soushen

高性能 Bing 搜索引擎 Skill - "搜神猎手" 使用 Playwright 底层 API 进行深度网页搜索和元素提取 功能： 1. Bing 搜索执行 - 返回结构化搜索结果（标题、链接、摘要、来源） 2. 深度页面分析 - 提取页面的所有关键元素（链接、表单、按钮、脚本、元数据） 触发条件： - 用户...

This skill implements exactly the kind of browser automation you would expect for a scraper, but be cautious: the code collects page cookies and session state and uses anti-detection flags that can make scraping stealthier. Before installing or running it, consider the following: - Do not run this on a machine where you are logged into sensitive accounts (Google, Microsoft, banking, etc.). The script collects cookies which could include authentication tokens. - Review the rest of scripts/bing_search.py (the file listing was truncated) to confirm there is no code that transmits extracted cookies or other data to any external server. If the code prints JSON to stdout, be careful where you direct that output. - Run the skill in an isolated environment (container or VM) and limit network access if you want to prevent accidental exfiltration. - If you only need link/form/script metadata, consider editing the code to remove cookies from the PageElements output (delete or sanitize cookies before returning/printing). - The anti-detection flags (e.g. --disable-web-security, AutomationControlled evasion) are legitimate for scraping but increase privacy/ethical concerns; ensure usage complies with target site terms of service. If you can provide the remainder of bing_search.py (the truncated portion), I can re-evaluate for any explicit exfiltration or remote endpoints and raise the confidence of this assessment.

Type: OpenClaw Skill Name: soushen Version: 1.0.0 The skill provides Bing search and deep web scraping capabilities using Playwright, but it exhibits high-risk behaviors including the extraction of browser cookies and the disabling of critical browser security features like '--disable-web-security' and '--no-sandbox' in `scripts/bing_search.py`. While these are common in advanced scraping tools, returning session cookies to the AI agent and bypassing security sandboxes creates a significant risk for session hijacking or data leakage if the agent is directed to sensitive or authenticated URLs.