser -p kernel frame "#iframe-id"   # Switch to iframe
agent-browser -p kernel snapshot -i          # Snapshot within iframe
agent-browser -p kernel click @e1            # Interact within iframe
agent-browser -p kernel frame main           # Return to main frame

Cross-Origin Iframes

Cross-origin iframes require executing a Playwright script directly against the Kernel session:

  1. Find the session ID: kernel browsers list
  2. Execute a Playwright script: kernel browsers exec <session-id> --code " const frame = page.frameLocator('#payment-iframe'); await frame.locator('#card-number').fill('4111111111111111'); await frame.locator('#submit').click(); "

See the kernel-cli skill for more details on executing Playwright code.

Waiting Strategies

Smart waits are critical for fast, reliable automation. Using condition-based waits instead of fixed timeouts can reduce execution time by 50%+ while improving reliability.

Smart Waits (Recommended)

# Wait for page load states
agent-browser -p kernel wait --load domcontentloaded  # DOM ready
agent-browser -p kernel wait --load networkidle       # Network settled

# Wait for specific URL pattern (great for redirects after login)
agent-browser -p kernel wait --url "**/dashboard"
agent-browser -p kernel wait --url "**/order-confirmation"

# Wait for text to appear (great for dynamic content)
agent-browser -p kernel wait --text "Password"        # Field appeared
agent-browser -p kernel wait --text "Order confirmed" # Success message

# Wait for JavaScript condition
agent-browser -p kernel wait --fn "window.appReady === true"
agent-browser -p kernel wait --fn "document.querySelector('.spinner') === null"

# Wait for element by CSS selector
agent-browser -p kernel wait "#login-form"
agent-browser -p kernel wait ".results-loaded"

Fixed Waits (Last Resort)

# Only when no condition is available
agent-browser -p kernel wait 2000

Element Refs Best Practices

Element refs (@e1, @e2, etc.) are ephemeral and change:

  • After page navigation
  • After significant DOM updates
  • Between browser sessions

Always take a fresh snapshot before interacting:

agent-browser -p kernel snapshot -i
# Now use the refs from this snapshot
agent-browser -p kernel click @e5

Filtering Snapshots

# Filter for specific elements
agent-browser -p kernel snapshot -i | grep -i "button\|submit"

# Scope to a specific area
agent-browser -p kernel snapshot -s "#main-content" -i

Login Patterns

Single-Page Form (Optimized)

Username and password on the same page:

agent-browser -p kernel open https://example.com/login
agent-browser -p kernel wait --load domcontentloaded

# Use semantic selectors for stability
agent-browser -p kernel find label "Email" fill "user@example.com"
agent-browser -p kernel find label "Password" fill "secret123"
agent-browser -p kernel find role button click --name "Sign In"

# Wait for actual redirect, not arbitrary timeout
agent-browser -p kernel wait --url "**/dashboard"

Two-Step Form (Optimized)

Username first, then password on a second screen:

agent-browser -p kernel open https://example.com/login
agent-browser -p kernel wait --load domcontentloaded

# Step 1: Username
agent-browser -p kernel find label "Username" fill "myuser"
agent-browser -p kernel press Enter

# Wait for password field to appear (not a fixed sleep!)
agent-browser -p kernel wait --text "Password"

# Step 2: Password
agent-browser -p kernel find label "Password" fill "secret123"
agent-browser -p kernel press Enter

# Wait for successful redirect
agent-browser -p kernel wait --url "**/home"

Comparison: Old vs Optimized

# OLD (slow, fragile):
agent-browser -p kernel wait 2000
agent-browser -p kernel fill @e1 "username"
agent-browser -p kernel wait 2000
agent-browser -p kernel fill @e3 "password"
agent-browser -p kernel wait 5000

# OPTIMIZED (fast, stable):
agent-browser -p kernel wait --load domcontentloaded
agent-browser -p kernel find label "Username" fill "username"
agent-browser -p kernel wait --text "Password"
agent-browser -p kernel find label "Password" fill "password"
agent-browser -p kernel wait --url "**/dashboard"

Modal Login

Login form appears in a modal overlay:

# Click login link to open modal
agent-browser -p kernel find text "Log In" click
agent-browser -p kernel wait --text "Password"  # Wait for modal

# Fill modal fields
agent-browser -p kernel find label "Email" fill "user@example.com"
agent-browser -p kernel find label "Password" fill "password123"
agent-browser -p kernel find role button click --name "Sign In"
agent-browser -p kernel wait --url "**/dashboard"

Fallback: JavaScript for Tricky Modals

Some modals don't expose accessible labels:

agent-browser -p kernel eval "document.querySelector('.login-link').click()"
agent-browser -p kernel wait 1000

agent-browser -p kernel eval "
  document.getElementById('username').value = 'user@example.com';
  document.getElementById('username').dispatchEvent(new Event('input', {bubbles: true}));
  document.getElementById('password').value = 'password123';
  document.getElementById('password').dispatchEvent(new Event('input', {bubbles: true}));
  document.querySelector('button[type=submit]').click();
"
agent-browser -p kernel wait --url "**/dashboard"

Handling New Tabs

Some links open in new tabs:

# Click link that opens new tab
agent-browser -p kernel click @e38
agent-browser -p kernel tab 1           # Switch to new tab (0-indexed)
agent-browser -p kernel wait 2000
agent-browser -p kernel snapshot -i     # Interact with new tab

Screenshots and Debugging

# Take screenshot
agent-browser -p kernel screenshot ~/Downloads/page.png

# Full page screenshot
agent-browser -p kernel screenshot ~/Downloads/full.png --full

# View console messages
agent-browser -p kernel console

# View page errors
agent-browser -p kernel errors

# Get current URL
agent-browser -p kernel get url

Session Management

Cleanup

Always close the browser when done to save the profile:

agent-browser -p kernel close

Multiple Sessions

Run parallel browser sessions with named sessions:

agent-browser -p kernel --session site1 open https://site1.com
agent-browser -p kernel --session site2 open https://site2.com
agent-browser -p kernel session list

Common Gotchas

  1. Refs change after navigation: Always re-snapshot after clicking links or submitting forms.
  2. Wait after actions: Add waits after clicks/submits that trigger page loads or AJAX.
  3. Profile not saving: Make sure to run agent-browser -p kernel close to save the profile state.
  4. Timeout too short: Increase KERNEL_TIMEOUT_SECONDS for workflows with user pauses or slow pages.
  5. Stealth not working: Some sites detect bots despite stealth. Use manual login fallback.
  6. eval for stubborn elements: If fill or click don't work, try eval with direct DOM manipulation.
  7. Cross-origin iframes: Can't interact via agent-browser commands. Use Kernel's Playwright execution.

Quick Reference

# Start session with profile persistence
export KERNEL_PROFILE_NAME=mysite
export KERNEL_TIMEOUT_SECONDS=600
agent-browser -p kernel open https://example.com

# Basic interaction with semantic selectors (recommended)
agent-browser -p kernel wait --load domcontentloaded
agent-browser -p kernel find label "Email" fill "user@example.com"
agent-browser -p kernel find label "Password" fill "secret"
agent-browser -p kernel find role button click --name "Submit"
agent-browser -p kernel wait --url "**/success"

# Alternative: snapshot + refs (for exploration)
agent-browser -p kernel snapshot -i
agent-browser -p kernel fill @eN "text"
agent-browser -p kernel click @eM

# Get session info for manual intervention
kernel browsers list
kernel browsers view <session-id>

# Cleanup
agent-browser -p kernel close

Selector Cheat Sheet

# Buttons and links
agent-browser -p kernel find role button click --name "Submit"
agent-browser -p kernel find role link click --name "Next"
agent-browser -p kernel find text "Click here" click

# Form inputs
agent-browser -p kernel find label "Email" fill "user@example.com"
agent-browser -p kernel find placeholder "Search" type "query"
agent-browser -p kernel find testid "username-input" fill "myuser"

# Smart waits
agent-browser -p kernel wait --load domcontentloaded
agent-browser -p kernel wait --text "Success"
agent-browser -p kernel wait --url "**/dashboard"
agent-browser -p kernel wait --fn "window.loaded === true"

适合场景

01

用户想查找某类 Agent Skill 时

02

需要根据任务场景推荐可安装能力包时

03

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

Codex

39.47%
按下载量换算1,052

Claude

29.59%
按下载量换算789

Cursor

18.43%
按下载量换算491

Gemini CLI

8.76%
按下载量换算234

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

可疑

权限和风险

敏感数据

该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。来源安全扫描存在 warning/failed 结果,不能写成本站确认安全。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills