Daily News Curator

Curates a personalized news briefing from RSS feeds, email newsletters, Hacker News, Reddit, and web sources tailored to your specific interests and industry. Unlike generic news apps, this skill learns your preferences, filters noise, and presents a focused reading list with AI-generated summaries so you stay informed in 5 minutes instead of 50.

Supported Platforms & Integrations

Platform	Setup Method	Auth Type	Notes
RSS/Atom Feeds	Direct URL subscription	None	Any valid RSS or Atom feed URL
Hacker News	HN Algolia API	None (public)	Top stories, Show HN, Ask HN filterable
Reddit	Reddit JSON API	Optional OAuth	Subreddit filtering, score thresholds
Twitter/X Lists	X API v2	OAuth 2.0	Curate from specific lists or accounts

When to Use This Skill

• Use this when you follow 10+ news sources and want one consolidated feed
• Use this when you want AI-summarized articles instead of reading full pieces
• Use this when you need industry-specific news filtered from general noise
• Consider alternatives when you need real-time breaking news alerts (this is a digest)

Quick Start

Copy
# Minimal configuration - news-curator.yml
skill: daily-news-curator
interests:
  primary: ["artificial-intelligence", "web-development", "startups"]
  secondary: ["cybersecurity", "climate-tech"]
  excluded: ["crypto", "celebrity-news"]
sources:
  rss:
    - https://techcrunch.com/feed/
    - https://feeds.arstechnica.com/arstechnica/technology-lab
    - https://www.theverge.com/rss/index.xml
  hackernews:
    enabled: true
    min_score: 100
  reddit:

Copy
claude /daily-news-curator

Expected Output

NEWS BRIEFING - March 15, 2026
Curated from 23 sources | 142 articles scanned | 12 selected

TOP STORIES (3):

1. "Anthropic Releases Claude Agent SDK for Enterprise" - TechCrunch
   Summary: New SDK allows enterprises to build autonomous agents with
   built-in safety constraints and audit logging. Supports Python and
   TypeScript. Available today on npm and PyPI.
   Relevance: AI + web development | Read time: 6 min
   Link: https://techcrunch.com/2026/03/15/...

2. "React 20 Introduces Server-First Architecture" - Verge
   Summary: Major React update rethinks component model with server
...

Advanced Configuration

Platform-Specific Setup

RSS Feeds with Custom Parsing

Copy
sources:
  rss:
    - url: https://techcrunch.com/feed/
      category: "tech"
      max_age_hours: 24
    - url: https://feeds.bloomberg.com/markets/news.rss
      category: "finance"
      keywords_filter: ["tech", "AI", "startup"]

Full Options Reference

Parameter	Type	Default	Description
interests.primary	array	[]	Main topics to prioritize (weighted 3x)
interests.secondary	array	[]	Supporting topics (weighted 1x)
interests.excluded	array	[]	Topics to always filter out
max_articles	number	15	Maximum articles in final digest
summary_length	string	"medium"	Summary detail: short, medium, long
sources.rss	array	[]	RSS feed URLs or objects with config

Core Concepts

Concept	Purpose	How It Works
Interest Scoring	Ranks articles by personal relevance	Matches article content against your interest keywords with TF-IDF weighting
Cross-Source Dedup	Prevents duplicate stories	Compares article titles and content using cosine similarity, keeps highest-scored version
Source Diversity	Prevents single-source dominance	Caps any single source at 30% of total articles to ensure breadth
Temporal Decay	Prioritizes fresh content	Articles lose relevance score over time; 12-hour-old story scores 50% of a 1-hour-old story

Architecture

RSS Feeds ────────> Feed Parser ──────┐
  Hacker News API ──> HN Fetcher ──────┤
  Reddit API ───────> Reddit Fetcher ──┼──> Deduplicator ──> Interest Scorer ──> Ranker ──> Summarizer ──> Output

Workflow Examples

Scenario 1: Developer staying current with AI

Input: 5 RSS feeds, HN top stories, 3 ML subreddits, primary interest in AI tooling Processing: Scans 142 articles, deduplicates 23 cross-posted stories, scores remaining 119 against AI tooling interest, selects top 15 with source diversity cap. Output: 15-article digest with 3 top stories fully summarized and 12 brief headlines with relevance tags.

Scenario 2: Startup founder tracking competitors

Input: RSS feeds from competitor blogs, industry news, VC funding feeds, tracked company names Processing: Highlights any mention of tracked companies, flags funding announcements, and groups competitive intelligence separately from general news. Output: Digest with dedicated "Competitor Watch" section showing 2 competitor blog posts and 1 funding announcement.

Best Practices

1. Start with 5-8 sources and expand gradually -- Too many sources on day one creates noise. Add one new source per week and evaluate whether it improves your digest quality.

2. Use excluded interests aggressively -- It is easier to exclude noise than to perfectly define signal. If crypto news keeps appearing in your tech feed, add it to.

3. Set appropriate score thresholds -- HN score of 100 captures major stories but filters low-engagement posts. Reddit varies by subreddit size: use 50 for small.

Common Issues

1. RSS feed returning empty results -- Some sites require a User-Agent header. The skill uses a standard browser UA by default, but some feeds may be geo-restricted.

2. Too many articles from one source -- The source diversity cap defaults to 30%. If one source still dominates, reduce its max_items in the per-source config or.

Privacy & Data Handling

All news fetching and summarization happens locally. Article content is fetched directly from source URLs and processed on your machine. No article text or reading patterns are sent to external analytics services. Feed subscription lists are stored in ~/.claude/config/news-curator/ and cached articles expire after 48 hours in ~/.claude/cache/news-curator/.