name: filter-ai-relevant version: "2.0.0" description: > DEPRECATED: This skill has been replaced by the analyze-article skill which provides more accurate AI-relevance filtering through LLM analysis rather than simple keyword matching. license: MIT deprecated: true deprecated_by: "analyze-article" deprecated_reason: "LLM-based analysis provides better relevance filtering than keyword matching" metadata: kubani: domain: news category: collection requires_approval: false confidence: 0.0 version: "2.0.0" status: deprecated

Filter AI-Relevant Articles (DEPRECATED)

⚠️ DEPRECATED: This skill has been deprecated as of v2.0.0.

AI relevance filtering is now handled by the analyze-article skill, which uses LLM analysis to accurately categorize and filter articles rather than simple keyword matching.

Migration Guide

Before (v1.0)

Filtering used keyword matching:

1. fetch-rss-feeds - Collect articles
2. filter-ai-relevant - Keyword-based filtering
3. analyze-article - LLM analysis

After (v2.0)

LLM analysis handles relevance determination:

1. fetch-rss-feeds - Collect articles (with dedup)
2. analyze-article - LLM categorizes AND filters for relevance

How to Migrate

If you were using filter-ai-relevant, update your workflow to rely on analyze-article for relevance filtering.

The new approach:

• analyze-article categorizes articles as: research, business, product, security, policy, general
• Articles can be filtered by category after analysis
• LLM provides more nuanced relevance determination than keywords

Replacement Skill

Why Deprecated

1. Better accuracy: LLM understands context, not just keywords
2. Richer output: Categories instead of binary relevance
3. No maintenance: No keyword list to maintain
4. Semantic matching: Understands synonyms and related concepts

Example Comparison

Keyword approach (old):

• "Meta releases new AI model" → filtered (contains "AI")
• "Facebook updates metadata API" → filtered (contains "meta")
• "Novel transformer architecture for proteins" → missed (no exact keyword)

LLM approach (new):

• "Meta releases new AI model" → category: product, AI-relevant
• "Facebook updates metadata API" → category: general, not AI-focused
• "Novel transformer architecture for proteins" → category: research, AI-relevant

See Also

• analyze-article - v2.0 with LLM categorization