Introducing Awesome Social Media Skills, an Open-Source Skill Library for AI Agents
Most "AI for social media" resources we've come across are lists of prompts — "write a viral post," "generate 10 caption ideas." Useful once, but not something an agent can execute reliably, and not something that composes into an actual pipeline.
So we open-sourced Awesome Social Media Skills: a curated, MIT-licensed collection of AI agent skills covering the full content lifecycle — research, content strategy, writing, video, images, repurposing, publishing, and analytics.
What's actually in it
26 skills across 9 categories, each following the same format: a clear purpose, defined inputs, specific step-by-step instructions, an exact output shape, a real example, and explicit quality checks — including what the skill must never claim (no promises of guaranteed engagement or virality, anywhere in the repo).
That consistency is the point. A skill written this way can be read and executed by Claude Code, Codex, OpenClaw, or Hermes without rewriting it per agent — the same idea we cover in Agent-native vs. API-first: what's actually different.
A few examples of what it covers:
- Turning one long-form video into a week of platform-native posts, without losing accuracy on the claims made in it
- Extracting a reusable brand voice from existing content so multiple writers or agents stay consistent
- A human-in-the-loop approval gate before anything publishes — see Approval workflows for AI-generated social content for the thinking behind why that gate matters
- Closing the loop from performance data back into what to create next
Where ReplyNodes fits — and where it doesn't
The repository is fully useful without ReplyNodes. Every skill produces markdown you can hand to any agent's context directly.
Publishing is the one step we didn't try to solve generically, because
that's the part we actually build. Skills route to a
Cross-Platform Publishing Formatter
that produces a structured, platform-tagged queue — and from there, any
MCP-compatible publishing layer can pick it up. We wrote up how ours does,
using the real tool names and config from
@replynodes/mcp, in the
repo's MCP quickstart.
If you'd rather understand the protocol first, start with
What is MCP, and why does social publishing need it?
It's meant to grow
26 skills is a starting point, not a finished catalog. The repo has a contribution guide, a portable skill template, and a quality-label system (Featured / Verified / Community) so it can grow past what we write ourselves. If you build agent skills for social content — or find a gap this doesn't cover yet — pull requests are open.