Pre-Built Social Research Skills for AI Agents, Not API Glue

What it does This repo packages social media research workflows—like finding outlier posts, mining comments for pain points, tearing down competitor ads, and prospecting influencers—as agent skills compatible with Claude Code, Cursor, Codex, and similar tools. Each skill pulls public data from TikTok, Instagram, YouTube, Reddit, X, LinkedIn, and elsewhere via the ScrapeCreators API, then formats it into business artifacts such as tables, briefs, or CSVs rather than dumping raw JSON.

The interesting bit The design explicitly rejects “endpoint routing” as the goal. Skills are baseline-aware, judging a post’s performance against a creator’s own history instead of raw vanity metrics, and they preserve exact audience language—hooks, objections, quotes—so the output feels like research, not a data pipe.

Key highlights

• Over a dozen distinct skills covering everything from transcript intelligence to ad library teardowns and product demand validation.
• Outputs are actionable artifacts: outlier tables, VOC reports, competitor briefs, trend evidence tables, and outreach-ready prospect lists.
• Baseline-aware analysis compares performance against a creator’s own typical metrics, not just absolute numbers.
• Cited outputs include source URLs for posts, videos, and comments whenever possible.
• Works with Claude Code, Cursor, Codex, Gemini CLI, and other agents supporting the Agent Skills spec.

Caveats

• Requires a ScrapeCreators API key; the underlying data layer is a paid service, not self-hosted scraping.
• Only accesses public social data, so private or logged-in content is explicitly out of scope.

Verdict Worth a look if you want AI agents to produce actual research briefs rather than raw social data dumps. Skip it if you need private account analysis or a free, standalone scraper.