← 返回 Skills 市场

Ffagen Minimax Vision Scraper

Name: Ffagen Minimax Vision Scraper
Author: ffagen

作者 ffagen · GitHub ↗ · v1.0.0 · MIT-0

cross-platform ⚠ suspicious

总下载

当前安装

版本数

在 OpenClaw 中安装

/install ffagen-minimax-vision-scraper

功能描述

Playwright截图 + MiniMax图像理解的高级网页抓取skill。绕过反爬虫，直接用AI理解截图内容。

使用说明 (SKILL.md)

MiniMax Vision Scraper

Playwright 截图 + MiniMax 图像理解，绕过反爬虫，直接用 AI 提取页面内容。

核心优势

对比项	传统抓取	Vision Scraper
反爬虫	❌ 易被屏蔽	✅ 截图绕过检测
JS渲染	⚠️ 复杂	✅ 截图即完整
内容理解	❌ 需解析HTML	✅ AI直接理解
动态内容	❌ 难抓	✅ 截图即所见

工作流程

URL → Playwright截图(Chrome) → MiniMax VLM图像理解 → AI分析结果

安装

cd ~/.openclaw/workspace/skills/ffagen__minimax-vision-scraper
npm install playwright
# 无需安装 chromium，直接使用系统已装的 Google Chrome

使用方式

直接调用

node scripts/screenshot.js \x3CURL> [prompt]

# 示例
node scripts/screenshot.js "https://news.sina.com.cn" "提取今日头条新闻"
node scripts/screenshot.js "https://finance.sina.com.cn" "提取股票行情数据"

环境变量

变量	默认值	说明
WAIT_TIME	5000	等待时间(ms)
HEADLESS	true	是否无头模式
SCREENSHOT_PATH	/tmp/screenshot-*.png	截图保存路径
MINIMAX_API_KEY	(内置)	MiniMax API Key
MINIMAX_API_HOST	api.minimaxi.com	API地址

在OpenClaw中调用

用 ffagen__minimax-vision-scraper 抓取 https://example.com，分析页面内容

反爬虫技术（致敬 playwright-scraper-skill）

✅ 隐藏 navigator.webdriver
✅ 真实 iPhone User-Agent
✅ 移动端视口 (375x812)
✅ 隐藏 Chrome 自动化特征
✅ 模拟 permissions.query
✅ Cloudflare 检测 + 额外等待

示例

抓取财经新闻：

node scripts/screenshot.js "https://finance.sina.com.cn" "提取所有财经新闻标题和摘要"

抓取股票数据：

node scripts/screenshot.js "https://stock.finance.sina.com.cn" "提取大盘指数和涨跌数据"

抓取商品信息：

node scripts/screenshot.js "https://item.jd.com/100000.html" "提取商品名称、价格、评价"

技术说明

截图引擎：使用系统已装的 Google Chrome（无需额外下载）
图像理解：MiniMax VLM API (/v1/coding_plan/vlm)
反爬策略：Playwright Stealth 模式，隐藏自动化特征

安全使用建议

This skill will take screenshots of target pages and upload them to a third-party Vision API. Before installing: 1) Do not rely on the embedded API key — remove it or supply your own MINIMAX_API_KEY; the built-in key could be abused or revoked and means your screenshots would be processed under someone else's account. 2) Assume any screenshot may contain sensitive data (logins, tokens, personal data); avoid running it against pages where you are authenticated or that contain secrets. 3) Update the macOS-specific Chrome path if you are not on macOS or let Playwright manage its browser. 4) If you plan to use this, review the code, replace/remove the embedded key, point MINIMAX_API_HOST to a trusted endpoint, and run in an isolated sandbox until you trust the behavior. Because of the embedded key and undeclared env requirements, treat this skill as suspicious and proceed only after remediation.

功能分析

Type: OpenClaw Skill Name: ffagen-minimax-vision-scraper Version: 1.0.0 The skill contains a hardcoded MiniMax API key in `scripts/screenshot.js`, which is a significant security vulnerability (credential leak). It also uses a hardcoded absolute path for the Google Chrome executable on macOS, which is brittle and environment-specific. While the tool performs its stated function of web scraping via screenshots, the ability to capture any URL and send it to an external endpoint (api.minimaxi.com) presents a risk of local file disclosure if the agent is directed to access sensitive local files via `file://` protocols.

能力评估

⚠ Purpose & Capability

The name/description (Playwright screenshots + MiniMax VLM) matches the code: the script screenshots a page and posts it to a MiniMax API. However registry metadata claims no required env vars or binaries while SKILL.md and scripts rely on system Google Chrome and an API key (MINIMAX_API_KEY). The script hardcodes an API key and uses a macOS-specific Chrome path, which is inconsistent with the 'no requirements' claim.

⚠ Instruction Scope

SKILL.md and scripts limit runtime actions to browsing, screenshotting, and POSTing the image to api.minimaxi.com. That is consistent with the stated purpose, but screenshots can contain sensitive data and the skill will upload full-page screenshots to a third-party API. The SKILL.md doesn't warn about this risk nor require the user to supply their own API key (the code falls back to an embedded key).

✓ Install Mechanism

No install spec in registry, but package.json/package-lock.json are present and only depend on playwright from npm — a standard dependency. There are no downloads from odd URLs or archives. The install risk is moderate and expected for a Playwright-based tool.

⚠ Credentials

The code uses sensitive environment variables (MINIMAX_API_KEY, MINIMAX_API_HOST) but the registry lists none. Worse, the script contains a hardcoded API key string that will be used if MINIMAX_API_KEY is not set. Requiring/using an API key to call an external service is expected, but embedding a fallback secret (and not declaring it) is inappropriate and could represent abuse of someone else's key or leakage of a secret.

✓ Persistence & Privilege

The skill does not request persistent/always-on privileges, does not change other skills' configs, and does not require elevated system privileges. It launches a local browser instance only.

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 ffagen-minimax-vision-scraper
安装完成后，直接呼叫该 Skill 的名称或使用 /ffagen-minimax-vision-scraper 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 of ffagen__minimax-vision-scraper. - Combines Playwright webpage screenshots with MiniMax visual language model for advanced scraping. - Bypasses anti-bot mechanisms by using screenshots and AI image understanding instead of parsing HTML. - Supports command-line usage and flexible prompts for detail extraction. - Includes built-in stealth technologies adapted from playwright-scraper-skill to evade detection. - Provides preset environment variable options and usage instructions for OpenClaw integration.

元数据

Slug ffagen-minimax-vision-scraper

版本 1.0.0

许可证 MIT-0

累计安装 0

当前安装数 0

历史版本数 1

常见问题