last30days-skill Scans 10 Platforms in One Command — No Wonder It Hit 1 on GitHub Trending

AI agents are great at writing code, summarizing documents, and answering questions from their training data. But ask one what people on Reddit, X, or Hacker News are actually saying about a topic right now, and you hit a wall. Training data is months old. Web search returns SEO-optimized pages, not community conversations. And manually checking ten platforms for sentiment on a single topic? Nobody has time for that.

That’s the exact gap last30days-skill fills. It’s a Claude Code skill — a plugin you install with a single command — that turns your AI agent into a cross-platform research engine. Give it a topic, and it goes out to Reddit, X/Twitter, YouTube, TikTok, Instagram, Hacker News, Bluesky, Polymarket, and the general web, pulls back what people are actually discussing, scores and deduplicates the results, and hands you a cited research briefing. As of March 26, 2026, it sits at #1 on GitHub Trending with over 8,300 stars and 750+ forks, and the number is still climbing.

What It Actually Does

The core loop is simple: you type /last30days followed by a topic. The skill then:

Parses your intent — classifying your query into one of five types (recommendations, news, prompting, comparison, or general) to tailor how it searches
Runs parallel searches across up to 10 platforms simultaneously
Scores results using a composite algorithm that weighs text relevance, engagement velocity, authority signals, and recency
Detects cross-platform convergence — if the same story or opinion shows up on Reddit, X, and Hacker News, it gets flagged as high-signal evidence and weighted accordingly
Synthesizes everything into a narrative with inline citations, a stats block showing platform-by-platform breakdowns, and actionable follow-up suggestions based on what it actually found

The output gets auto-saved as a Markdown file to ~/Documents/Last30Days/, so over time you build a personal research knowledge base without extra effort.

There’s also a depth dial. Quick mode (--quick) pulls 8–12 sources per platform for a fast overview. Default mode balances thoroughness at 20–30 sources. Deep mode (--deep) goes all in with 50–70 Reddit threads and 40–60 X posts — useful when you need comprehensive coverage for a report or strategic decision.

The Comparison Mode That Makes It Stand Out

One feature that separates last30days-skill from a simple multi-platform search aggregator is its comparative analysis mode. Ask “Cursor vs Windsurf” or “React Native vs Flutter in 2026,” and it runs three parallel research passes — one for each item plus a direct comparison search — then structures the output as a side-by-side community sentiment analysis.

This matters because comparison questions are notoriously hard to answer with standard search. Google gives you affiliate-driven listicles. ChatGPT gives you hedged answers based on outdated training data. last30days-skill gives you what developers and users are actually saying this month, with engagement numbers attached. You can see that 45 Reddit threads favor one tool while X sentiment leans the other way, and make your own call.

The Polymarket integration adds another layer. For topics with active prediction markets — elections, crypto, product launches — the skill pulls current odds and price movements, weaving them into the narrative with specific percentages rather than vague statements.

How It Stacks Up Against Perplexity, Grok, and ChatGPT Deep Research

The obvious question: why not just use Perplexity or Grok’s deep research mode?

Perplexity is strong at factual, citation-based research across web pages, but it doesn’t dig into Reddit comment threads, YouTube transcripts, TikTok trends, or prediction markets. It’s an answer engine optimized for verifiable facts, not community sentiment aggregation.

Grok has deep X/Twitter integration thanks to the xAI connection, and it’s fast — roughly 10x faster than ChatGPT’s Deep Research for web-based queries. But its strength is concentrated on X data. It doesn’t systematically scan Reddit, TikTok, Instagram, or Polymarket in the same structured way.

ChatGPT Deep Research is thorough but slow, and it operates as a black box — you get a polished report but limited visibility into what sources were checked and what was filtered out.

last30days-skill takes a different approach entirely. It’s not a hosted product; it’s an open-source tool that runs inside your existing AI agent workflow. The key differentiators:

Platform breadth: 10+ sources vs. web-focused or single-platform bias
Transparency: you see the stats block with exact counts — how many Reddit threads, X posts, YouTube videos, and HN stories were analyzed, with engagement metrics for each
Cross-platform signal detection: stories that appear across multiple platforms get boosted, something no single-platform research tool can replicate
Local-first data: everything saves to your machine as Markdown, not locked in a SaaS product
Agent-composable: with the --agent flag, other AI agents can call it programmatically and get structured output, enabling multi-agent research workflows

The tradeoff is setup friction. Perplexity and Grok work out of the box. last30days-skill requires API keys and a compatible agent environment.

Setup, Pricing, and the API Key Situation

last30days-skill itself is free and open source. The costs come from the data access layer.

The most important key is ScrapeCreators (SCRAPECREATORS_API_KEY), which covers Reddit, TikTok, and Instagram with a single credential. ScrapeCreators operates on a pay-as-you-go credit system: 100 free credits to start, then $10 for the Solo Dev plan, $47 for Freelance, and $497 for Business. Credits never expire, so you only pay when you need more.

For X/Twitter data, you can use your own auth tokens (AUTH_TOKEN and CT0) or fall back to the xAI API. Bluesky access requires an app password. YouTube works through yt-dlp for transcript extraction — no API key needed.

Installation varies by platform:

Claude Code: Install via plugin marketplace with two commands
Gemini CLI: Install as an extension
OpenAI Codex CLI: Manual installation supported
Any environment: Clone the repo and configure manually

The skill supports per-project configuration through a .claude/last30days.env file, so different repositories can use different API keys — useful for teams where personal and work accounts need separation.

Version 2.9.5 (current as of March 2026) added Bluesky/AT Protocol support, the comparative analysis mode, config validation on startup, and expanded the test suite to 455+ tests.

Why 8,000+ Developers Starred It This Week

The star count tells part of the story, but the why is more interesting. Three trends are converging:

The AI agent skill ecosystem is exploding. Claude Code’s skill system, Gemini CLI’s extensions, and Codex CLI’s plugin architecture have created a market for specialized agent capabilities. last30days-skill is one of the first skills to demonstrate that a single plugin can meaningfully expand what an AI agent can do.

Real-time data is AI’s biggest blind spot. LLMs are trained on static snapshots. The gap between what an AI “knows” and what the internet is discussing right now keeps growing as the pace of change accelerates. A skill that bridges this gap addresses a pain point every AI power user feels daily.

Research workflows are going agent-native. Instead of alt-tabbing between Reddit, X, YouTube, and a dozen browser tabs, developers are increasingly asking their AI agents to do the synthesis. last30days-skill fits naturally into this workflow because it lives where the work already happens — inside the terminal.

The forks (750+) suggest active modification and extension by the community, and there are already third-party variants appearing on GitHub, including versions that integrate with Composio and other agent frameworks.

FAQ

Is last30days-skill free to use?
The skill itself is free and open source. You’ll need API keys for data access — ScrapeCreators (starts free with 100 credits, then $10+) covers Reddit, TikTok, and Instagram. X/Twitter access uses your own auth tokens at no additional cost. Bluesky and YouTube access are also free.

Does last30days-skill work only with Claude Code?
No. While Claude Code is the recommended platform, it also supports Gemini CLI and OpenAI Codex CLI. Manual installation works with other agent environments as well.

How long does a research run take?
Typically 2–8 minutes depending on the topic and depth setting. Quick mode (--quick) runs faster with fewer sources; deep mode (--deep) is more thorough but takes longer.

What are the main competitors to last30days-skill?
For general AI research, Perplexity and Grok Deep Research are the closest comparisons, though they focus on web pages and X respectively. For agent-native research skills, Orchestra Research’s AI-Research-SKILLs library covers similar ground but with a different architecture. No other tool currently matches the 10-platform parallel search with cross-platform signal detection.

Can I use last30days-skill in automated workflows?
Yes. The --agent flag produces non-interactive, structured output that other AI agents can consume directly. This makes it composable in multi-agent pipelines — for example, a monitoring agent could run daily research on a topic and feed the results into a reporting workflow.

Top AI Product

Leave a comment Cancel reply