Morning Briefing AI News

AI News Skill

Fetch and summarize AI industry news from curated RSS feeds. Designed to catch major model releases (GPT, Claude, Gemini, Grok, GLM, etc.) and important AI news.

Usage

Invoke with /ai-news or ask for "AI news", "morning briefing", "what's new in AI".

Feeds

This skill fetches from 15 curated sources:

Aggregators (daily/weekly digests):

Lab Blogs (via Olshansk/rss-feeds):

AI Coding Tools (via Olshansk/rss-feeds):

Instructions

Step 0: Idempotency Check (Save Mode Only)

If --save was requested AND --regenerate was NOT specified: 1. Determine today's briefing path (same logic as Step 7.5) 2. Check if ai-news-YYYY-MM-DD.md already exists at that path 3. If it exists → read and return its contents to the user with note: (Cached — run with --regenerate to refresh) 4. If it does not exist → proceed to Step 1

If --save was NOT requested, skip this step entirely.

Step 1: Load Feed Configuration

Read the feed list from feeds.json in this skill directory.

Step 2: Fetch Feeds (Parallel)

For each feed in feeds.aggregators, feeds.labs, and feeds.tools, use web_fetch to retrieve the RSS content:

web_fetch(url: "", extractMode: "text")

Fetch priority 1 feeds first, then priority 2.

Step 3: Parse RSS Items

RSS feeds are XML. Extract items using these patterns:

  
  ...
  ...
  ...

For Atom feeds:

  
  
  ...
  
...

Step 4: Filter by Date

Compute cutoff = current datetime - exactly 86400 seconds (24 hours). Discard any item where pubDate or published is earlier than cutoff. Do not round or approximate — if an item is 25 hours old, it is excluded.

Weekly mode (user says "week" or "last 7 days"): use 604800 seconds instead.

Step 5: Deduplicate

Same story may appear in multiple feeds. Dedupe by: 1. Exact title match 2. Similar title (>80% overlap) 3. Same link URL

Keep the version from the highest-priority source.

Step 6: Categorize

Assign each item to a category:

| Category | Keywords | |----------|----------| | Model Releases | "release", "launch", "announce", model names (GPT, Claude, Gemini, Grok, GLM, Llama) | | Research | "paper", "research", "study", "benchmark" | | Industry | "funding", "acquisition", "hire", "layoff", "IPO" | | Product | "feature", "update", "API", "pricing" | | Opinion | "think", "believe", "future", "prediction" |

Step 7: Format Output

# AI News Briefing
Date: [today's date]
Sources: [count] feeds checked
Period: Last 24 hours
Model Releases
1. [Title] — [1 sentence summary]. [Source]
2. ...
Research
1. [Title] — [1 sentence summary]. [Source]
...
Industry News
1. [Title] — [1 sentence summary]. [Source]
...
Coverage Gap: Chinese AI labs (Zhipu, DeepSeek, Baidu, Alibaba) don't publish RSS. Check manually for major releases.

Step 7.5: Save to File (Optional)

Only activate this step if the user explicitly requested it — via --save, save to file, or similar phrasing. Do NOT save by default.

If saving is requested: 1. Determine the save directory: - If the user specified a path in their request, use that path exactly - Otherwise default to ~/ai-news/ where date is today's date in YYYY-MM-DD format 2. Write the full formatted briefing to

   # AI News Links — YYYY-MM-DD
   ## Model Releases
   - Title
   ## Research
   - Title
   ## Industry News
   - Title
   

   Saved to /ai-news-YYYY-MM-DD.md
   Links  → /ai-news-YYYY-MM-DD-links.md
   

Invoke with /ai-news or ask for "AI news", "morning briefing", "what's new in AI".