analyze-article

X-McKay's avatarfrom X-McKay

Use LLM to analyze a raw article. Generates summary, extracts entities, categorizes content, assigns importance score, and detects breaking news.

1stars🔀0forks📁View on GitHub🕐Updated Jan 11, 2026

When & Why to Use This Skill

This Claude skill leverages advanced LLMs to perform deep analysis and enrichment of raw news articles. It automatically generates concise summaries, extracts key entities (companies, people, technologies), categorizes content into specific domains, and calculates an importance score to help users prioritize breaking news and significant industry developments.

Use Cases

  • Automated News Triage: Efficiently process high volumes of news feeds by automatically identifying breaking news and high-impact articles based on AI-driven importance scores and 'breaking' flags.
  • Competitive Intelligence: Monitor industry trends and competitor activities by extracting mentions of specific companies, people, and emerging technologies from raw news sources.
  • Content Curation & Knowledge Management: Streamline the organization of research by automatically categorizing articles and generating structured metadata for searchable databases or newsletters.
  • Primary Source Prioritization: Automatically boost the visibility of official company announcements and blog posts to ensure primary source information is highlighted over secondary reporting.
nameanalyze-article
descriptionEnriched article with AI-generated metadata
domainnews
categorydiagnostic
mcp-servers[]
requires-approvalfalse
confidence0.85
- nameprocessed_article
typeProcessedArticle

Analyze Article

Use LLM to analyze and enrich a single news article.

Preconditions

  • Article has valid title and summary/content
  • LLM API (vLLM) is accessible

Actions

Step 1: Prepare Content

Combine article title and summary for analysis. Truncate if longer than 2000 characters.

Step 2: Call LLM for Analysis

Send article to LLM with structured prompt requesting:

  1. Summary: 2-3 sentence summary of key points
  2. Category: One of research, business, product, security, policy, general
  3. Entities: List of companies, people, technologies, models mentioned
  4. Importance Score: 1-10 rating where:
    • 1-3: Minor news, incremental updates
    • 4-6: Notable news, meaningful developments
    • 7-8: Important news, significant impact
    • 9-10: Major news, industry-changing announcements
  5. Is Breaking: True if major breaking news requiring immediate alert
  6. Breaking Reason: Explanation if breaking (e.g., "Major model release")

Step 3: Parse LLM Response

Extract structured data from JSON response. Handle:

  • Qwen3 thinking tags (<think>...</think>) - strip them
  • Markdown code blocks around JSON
  • Invalid JSON - return defaults

Step 4: Apply Importance Boost

Boost importance score by +2 (max 10) for official company blog posts, as these represent primary source announcements.

Step 5: Generate Content Hash

Create hash from title + URL for deduplication tracking.

Success Criteria

  • Article enriched with AI summary
  • Valid category assigned
  • Entities extracted (may be empty)
  • Importance score between 1-10
  • Breaking news flag set appropriately

Fallback Behavior

If LLM call fails:

  • Use original summary as AI summary
  • Set category to general
  • Set importance to 5
  • Set is_breaking to false
  • Return entities as empty list