| name | agent-news |
| description | Fetch, parse, and summarize news from machine-readable sources designed for AI agents. Use when the user asks for news, headlines, or current events via structured feeds — RSS/Atom, JSON APIs, or llms.txt-compatible endpoints — without scraping human-facing GUIs.
|
Agent News Skill
Retrieves news from AI-friendly structured sources (JSON APIs > RSS/Atom > llms.txt). Never scrapes GUI pages when a feed exists.
Usage
python3 _scripts/agent_news.py
python3 _scripts/agent_news.py --topic "AI"
python3 _scripts/agent_news.py --sources "BBC,TechCrunch" --limit 20
python3 _scripts/agent_news.py --use-api
python3 _scripts/agent_news.py --use-api --topic "AI agents"
Source Hierarchy
- JSON APIs (requires
--use-api flag + keys in .env):
- World News API (
WORLD_NEWS_API_KEY) — global coverage, topic search + top news
- NewsAPI.org (
NEWS_API_KEY) — top headlines from major outlets
- GNews (
GNEWS_API_KEY) — breaking news by topic
- RSS/Atom — BBC, Reuters, NYT, TechCrunch, VentureBeat, OpenAI, Anthropic, etc.
- llms.txt — emerging machine-readable index standard
- tophub.today — Chinese news (legacy, via
tophub_news.py)
Key Principles
- Only report articles actually returned by the feed (no hallucination)
- Always include outlet name, original URL, and published timestamp
- Cache results locally if running multiple queries (respect rate limits)
Output
Sources/Agent News Digest - YYYY-MM-DD.md