ai-news-digest_skill

Overview

This skill aggregates AI news from 20+ sources, deduplicates overlapping items, classifies by theme, and generates an auditable digest in Markdown or JSON. It supports multi-language output (default Chinese), image export, and preserves source links and raw metadata for traceability.

How this skill works

The skill loads a prioritized source list (RSS/Atom first, HTML fallback), fetches and normalizes entries, then performs deduplication and topic classification. It sorts results by time and source weight, optionally extracts full text, translates summaries if needed, and renders a templated Markdown digest (or JSON/image) with per-item metadata and source tracing.

When to use it

• Daily or ad-hoc AI news briefings for teams or newsletters
• Monitoring a specific time window (today/yesterday/specified date range)
• Compiling multi-source summaries while avoiding duplicate coverage
• Generating translated digests for multilingual audiences
• Producing social-share images or Markdown-ready reports for publication

Best practices

• Prefer RSS/Atom-enabled sources; falling back to HTML only when RSS is unavailable
• Keep default time zone (Asia/Shanghai) unless you need a specific regional window
• Set sensible max and max-per-topic limits to maintain digest readability
• Do not attempt to bypass paywalls; include only public meta information for restricted content
• Enable LLM translation only when higher-quality translations are needed and API keys are configured

Example use cases

• Generate today’s Chinese AI digest from all default sources and save as digest.md
• Produce a JSON feed of AI product launches from the last 7 days for downstream processing
• Create a landscape shareable image of top 5 AI research items for social distribution
• Fetch yesterday’s AI policy updates limited to the policy topic and include source tracebacks
• Run a smoke test to verify feed connectivity and list any failed sources for investigation