使用方法
- 复制安装命令，在自己的智能体环境安装前先检查源码包。
- 从原始文档核对所需工具、凭据和支持的宿主。
- 说明输入材料与期望交付物，并对照来源检查输出。
npx skills add https://github.com/crawlora-org/crawlora-skills --skill news-briefing-research开始前需了解
- 安装不包含模型订阅、API 额度或访问文件的权限。
- 仓库星数和安装量反映活跃情况，不等于独立质量或安全测试。
仓库原始内容
按源仓库内容呈现，保留标题、案例、代码、表格、链接以及原文引用的演示图片。
News briefing research
Gather recent public news from multiple publishers and GDELT, then return a dated briefing with source links and clearly separated reporting from inference.
When to use this skill
- Summarize recent coverage of a topic, company, person, or event.
- Compare how multiple outlets frame the same story.
- Track coverage volume, sentiment, entities, or timelines in a bounded window.
- Find original article pages and live-story updates.
Research workflow
- Define topic terms, geography, language, date window, source set, and update
cutoff before searching. A headline query is not a complete news corpus.
- Search GDELT and publisher-specific endpoints separately. Deduplicate by
canonical URL, title, publisher, and publication timestamp while retaining syndicated versions when they add distinct reporting.
- Prefer the original article and record publication/update times, author,
outlet, URL, and whether the item is an opinion, live update, or report.
- Compare claims across sources and label unverified allegations, corrections,
missing context, and model-generated summaries. Do not infer causality from co-occurrence or coverage volume.
Examples
sh
scripts/crawlora.sh /gdelt/search query="semiconductor exports" | jq '.'
scripts/crawlora.sh /bbc/search query="semiconductor exports" | jq '.'
scripts/crawlora.sh /cnn/search query="semiconductor exports" | jq '.'
scripts/crawlora.sh /guardian/search query="semiconductor exports" | jq '.'
scripts/crawlora.sh /yahoo-news/search query="semiconductor exports" | jq '.'Notes and limits
- Public articles and feeds only; no paywall bypass, private account access,
or publication actions occur.
- Search indexes differ by source, region, and freshness. State the collection
window and coverage limits rather than calling the result exhaustive.
- Quote sparingly, link the original page, and preserve corrections or updates.
- Treat sentiment and entity extraction as source-derived signals, not facts.
See `reference/endpoints.md` for the complete generated endpoint list.
