检测网页是否含有针对 AI 的 GEO 投毒虚假信息，提供可信度评分与详细风险分析报告。

What to consider before installing: - Functionality gap: The code in src/index.ts/dist/index.js does not fetch the target URL; it analyzes a hard-coded sampleContent. If you expect real webpage analysis, this skill as shipped will not do that — it is effectively a demo. You would need to (a) integrate a safe page-fetcher (for example the platform's sanctioned web_fetch API) or (b) modify the handler to accept fetched page content before trusting results. - Low security surface: There are no external network calls, no required secrets, and no install script. That means the immediate security risk is low, but also explains why it cannot analyze real pages. - Code quality issues: The implementation includes some problematic constructs (e.g., regexes that look like incorrect alternation patterns, and object keys with spaces such as 'GEO 检测' that may cause runtime/compilation errors). These bugs may break the skill in real use and lead to incorrect scores. - If you want to use it in production: review and fix the page-fetch integration to use the platform-approved web fetch mechanism, run the included tests locally, fix the regexes and object property naming, and re-audit before granting any network privileges. Test the skill in an isolated environment first. - If the author provides an update or documentation showing safe, explicit web-fetch integration (and the code is corrected), that would reduce concerns. As-is, treat this as a demonstration/example rather than a drop-in tool for real webpage analysis.

Type: OpenClaw Skill Name: geo-shield Version: 1.0.0 The GEO-Shield skill bundle is designed to detect Generative Engine Optimization (GEO) and misinformation in web content. Analysis of the source code in `src/index.ts` and `dist/index.js` shows it uses regex-based pattern matching, domain reputation scoring, and content quality heuristics to generate a trust report. While the current implementation uses a hardcoded 'sampleContent' placeholder rather than performing an actual network fetch for the provided URL, there are no indicators of malicious behavior, data exfiltration, or prompt injection attacks. The code is transparent and aligns with its stated purpose.