ScrapeCreators Ships 13 Open-Source Agent Skills for Social Media Research
ScrapeCreators shipped a plugin bundle of 13 agent skills that lets coding assistants run full social media research workflows across TikTok, Instagram, YouTube, Reddit, X, LinkedIn and more.
Takeaways
- ScrapeCreators released 13 open-source agent skills for social media research across 10-plus platforms.
- Installs via
npx skills addinto Claude Code, Cursor, Codex, Gemini CLI, Windsurf and more. - Workflows include outlier post finding, comment mining, ad library teardowns, transcript intelligence, and influencer prospecting.
- Skills are MIT licensed; ScrapeCreators API starts at 100 free credits, then $47 for 25,000.
- Cache hits cost zero credits and purchased credits never expire, lowering real cost for repeat polling.
- Limits: public data only, no general X search, Instagram hashtag pages cap at 11.
ScrapeCreators has released an open-source bundle that gives coding agents reusable social-media research procedures. The package contains 13 skills: one shared API layer and 12 workflows for discovery, audience mining, competitive analysis, market research, and content repurposing. It runs in Claude Code, Cursor, Codex, and Gemini CLI. At the time of writing, the GitHub repository had attracted roughly 4,000 stars.
Thirteen skills, one data layer
Each module follows the Agent Skills specification, a format for packaging instructions, tool calls, and output schemas into components that an agent can load when needed. The shared scrapecreators-api skill routes requests and handles pagination against ScrapeCreators’ public-data API. The remaining skills combine those calls into finished artifacts such as CSV files, research briefs, and teardown reports.
| Function | Skills | Purpose |
|---|---|---|
| Data layer | scrapecreators-api |
Routes API requests and handles pagination. |
| Discovery | outlier-post-finder, trend-discovery, influencer-prospecting |
Finds posts outperforming a creator’s usual results, identifies niche formats and hashtags, and builds creator lists with fit scores. |
| Audience mining | transcript-intelligence, comment-mining |
Extracts hooks, claims, quotations, objections, questions, and audience language from TikTok, Instagram, YouTube, Facebook, X, LinkedIn, Rumble, and Reddit. |
| Competitive analysis | competitor-social-research, ad-library-teardown, creator-profile-teardown |
Produces briefs covering content pillars, hooks, offers, calls to action, and recommendations. Ad analysis covers Meta, Google, and LinkedIn. |
| Market research | audience-research, social-listening-brief, product-demand-research |
Combines signals from social platforms and Reddit threads into audience, demand, and listening reports. |
| Repurposing | content-repurposing |
Converts transcripts and posts into LinkedIn posts, X threads, scripts, and newsletter drafts. |
| Total | 13 skills | One API layer and 12 research workflows. |
One command, several runtimes
With Node.js and npm available, developers can install the bundle from the command line:
npx skills add ScrapeCreators/social-media-research-skills
Authentication requires a ScrapeCreators API key supplied through an environment variable. The installed skills still depend on the hosted ScrapeCreators API for data collection.
After installation, a developer can use natural-language requests such as:
- Find outlier YouTube Shorts for a specified creator.
- Mine comments on an Instagram Reel for buying intent and preserve the original wording.
- Analyze a competitor’s active Meta, Google, and LinkedIn ads for hooks, offers, and calls to action.
- Turn a video transcript into a LinkedIn post, an X thread, and a newsletter draft.
The agent selects the relevant skill, sends requests through the shared API layer, and formats the results according to that workflow’s output schema.
Recipes keep the evidence attached
ScrapeCreators exposes more than 37 endpoints, while each workflow defines a narrower business outcome, the required sequence of calls, and the final deliverable. This orchestration reduces the endpoint selection and prompt design that developers would otherwise implement for each research task.
- Baseline-aware scoring: Performance analysis compares each post with that creator’s usual results.
- Source citations: Referenced posts and ads retain their original URLs.
- Verbatim evidence: Comments and transcript excerpts preserve the source language.
- Structured output: Workflows return consistent briefs, reports, drafts, or tabular files.
Exact audience language preserves objections, questions, and buying phrases that product and marketing teams may need for copy or classification. Explicit workflow instructions also prevent the model from smoothing those phrases into generic summaries.
API calls determine the bill
The MIT license covers the skill definitions and orchestration code. Requests to the underlying data API consume paid credits under ScrapeCreators’ published tiers.
| Plan | Price | Credits |
|---|---|---|
| Free | Free | 100 |
| Freelance | $47 | 25,000 |
| Business | $497 | 500,000 |
| Enterprise | Custom | Custom |
- Most endpoints charge one credit per request, while some endpoints consume more.
- A single workflow can issue several requests, including paginated calls, so one skill run does not necessarily equal one credit.
- A cache hit costs zero credits when the request includes
cache_max_age, which accepts values from1dthrough30d. The response reportscredits_charged: 0. - Purchased credits do not expire.
A workflow’s effective cost depends on its endpoint mix, pagination depth, and cache reuse. Repeated dashboard queries against the same creators or posts can benefit substantially from cached responses.
Coverage stops at public data
Platform support varies, and the repository limits its claims to public data. Logged-in feeds, private accounts, and private endpoints fall outside the documented coverage.
- X: The vendor does not document a general search endpoint, which prevents broad keyword searches across the platform.
- Instagram: Hashtag discovery uses Google-indexed public posts instead of a native hashtag feed.
- Scan depth: Instagram discovery pagination stops at page 11, limiting the depth of trend scans.
The default workflows choose endpoint order, analysis steps, and output structure for the caller. Teams that need request-level cost tuning or different coverage can fork the MIT-licensed definitions, although replacing ScrapeCreators as the data provider would require changes to the shared API layer.
A pattern for vertical agents
The Agent Skills format remains new, and many published examples wrap a single task. ScrapeCreators uses a layered design: one skill handles transport and pagination, while specialized workflows define evidence requirements, analysis steps, and output contracts.
That architecture applies beyond social research. Developers building API-backed agents can use the same separation between a shared data layer and domain-specific workflows, then add citations, verbatim evidence rules, cost controls, and structured outputs at the workflow level.