An AI agent-driven multi-platform aggregated search engine for the last 30 days, ranked by social engagement signals across 15+ sources including Reddit, X, YouTube, HN, and Polymarket.
last30days-skill is a research-oriented skill for AI agents (CLI command /last30days), currently at v3.0.0, written in Python (≥3.12). It simultaneously searches 15+ data sources including Reddit (with comments and vote counts), X/Twitter, YouTube (full transcripts), TikTok, Instagram Reels, Hacker News, Polymarket, GitHub, Threads, Pinterest, Bluesky, Perplexity Sonar, and the Web — ranking by real engagement signals such as upvotes, likes, views, and Polymarket odds rather than traditional SEO ranking.
Core mechanisms include: a v3 intelligent topic resolution engine (pre-search identification of relevant accounts, subreddits, channels, and tags, with bidirectional person→company and product→founder resolution plus entity disambiguation), cross-source cluster merging (entity-overlap detection to unify multi-platform appearances of the same event into a single entry), a dual scoring system (social engagement scoring + Best Takes humor/virality scoring), and a per-author cap of 3 results to prevent voice monopolization. Supports X vs Y single-pass comparison queries (~3 minutes to generate comparison tables covering architecture, memory, security features), GitHub person mode (--github-user flag), and ELI5 mode.
Zero-config access to Reddit, HN, Polymarket, and GitHub; remaining sources unlocked via BYOK (X requires browser login, YouTube requires yt-dlp, TikTok/Instagram etc. require ScrapeCreators key, Perplexity requires OpenRouter key, Web search requires Brave Search key). Supports Claude Code, claude.ai, Gemini CLI, and OpenClaw as host platforms. Architecturally, a Python pre-research module handles topic resolution, multi-source parallel search with timeout budget fallback ensures resilience, backed by 1,012 test cases.