Open‑Source AI Skill that Scrapes 7 Platforms to Deliver Verified Trend Reports

Signal Mining beyond Simple Aggregation

The open‑source project last30days implements an AI research skill that crawls multiple platforms, identifies community‑validated hotspots (likes, shares, discussions, or real‑money bets), and synthesises a narrative report with citations.

Covered Signal Sources

Data is collected from seven platforms: Reddit (including top‑level comments), X (Twitter), YouTube, Hacker News, Polymarket (prediction markets), Instagram Reels, and Bluesky. This mix spans technical discussions, short‑form videos, and market‑based probability signals.

Multi‑Source Insight Example

Querying “Claude Code vs Cursor” returns technical blog comparisons, Reddit sub‑forum rants, YouTube review snippets, Hacker News critiques, and Polymarket odds on each tool’s market share.

Technical Core – Multi‑Signal Quality Ranking Pipeline

The system applies a weighted scoring pipeline rather than equal‑weight aggregation. For each result it computes a composite score from:

Bidirectional text similarity with synonym expansion

Standardised interaction heat (likes, retweets, views)

Source authority weight

Cross‑platform convergence detection using mixed‑triple‑tag Jaccard similarity

Time‑decay factor

For Polymarket predictions the ranking additionally incorporates text relevance, 24‑hour trading volume, liquidity depth, price‑change velocity, and outcome competitiveness.

“AI reshapes itself every month. This skill keeps you in sync.” – README opening line

v2.9.5 – Comparison Mode

Version 2.9.5 adds a “comparison mode.” A query formatted as “X vs Y” launches three parallel research threads and produces a side‑by‑side report that includes strengths, weaknesses, a direct comparison table, and a data‑driven conclusion.

Integration into Claude Code Workflow

The skill is packaged as a Claude Code plugin. After adding the source in Claude’s plugin market, a single command installs it. Users invoke the skill with a slash command, e.g., /last30days Sora video generation, which triggers automatic crawling of the past 30 days across all platforms and generates a report.

Since v2.9.1 the tool auto‑archives each full briefing as a Markdown file named after the query in ~/Documents/Last30Days/, gradually building a personal research library.

Intended Users

AI developers and researchers – rapid insight into community practices, pitfalls, and solutions for topics such as LoRA or RAG.

Tech media and content creators – discover topic ideas and verify emerging trends.

Product managers and market analysts – fast competitor or market perception reports via comparison mode.

Investors and trend observers – leverage Polymarket data to gauge public expectations on events like major model releases.

Outlook

The project has accumulated tens of thousands of GitHub stars, indicating strong demand for automated, personalised information‑processing tools. It moves from a passive subscription service toward an active research assistant capable of analysing month‑long sentiment shifts, highlighting developer pain points, and rendering charts from a single prompt.