← 返回 Skills 市场
219
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install persistent-browser
功能描述
用 Playwright 持久化上下文（main-identity）抓取需要登录态的网站（YouTube、GitHub、HuggingFace、Reddit、Kaggle、X/Twitter）。当用户要求外网搜索或指定这些网站时自动触发。
安全使用建议
This skill instructs the agent to open and modify a hard-coded local browser profile (including deleting a lock file) to scrape logged-in content. That profile can contain cookies, sessions, and other sensitive data. Before installing: (1) do not install unless you fully trust the skill author and understand why it needs your browser profile; (2) prefer a design that uses explicit API tokens or a dedicated, sandboxed browser profile rather than your main identity; (3) require the skill to declare dependencies (Playwright, browser) and the config path it will use, and change the path to a profile you control; (4) avoid running with --no-sandbox and headful flags on untrusted code; (5) if you test it, run inside an isolated VM/container with a throwaway profile and monitor file access; (6) consider disabling autonomous invocation so the skill runs only with explicit user approval.
功能分析
Type: OpenClaw Skill
Name: persistent-browser
Version: 1.0.0
The skill is designed to access a persistent browser profile located at a hardcoded path (/home/kncao/.openclaw/browser-profiles/main-identity), which contains sensitive session data and login credentials for platforms like GitHub, X/Twitter, and YouTube. It explicitly disables security features (--no-sandbox) and requires non-headless mode to bypass bot detection. While the stated purpose is scraping, the access to a persistent 'main-identity' profile poses a high risk of session hijacking or unauthorized access to private user data if the agent is misdirected. (SKILL.md)
能力评估
Purpose & Capability
The described purpose (scraping sites that require login via a persistent Playwright context) is coherent with using launch_persistent_context. However, the SKILL.md hardcodes a specific user_data_dir (/home/kncao/.openclaw/browser-profiles/main-identity) and instructs manipulating its files. The skill metadata declares no required config paths or credentials, so the instructions demand access beyond what was declared and beyond a typical scraper's minimal needs.
Instruction Scope
The runtime instructions explicitly tell the agent to read/write a local browser profile and to rm -f the SingletonLock before each run. That is file-system access and modification of another profile on disk (potentially containing cookies, sessions, credentials). The instructions also require running Playwright headful with args (including --no-sandbox and anti-detection flags), which expands runtime privileges and evasion tactics. These actions go beyond simple page fetching and implicate sensitive local data and destructive operations.
Install Mechanism
This is instruction-only (no install spec), which limits installation-time risk. However, the SKILL.md assumes Playwright (and a browser) are available but does not declare required binaries or packages. The missing dependency declarations are an incoherence: the skill will fail or behave unpredictably unless Playwright and appropriate browsers are present.
Credentials
No environment variables or credentials are requested, yet the skill asks to use a persistent browser profile that likely contains cookies, tokens, and session state. Access to that profile is disproportionate and privacy-sensitive. The skill gives no guidance for using a dedicated/sandboxed profile or requesting explicit user consent for accessing such data.
Persistence & Privilege
The skill is not marked always:true, but disable-model-invocation is false (normal), meaning the agent could autonomously invoke this skill when triggered by web-search intents. Autonomous invocation combined with the ability to read/modify a local logged-in browser profile increases the blast radius and privacy risk if invoked without explicit user confirmation.
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install persistent-browser - 安装完成后，直接呼叫该 Skill 的名称或使用
/persistent-browser触发 - 根据 Skill 的参数说明提供必要输入，即可获得结构化输出
版本历史
v1.0.0
Initial release: Enables persistent browser context scraping for login-required websites.
- Uses Playwright with a persistent user profile for sites needing authentication (YouTube, GitHub, HuggingFace, Reddit, Kaggle, X/Twitter).
- Automatically triggers when accessing specified sites or when "external web search" is requested.
- Ensures headless mode is disabled for proper page rendering, especially on X/Twitter.
- Cleans up browser profile lock files before scraping to avoid errors.
- Customizes waiting times per site for reliable JavaScript/SPAs rendering.
- Extracts plain text content, avoiding screenshots.
元数据
常见问题
Persistent Browser Scraper 是什么？
用 Playwright 持久化上下文（main-identity）抓取需要登录态的网站（YouTube、GitHub、HuggingFace、Reddit、Kaggle、X/Twitter）。当用户要求外网搜索或指定这些网站时自动触发。 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 219 次。
如何安装 Persistent Browser Scraper？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install persistent-browser」即可一键安装，无需额外配置。
Persistent Browser Scraper 是免费的吗？
是的，Persistent Browser Scraper 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Persistent Browser Scraper 支持哪些平台？
Persistent Browser Scraper 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。
谁开发了 Persistent Browser Scraper？
由 ckncg（@ckncg）开发并维护，当前版本 v1.0.0。
推荐 Skills