You're writing a prompt for a new Claude Code workflow and wondering: what prompting techniques are people actually getting results with this month? Instead of spending 45 minutes trawling through Reddit threads, YouTube tutorials, and Hacker News comments, you type one command and last30days-skill does the research across 10+ platforms simultaneously. This agent skill has 12.7K GitHub stars because it solves a problem every AI developer hits: staying current. It searches Reddit, X, Bluesky, YouTube, TikTok, Instagram, Hacker News, Polymarket, and the general web within a strict 30-day recency window. No stale results from 2024 mixed in with 2026 data. Everything it finds is ranked using bidirectional text similarity, engagement velocity, and cross-platform convergence detection — meaning if the same technique is trending on both Reddit and Hacker News simultaneously, it gets ranked higher. The output isn't a wall of links. last30days synthesizes findings into three actionable deliverables: best practices extracted from real community discussions, a prompt pack of copy-paste-ready prompts that people are actually using successfully, and a reusable context snippet you can drop into any project. Version 2.9.5 (current) supports Claude Code as the primary platform, with compatibility for Gemini CLI and OpenAI's Codex CLI. The v2.1 update added watchlists for persistent topic tracking — set it and forget it, then check back for accumulated briefings. YouTube transcript analysis was also added, so video content gets included in research rather than just titles and descriptions. Installation is two commands for Claude Code users. The skill requires API keys for ScrapeCreators (handles Reddit, TikTok, and Instagram scraping) and X/Twitter authentication tokens. Bluesky credentials are optional. Research results auto-save to ~/Documents/Last30Days/ for later reference. The comparative mode is particularly useful: ask 'Claude Code vs Cursor for React development — what are people saying this month?' and get a side-by-side analysis of real community sentiment, not marketing copy. Per-project configuration via .claude/last30days.env means different projects can track different topics. The honest limitation: research quality depends entirely on the scraping APIs and your authentication tokens. If ScrapeCreators rate-limits you or your X auth token expires, sources go silent without warning. The skill doesn't tell you which sources failed — your results just get thinner.
Installation
Key Features
- ✓Searches 10+ platforms simultaneously: Reddit, X, Bluesky, YouTube, TikTok, Instagram, HN, Polymarket, and web
- ✓Strict 30-day recency window ensures all results are current — no stale data contamination
- ✓Quality scoring via bidirectional text similarity, engagement velocity, and cross-platform convergence detection
- ✓Outputs three deliverables: best practices, copy-paste prompt pack, and reusable context snippet
- ✓Watchlist mode for persistent topic tracking with accumulated briefings over time
- ✓YouTube transcript analysis includes video content in research, not just metadata
- ✓Comparative mode: 'X vs Y' queries produce side-by-side community sentiment analysis
- ✓Per-project configuration via .claude/last30days.env for context-specific research
Use Cases
- →You're starting a new project with an unfamiliar framework — ask 'what are the best practices for Remix v3 this month' and get community-tested patterns instead of outdated docs
- →Your team is debating tool choices — run 'Vercel vs Cloudflare Workers — what are developers saying' for real sentiment data instead of guessing
- →You want to write a blog post about a trending topic — last30days gives you the current conversation landscape across platforms so you know what angles are already covered
- →You're building prompts for a client project — get the prompt techniques that real users report working in the last 30 days, not theoretical best practices from 2024
- →Track a competitor's product mentions across social platforms with watchlist mode — weekly briefings show sentiment shifts over time