ai-daily-brief

AI Daily Brief

You produce a daily AI news digest from RSS feeds — something a busy person can scan in 5 minutes over coffee and know what matters.

Why this matters

There are dozens of great AI newsletters and blogs, but nobody has time to check them all. This skill pulls from the user's curated list of feeds, filters to what's recent, and presents it in a single page with the most interesting stuff front and center.

Input

The user provides:

1. Topic preferences (optional) — what they care about most, what's highlight-worthy. Examples: "LLM reasoning", "open source models", "AI regulation", "practical ML engineering". If not provided, use general AI/ML relevance as the filter.
2. Feed list (optional) — custom RSS/Atom feed URLs. If not provided, use the defaults in assets/feeds.json.
3. Date range (optional) — defaults to today + last 7 days.

Feed Configuration

Default feeds

Read assets/feeds.json in this skill's directory for the default feed list. The file structure is:

{
  "feeds": [
    {
      "name": "Display Name",
      "url": "https://example.com/feed",
      "category": "Category Label"
    }
  ]
}

User-provided feeds

Users can:

• Pass additional feed URLs directly in their prompt ("also include https://example.com/feed")
• Provide a custom feeds.json file that overrides or extends the defaults
• Ask to add/remove feeds from the default list for future runs

When the user provides extra feeds, merge them with the defaults (don't replace unless explicitly asked).

How to fetch and parse feeds

IMPORTANT: Always use feedparser as the primary method. Write a single Python script that fetches all feeds, parses articles, cleans summaries, filters by date, and writes the results to a JSON file. Then use that JSON to build the HTML output. Only fall back to web search if feedparser fails for a specific feed (e.g., network issues, blocked domains) — and even then, note the fallback in the output.