Awesome Social Media Skills

A curated collection of AI agent skills, workflows, prompts, tools, and
integrations for researching, creating, repurposing, publishing, and
growing social-media content.

Build a social-media agent that can go from idea → research → content →
approval → publishing → analytics.

Idea
  ↓
Research
  ↓
Content strategy
  ↓
Create content
  ↓
Review and approve
  ↓
Publish or schedule
  ↓
Analyze performance
  ↓
Improve future content

Jump to: Why this exists · Quick start · Browse by workflow · Browse by platform · Featured skills · Agent workflows · Skill categories · Agent compatibility · Add your skill · Publish with ReplyNodes · Community

Why this exists

Most AI agent skill collections are scattered across gists, prompt
libraries, and one-off repos, and most social-media prompt content is
generic ("write a viral post") rather than structured enough for an agent
to execute reliably. This repository is a single, portable place for
skills that cover the whole social-media content lifecycle — not just
"write a caption" — in a consistent format any agent can read.

Every skill follows the same structure (templates/SKILL.md):
clear purpose, defined inputs, specific instructions, a defined output
format, a real example, and explicit quality checks. That consistency is
what makes skills portable across Claude Code, Codex, OpenClaw, and other
agent runtimes, and MCP-based tool chains.

What can AI social-media agents do?

An agent built from these skills can move through the full content
lifecycle, closing the loop back into research so each cycle improves on
the last:

flowchart LR
    R[Research] --> S[Content strategy]
    S --> C[Create content]
    C --> V[Review & approve]
    V --> P[Publish or schedule]
    P --> A[Analyze performance]
    A -.-> R

Individually, agents built on these skills can: extract a brand voice from
existing content, research real audience pain points, turn one video or
article into a week of platform-native posts, draft content and hand it to
a human for approval, and analyze what published to inform what to make
next. See Complete agent workflows for full
walkthroughs.

Quick start

1. Browse a skill category below, or search skills/ for
   something specific.
2. Copy the skill's markdown into your agent's system prompt, skill
   directory, or tool description — see Agent compatibility
   for runtime-specific notes.
3. Supply the inputs the skill lists, and let your agent follow its
   instructions.
4. Chain skills into a workflow — see Complete agent workflows
   for tested sequences, or docs/use-cases.md for more.

Browse by workflow

Browse by platform

Skills tagged for each platform, via their frontmatter platforms field:

Cross-platform formatting and queuing for all of the above happens in
Cross-Platform Publishing Formatter.

Featured skills

Skills that are broadly useful as a starting point, regardless of what
you're building:

• Brand Voice Extractor — turns sample content into a reusable voice profile every other writing skill can use
• Audience Pain-Point Researcher — grounds content decisions in real signal instead of guesses
• YouTube-to-Social Repurposer — the highest-leverage repurposing skill for anyone already making video
• Content Quality Reviewer — the approval gate every workflow in this repo routes through

Complete agent workflows

Full skill sequences with walkthroughs, in examples/:

• Content pipeline — draft, human approval, publish
• YouTube to social — one video into a week of posts
• Blog to social — one article into a thread, post, and carousel
• Agent content calendar — research a niche into a scheduled calendar

More combinations, including product launches and performance-driven
iteration, are in docs/use-cases.md.

What can you build?

• An agent that turns one YouTube video into a week of social posts
• An agent that researches a niche and creates a content calendar
• An agent that turns product updates into platform-specific content
• An agent that maintains a consistent brand voice across contributors
• An agent that drafts content, requests human approval, and schedules publishing
• An agent that analyzes performance and improves future content

None of these are guarantees of a specific outcome — they're patterns for
assembling skills into a working pipeline. What each pipeline produces
depends on the inputs and source material you give it.

Skill categories

26 skills today. See docs/quality-standards.md
for what Featured, Verified, and Community labels mean as this
grows.

A machine-readable index.json is generated from every
skill's frontmatter (via scripts/generate_index.py)
and validated in CI — useful if you're building search, a website, or
tooling on top of this repository instead of browsing it by hand.

Agent compatibility

Skills are plain markdown with structured frontmatter — not tied to one
agent runtime. Compatibility labels describe whether a skill's format is
ready to use as-is in a given environment; they are not claims of
official support or endorsement by that agent's maintainers.

Full details: docs/agent-compatibility.md.

Add your skill

1. Copy templates/SKILL.md into skills/<category>/.
2. Follow docs/skill-format.md for what belongs in
   each section, and docs/submission-guide.md
   for the full walkthrough.
3. Add it to that category's README table.
4. Open a pull request using the template — see
   CONTRIBUTING.md for acceptance criteria and the
   disclosure requirement for commercial submissions.

Also welcome: workflows, prompts, MCP servers, tools, and integrations —
see CONTRIBUTING.md for how to submit each.

Publish with ReplyNodes

These skills help agents research and create content. Publishing —
actually posting or scheduling it — is a separate concern, kept
intentionally decoupled so any scheduling tool or MCP-compatible publishing
layer can sit downstream of Cross-Platform Publishing Formatter.

ReplyNodes is one such layer: it provides
publishing and scheduling for agents through APIs and an
MCP server
(@replynodes/mcp on npm),
so an agent that has drafted and had content approved using the skills
above can hand it off to be published or scheduled without a manual
copy-paste step.

Research
  ↓
Create content
  ↓
Review
  ↓
Approve
  ↓
Publish or schedule with ReplyNodes
  ↓
Analyze results

ReplyNodes is optional. Every skill and workflow in this repository is
useful without it — this section exists because publishing is a real step
in the lifecycle this repo covers, and ReplyNodes is the specific tool we
build and can speak to directly.

See docs/mcp-quickstart.md for a technical
walkthrough — setup, available tools, and where it fits relative to the
publishing skills above.

Community

Featured Skill: Brand Voice Extractor
— the skill most other writing skills in this repo depend on.

Featured Workflow: YouTube to social
— the highest-leverage starting point if you already publish video.

Skill of the Week: rotated manually by maintainers; see open
issues for the current pick and nominate one by opening an
issue.

Contributing: see CONTRIBUTING.md — most submissions
are a single pull request adding one file plus a table row. First-time
contributors are welcome; issues labeled good-first-issue (once tagged)
are a good place to start.

Questions or ideas: open a discussion or issue.

This repository does not publish fabricated stars, testimonials, or
adoption numbers, and doesn't intend to start.

License

MIT — skills, docs, and templates in this repository are free
to use, adapt, and redistribute. Individual submitted tools and
integrations may carry their own license; check the entry before assuming
MIT applies to a linked third-party project.