专业的学术情报提取工具。支持 Nature/Science/Cell 等全球主流期刊，自动化抓取过去 N 天内新增的 Article 或 Review，并深度提取 PMID 与 Abstract 全文，为 AI 科普总结提供核心数据源。

This skill appears to do what it says: it scrapes PubMed search results and article pages for PMIDs, titles, and abstracts and saves them as JSON in ~/Documents/Journal_Intel/. Before installing or running, consider: (1) The description's phrase “deep access / 全文” may imply retrieving paywalled full text, but this script only fetches PubMed pages/abstracts — if you expect full articles you'll need different code or credentials. (2) Respect PubMed/NLM terms of use and robots.txt; if you plan frequent runs increase the delay or use official APIs (e.g., Entrez E-utilities) to avoid throttling. (3) The SKILL.md assumes a virtualenv (venv/bin/python3) but no install step is provided; you should create a virtualenv and pip install -r requirements.txt before running. (4) The script writes files to your Documents folder — confirm you’re comfortable with that path and disk use, or modify the save location. (5) No credentials or external endpoints are requested, and the code does not exfiltrate data beyond contacting PubMed. If you need stronger guarantees, inspect/modify the script to use Entrez APIs (with an API key) and add explicit error handling and rate limiting.

Type: OpenClaw Skill Name: journal-intel-extractor Version: 1.0.0 The skill is a legitimate academic scraping tool designed to extract article abstracts from PubMed. The code in main.py uses standard libraries (requests, BeautifulSoup) to fetch data from a specific domain (pubmed.ncbi.nlm.nih.gov), implements basic rate limiting, and saves the results to a local directory (~/Documents/Journal_Intel) as described in SKILL.md. There are no signs of data exfiltration, malicious execution, or prompt injection.