Stream

Stream — News & Market Intelligence Skill

What This Does

Fetches and summarizes news from configured RSS feeds, financial sources, X/Twitter influencers, and government press releases. Outputs a categorized, concise briefing.

Feed Configuration

Feeds are stored in ~/.katana/memory/skills/productivity/news-stream/feeds.json. This file contains all RSS/JSON feed URLs organized by category. The agent reads this file, curls each URL, parses the response, and summarizes.

How to Fetch Feeds

JSON Feeds (rss.app format)

curl -s "https://rss.app/feeds/v1.1/FEED_ID.json"

Response structure:

{
  "items": [
    {
      "title": "Headline",
      "url": "https://...",
      "date_published": "2026-03-04T12:00:00Z",
      "content_text": "Article summary or full text"
    }
  ]
}

Extract: items[].title, items[].content_text, items[].date_published, items[].url

XML Feeds (standard RSS)

curl -s "https://feeds.content.dowjones.io/public/rss/mw_bulletins"

Response structure:

<rss>
  <channel>
    <item>
      <title>Headline</title>
      <link>https://...</link>
      <pubDate>Mon, 04 Mar 2026 12:00:00 GMT</pubDate>
      <description>Summary text</description>
    </item>
  </channel>
</rss>

Extract from each <item>: <title>, <description>, <pubDate>, <link>

HTML Feeds (Federal Reserve)

curl -s "https://www.federalreserve.gov/feeds/press_all.xml"

This is actually XML/Atom. Parse <entry> elements for <title>, <link>, <updated>.

Processing Rules

1. Fetch all feeds from feeds.json
2. Filter to last 24 hours (or since the timestamp in the last stream check)
3. Deduplicate — if multiple feeds have the same story, keep the original source
4. Categorize based on the category field in feeds.json:
   • markets → 📊 Markets section
   • ai-tech → 🤖 AI & Tech section
   • economy → 💰 Economy section
   • government → 🏛️ Fed & Government section
   • x-feed → 🐦 X Chatter section
5. Summarize each item in 1-2 sentences. Strip HTML tags, marketing fluff, and repetitive disclaimers.
6. Rank by importance within each category. Market-moving events first.
7. Flag urgent items at the top if any are time-sensitive or market-moving.

Integration

Morning Routine (Sensei)

When Sensei runs the morning check-in, it should call /stream and include the top 3-5 headlines in the briefing. Check ~/.katana/memory/core/user.md for the user's interests, active investments, or tracked topics — prioritize news that matches.

Memory

After each stream pull, optionally log a one-line summary to ~/.katana/memory/work.md:

## 2026-03-04 — Stream
Checked 15 feeds, 23 new items. Key: Fed held rates, NVDA earnings beat, new AI regulation proposed in EU.

Feed Management Commands

• "Add a feed" → Ask for URL, name, category. Append to feeds.json.
• "Remove [name]" → Remove matching entry from feeds.json.
• "List feeds" → Display all configured sources with categories.
• "Only show [category]" → Filter output to one section.

Files

• feeds.json — Feed URL configuration (edit this to add/remove sources)
• SKILL.md — This file (agent instructions)