web-access
This skill provides web access capabilities for AI agents, including search, web scraping, and browser automation via Chrome DevTools Protocol (CDP). It supports both programmatic interaction and GUI-based navigation to handle dynamic content, login-protected sites, and anti-scraping measures.
Is web-access safe to install?
Review the source first: our audit of web-access's source files found 3 shell commands, 13 external URLs, file reads and writes (high risk). Every command and URL listed appears verbatim in the skill's source. The skill executes shell commands, interacts with local browser processes, and performs network requests. It can read and write to local configuration files and site-pattern references.
How we audit skills: our security review methodology.
Who is this skill for?
AI agents requiring internet connectivity, web research, or interaction with browser-based interfaces.
What can you do with it?
- Searching for information via web search engines
- Extracting specific content from known URLs
- Navigating and interacting with websites requiring login states
- Scraping dynamic or anti-scraping protected content from social media and web platforms
- Performing GUI-based browser tasks like clicking, scrolling, and form submission
- Retrieving local browser history or bookmarks for context
- Sampling video content via browser frame capture
How good is this skill?
Quality score: 5/10. The skill documentation is comprehensive, providing clear operational logic, specific API endpoints, and safety guidelines for browser automation.
What does the skill file contain?
# web-access Skill
## 前置检查
在开始联网操作前，先检查 CDP 模式可用性：
```bash
node "${CLAUDE_SKILL_DIR}/scripts/check-deps.mjs"
```
**Node.js 22+** 必需（使用原生 WebSocket）。
按脚本输出处理：
- `exit 0` → 继续
- `exit 2` → 需询问用户偏好，写入 `${CLAUDE_SKILL_DIR}/config.env` 的 `WEB_ACCESS_BROWSER`
- `exit 1` → 按 stdout 错误信息处理。若提示包含「Agent 处理顺序」，按其步骤执行（如先用系统命令打开浏览器后重跑），自动可解则不打扰用户；仍失败再向用户求助
支持参数 `--browser <chrome|edge>` 表达本次临时覆盖（不写 config.env）。
切换浏览器时，proxy 是长驻进程，需先 `pkill -f cdp-proxy.mjs` 再重跑 check-deps。
检查通过后并必须在回复中向用户直接展示以下须知，再启动 CDP Proxy 执行操作：
```
温馨提示：部分站点对浏览器自动化操作检测严格，存在账号封禁风险。已内置防护措施但无法完全避免，Agent 继续操作即视为接受。
```
## 浏览哲学
...Frequently asked questions
How does the skill handle login-protected websites?
The skill uses CDP to connect to the user's existing browser instance, which naturally carries the user's login state. If content is inaccessible, the agent prompts the user to log in manually.
Can I use this skill to search my own browser history?
Yes, the skill includes a script to search local Chrome or Edge bookmarks and history using keywords, time filters, and sorting options.
What is the difference between WebFetch and CDP mode?
WebFetch is for static content extraction and does not handle login states. CDP mode provides a full browser environment for interactive tasks, login-required pages, and dynamic content.
How does the skill minimize impact on my browser?
It creates and manages its own background tabs for tasks and closes them upon completion, leaving the user's existing tabs untouched.
Related skills
agent-browser
506.7KAI agents and developers requiring programmatic web interaction, exploratory testing, or automation of Electron desktop applications
The agent-browser CLI provides browser automation for AI agents using Chrome or Chromium via CDP. It supports page navigation, form interaction, data extraction, and testing. The tool utilizes accessibility-tree snapshots and element references for interaction.
extract-design-system
124.5KDevelopers and designers who want to initialize design systems based on existing public websites
This skill extracts design primitives from public websites and generates starter token files for projects.
browser-act
74.5KAI agents and developers requiring browser-based automation, web scraping, or interaction with JavaScript-heavy websites
Browser-act provides a command-line interface for AI agents to perform browser automation, including navigation, data extraction, form interaction, and session management.
anti-detect-browser
27.3KDevelopers and AI agents requiring isolated browser environments, multi-account management, or automated web interactions that bypass anti-bot detection
The anti-detect-browser SDK provides tools to launch Chromium instances with unique, persistent device fingerprints for multi-account management, web scraping, and AI agent automation. It integrates with Playwright to isolate browser sessions, cookies, and storage, and offers an MCP server mode for AI agent control.