电池行业国际市场动态监控系统。监控中国、美国、印度、俄罗斯、韩国的政策补贴、法规、安全事件和行业投资动态,优先使用官方/行业 RSS 与 Google News RSS 获取稳定新闻源,再抓取详情、分类利好利空并生成偏证券研究风格的周报。支持非中文新闻的原文摘要、中文翻译、以及每条新闻的单条行业解读。用户要求"运...

This skill appears to implement the described news-monitoring and report-generation workflow, but review the following before installing: - Environment variables: the code reads GNEWS_API_KEY and NEWS_API_KEY from the environment though the skill metadata does not declare them. Only set those keys if you trust the API provider and you understand rate limits and data sharing. - Local LLM call: scripts/analyze_sentiment.py will POST news content to http://localhost:11434 by default (calls a local Ollama/vLLM-like API). If you do not run a local LLM, the code falls back to a template, but if you do run one, be aware that parsed article text and titles will be sent to that local service. Ensure the local service is trusted and isolated if the content is sensitive. - Network activity and data written: the scripts fetch arbitrary news URLs discovered via RSS and web search and write parsed outputs to data/*.json/.md/.docx. Run the skill in an isolated environment (dedicated VM/container) if you are concerned about network exposure or want to limit what the system fetches. - Dependency installation: there is no automated install spec; install requirements.txt in a virtual environment and audit third-party packages if you require higher assurance. - If you need higher assurance: ask the maintainer for explicit documentation of optional API keys and local-LLM behavior, or request a version that documents all environment variables and network endpoints. If you can't obtain that, treat API keys and enabling a local LLM cautiously.

Type: OpenClaw Skill Name: battery-market-watch Version: 1.1.0 The bundle is a legitimate news monitoring and reporting tool focused on the international battery market. It implements a multi-step workflow including news searching via RSS and Google News (scripts/search_news.py), content extraction (scripts/fetch_detail.py), and sentiment analysis using a local LLM or keyword templates (scripts/analyze_sentiment.py). The final reports are generated in Word and Markdown formats and saved to a local data directory, with a convenience feature in scripts/generate_report.py that attempts to copy the results to the user's desktop. No evidence of data exfiltration, malicious prompt injection, or unauthorized persistence was found; all network and file operations align with the stated purpose of the skill.