Social Media Research Skills for AI Agents: ScrapeCreators' Free Toolkit

Most people asking an AI agent to research a competitor's social accounts end up doing the same thing twice: writing scraping logic, then writing the analysis prompt on top of it. Social Media Research Skills for AI Agents, an open-source repo from ScrapeCreators, tries to remove the first half of that. It packages a set of research workflows an agent like Claude Code can install and run directly, instead of improvising a scraper from scratch every time.

The repo is on GitHub at ScrapeCreators/social-media-research-skills. As of today, it sits at 1,411 stars and 15 forks, up from the 1,036 stars logged in an earlier snapshot just a day or two prior, which puts real-world growth ahead of what that snapshot suggested.

What It Actually Does

The repo is not a scraper. It is a library of 13 "skills," written to the Agent Skills spec, that sit on top of the ScrapeCreators API and turn a research request into a finished output rather than a raw JSON dump. The skills cover:

• Outlier post finder: surfaces posts, reels, shorts, or tweets that beat a creator's own baseline, not just posts with high raw numbers.
• Transcript intelligence: pulls hooks, claims, and quotable lines out of video transcripts across TikTok, Instagram, YouTube, X, LinkedIn, Facebook, Rumble, and Reddit.
• Comment mining: extracts objections, questions, and the audience's actual language from a post's comment section.
• Competitor social research and creator profile teardown: build a brief on what a competitor or creator is doing and why it works.
• Ad library teardown: reads active Meta, Google, and LinkedIn ads for hooks, offers, and CTAs.
• Trend discovery, influencer prospecting, audience research, social listening brief, product demand research, and content repurposing round out the rest.
• scrapecreators-api: a routing skill the other twelve fall back on when they need a specific endpoint.

Platform coverage is broad: TikTok, Instagram, YouTube, Reddit, X/Twitter, LinkedIn, Facebook, Threads, Bluesky, Pinterest, Rumble, and multiple ad libraries.

How Someone Would Actually Use It

Installation is a single command:

npx skills add ScrapeCreators/social-media-research-skills

It works with Claude Code, Cursor, GitHub Copilot, Gemini CLI, Windsurf, OpenAI Codex, and VS Code, since it follows the shared Agent Skills spec rather than a proprietary format. After install, it needs a ScrapeCreators API key set as an environment variable (SCRAPECREATORS_API_KEY), obtained from scrapecreators.com.

From there, usage is plain-language prompting, not endpoint calls. The README's own examples give a sense of the register:

"Find the outlier posts for @starterstory on YouTube Shorts from the latest page of videos."

"Mine the comments on this viral Instagram Reel. I want objections, questions, buying intent, and exact audience language."

The agent picks the right skill, calls the ScrapeCreators API underneath, and returns a structured brief, an outlier table, or a comment-theme report, complete with source links where relevant. That's the actual value: someone doing a competitor scan does not have to specify which TikTok or Instagram endpoint to hit, or write parsing logic for the response. They describe the outcome and the skill routes the request.

The design principles stated in the README are worth noting because they explain the shape of the outputs: results are meant to be baseline-aware (a post's performance judged against that account's own norm, not raw view counts alone), source-cited, and built around preserving exact language from comments and transcripts rather than paraphrasing it away.

Who This Is Genuinely For

This fits anyone already running social competitor scans, content research, or influencer vetting through an AI agent workflow, marketing teams, solo creators doing niche research, or anyone building an AI agent competitor research habit into their weekly process. It is closer to infrastructure than to a finished dashboard: the payoff shows up when someone is already comfortable prompting an agent and wants the data-pulling layer handled instead of hand-rolled.

It is not a good fit for someone who wants a point-and-click social listening tool with a visual dashboard. There is no UI here, only skill files an agent reads and executes. It also assumes some tolerance for iterating on prompts to get the output shape right, since the skills are workflows, not fixed reports.

Limitations Worth Knowing

The skill library itself is free and open-source, but it is not a standalone free tool. Every skill calls the ScrapeCreators API underneath, and that API is a separate paid service. ScrapeCreators does offer free credits to start with no card required, and unused credits do not expire, but any real volume of research eventually runs into paid usage. Treat "free, open-source" as describing the skill code, not the data pipeline behind it.

The GitHub repo also does not carry a published license file as of this writing, which matters for anyone planning to fork or redistribute the skills commercially rather than just install and use them as intended.

Data coverage is limited to public social information. The README is explicit that ScrapeCreators does not access logged-in or private data, so anything behind a login wall, like private accounts or gated ad details, is out of reach.

The Bottom Line

For an AI operator already using Claude Code or a similar agent for research work, this repo is a reasonable way to standardize social media pulls instead of reinventing them per project. The skill layer is free to install; budget for the API key separately, and check the license terms before building anything commercial on top of it.

Source: github.com/ScrapeCreators/social-media-research-skills