News Digest

Generate a professional, neutral news briefing using RSS feeds as the primary data source. RSS feeds return structured headlines and summaries, which is far more token-efficient than web search. Only fall back to WebSearch when RSS coverage is insufficient.

Workflow

1. Determine scope

Parse the user's request to figure out:

• Timeframe: "today" (last 24h) or "this week" (last 7 days). Default to today if unclear.
• Language: If the user writes in Chinese or asks for Chinese news, produce the digest in Chinese. Otherwise default to English. The user can also explicitly request a language.
• Categories: Use the user's requested categories if specified. Otherwise use the defaults below.

Default categories:

• International (geopolitics, diplomacy, conflicts, major world events)
• AI & Machine Learning (models, research, industry moves, regulation)
• Technology (products, startups, platforms, cybersecurity)
• Economics & Finance (markets, central banks, trade, macro indicators)

The user can add, remove, or replace categories — be flexible.

2. Fetch RSS feeds

Use WebFetch to fetch RSS feeds in parallel. Pick 2-3 feeds per category from the list below based on the user's language preference. RSS feeds return XML with <item> elements containing <title>, <description>, <link>, and <pubDate> — extract these to build your story list.

RSS Feed Directory:

International (English)

• BBC World: http://feeds.bbci.co.uk/news/world/rss.xml
• Al Jazeera: https://www.aljazeera.com/xml/rss/all.xml
• NPR World: https://feeds.npr.org/1004/rss.xml

AI & Technology (English)

• TechCrunch AI: https://techcrunch.com/category/artificial-intelligence/feed/
• Ars Technica AI: https://arstechnica.com/ai/feed/
• The Verge: https://www.theverge.com/rss/index.xml
• Wired AI: https://www.wired.com/feed/tag/ai/latest/rss
• MIT Tech Review: https://www.technologyreview.com/feed/

Economics & Finance (English)

• CNBC Top News: https://search.cnbc.com/rs/search/combinedcms/view.xml?partnerId=wrss01&id=100003114
• CNBC Economy: https://search.cnbc.com/rs/search/combinedcms/view.xml?partnerId=wrss01&id=20910258
• MarketWatch: https://www.marketwatch.com/rss/topstories

Chinese Sources (中文)

• 36Kr 快讯: https://36kr.com/feed
• 澎湃新闻 (via RSSHub): https://rsshub.app/thepaper/channel/25950
• 36Kr 快讯 (via RSSHub): https://rsshub.app/36kr/newsflashes
• China Daily: https://www.chinadaily.com.cn/rss/china_rss.xml
• China Daily Business: https://www.chinadaily.com.cn/rss/business_rss.xml

Feed selection strategy:

• For English digests: use the English feeds for each requested category
• For Chinese digests: use Chinese feeds + English feeds (translate English headlines to Chinese)
• Fetch 2-3 feeds per category in parallel — this gives enough coverage without waste

3. Filter and supplement

After parsing RSS items:

• Filter by date: Only keep items whose <pubDate> falls within the requested timeframe.
• Deduplicate: If the same story appears in multiple feeds, keep only one.
• Check coverage: If any category has fewer than 3 stories after filtering, use WebSearch as a fallback for that category only. This keeps token usage low — most of the time RSS provides enough stories.

4. Compile the digest

Write the digest in plain text using this structure:

NEWS DIGEST — [Date or Date Range]
============================================

[CATEGORY NAME]
--------------------------------------------
- [Headline]: [1-2 sentence summary]. ([Source])
- [Headline]: [1-2 sentence summary]. ([Source])
...

[NEXT CATEGORY]
--------------------------------------------
- ...

For Chinese output:

新闻简报 — [日期或日期范围]
============================================

[分类名称]
--------------------------------------------
- [标题]：[1-2句摘要]。（[来源]）
...

Guidelines:

• Aim for 5-8 stories per category. More is better than fewer, but don't pad with trivial items.
• Each story gets 1-2 sentences max. Lead with the most newsworthy fact.
• Write in neutral, factual tone. No editorializing.
• Always attribute the source in parentheses.
• Order stories within each category by significance, most important first.
• If a story spans multiple categories, place it in the most relevant one only.
• Include the date or date range in the header.

5. Present to the user

Output the digest directly. No preamble — just start with the digest itself.

If any category returned very few results (fewer than 3 stories), note this briefly at the end: "Note: Limited results found for [category] in the specified timeframe."