mvanhorn/last30days-skill

Metadata

• **Full name:** mvanhorn/last30days-skill
• **Description:** AI agent skill that researches any topic across Reddit, X, YouTube, HN, Polymarket, and the web - then synthesizes a grounded summary
• **Stars:** 20,980
• **Last pushed:** 2026-04-11
• **License:** MIT
• **Topics:** ai-skill, claude-code, deep-research, reddit, research, social-media, twitter, youtube, web-search, polymarket, hackernews

Archetype

**utility-with-skill** - Single-purpose research aggregation skill with a Python backend. Searches 13+ platforms in parallel, scores results by engagement metrics (upvotes, likes, real money), and synthesizes into grounded summaries.

Structure

SKILL.md                  # Runtime skill spec (source of truth)
skills/last30days/        # Skill definition
agents/                   # Agent configurations
hooks/                    # Hook definitions
scripts/                  # Python research engine
vendor/                   # Vendored dependencies
tests/                    # Test suite
fixtures/                 # Test fixtures

Extractable Value

Processes (specializations/)

1. **multi-source-research-synthesis** - The v3 research methodology is interesting: intelligent pre-research (entity resolution before search), parallel multi-source search, engagement-weighted scoring (upvotes > editorial SEO), cross-source cluster merging (dedup same story across platforms), synthesis with source citations. This is a research methodology pattern, not just a tool. **Placement: specializations/shared/multi-source-research**

Plugin Ideas

1. **last30days-integration** - A babysitter marketplace plugin that wraps /last30days as a research task. The pre-meeting research, competitive analysis, and trend monitoring use cases map well to babysitter orchestration tasks. The plugin would provide a research task type that delegates to the last30days engine. However, this is borderline since it requires the user to have last30days installed separately.

SKIP

• The actual platform-specific API integrations (Reddit JSON, X API, YouTube transcripts) - tool-specific implementation details
• ELI5 mode - presentation concern, not methodology

Processes

1. Multi-Source Research Synthesis

• **Source**: V3 research methodology with intelligent pre-research and engagement-weighted scoring
• **Placement**: specializations/shared/multi-source-research-synthesis.js
• **Description**: Research methodology pattern: entity resolution before search → parallel multi-source search → engagement-weighted scoring (upvotes > editorial SEO) → cross-source cluster merging → synthesis with source citations.

Plugin Ideas

• **Last30Days Research Integration**: Babysitter marketplace plugin that wraps /last30days as a research task for pre-meeting research, competitive analysis, and trend monitoring.

Library Mapping

Plugin Marketplace Mapping

Priority Assessment

**LOW-MEDIUM** - The multi-source research synthesis methodology is conceptually interesting but the actual value is tightly coupled to the platform integrations (API keys, browser sessions). The methodology pattern (entity-resolve -> parallel-search -> engagement-score -> cluster-merge -> synthesize) could inform a generic research process, but it's thin without the platform access. Better as inspiration than a direct port.