AI OS Skills: 51 Marketing Skills for AI Agents (2026)

AI OS Skills Clawdbot is a set of 51 automation skills for AI Operating Systems, published by Kevin Badi in an open-source repo called ai-os-skills. The skills are designed to run on Clawdbot, an agent runtime. They cover content creation, video cloning, publishing, and reporting, so an agent can turn one idea into carousels, infographics, shorts, and scheduled posts.

  idea or YouTube URL
        |
        v
  +------------------+
  |  Clawdbot agent  |  reads skill folders
  +------------------+
        |
        +--> carousel-gen      --> carousel-publish
        +--> infographic-gen   --> infographic-publish
        +--> klap-generate-shorts / short-video / veo-video
        +--> youtube-to-heygen-video
        |
        v
  published posts + reports

The repo is Kevin’s own project. You can read the source at github.com/kevinbadi/ai-os-skills. This post walks the index, explains how it relates to Claude marketing skills, and shows where Creator OS fits once a skill produces a file that needs to go live.

What Are AI OS Skills and Why Clawdbot Powers Them

AI OS Skills are folders of instructions an AI agent loads on demand, and Clawdbot is the runtime that decides when to load each one.

An AI Operating System is the layer that gives a model memory, tools, and a schedule. Skills are the small, named units of work inside it. Each skill in the repo has a name, a short description, and the steps the agent follows: which tool to call, what to feed it, where to put the output. The repo lists 51 of them, grouped into content creation, video cloning, publishing, and the rest of the marketing loop.

Clawdbot matters because the skills are written for it. It is the agent runtime the collection is built on. If you run a different agent, the skill text still describes the workflow, but the wiring is yours to do. That distinction is worth holding onto: the repo is a library of marketing playbooks written in agent-readable form, not a SaaS product with a login.

Get started for $0

AI OS Skills Index: All 51 Marketing Skills for AI Agents

The index is organized by job, and the jobs map cleanly onto a marketing week.

Two things stand out. First, almost every creation skill has a matching publish skill. The repo assumes the agent finishes the job, not just drafts it. Second, the source material is usually a YouTube video. Six skills start from a YouTube URL and fan out into shorts, avatar video, or text posts. That is a realistic starting point for a creator who already has footage.

Content Creation Skills: Carousels, Infographics, Stories, and Shorts

The creation group turns one piece of source material into several formats, each with its own skill.

A typical chain looks like this. The agent pulls a transcript from a video. carousel-gen writes the slides and generates images. Then the agent hands the images to a publish skill. Meanwhile infographic-gen builds a single tall image for a different audience, and stories-gen cuts a vertical set for Instagram or Facebook Stories. nano-banana-image-gen fills gaps when a slide needs a custom visual.

For moving output, klap-generate-shorts cuts short-form clips from a YouTube video through Klap, short-video and shorts-creatorclaw handle short-form pipelines, and veo-video generates AI video with Google Veo 3. If you want a deeper look at how generated footage fits a real workflow, the post on AI video generation locally with coding agents is a good companion.

Use a prompt like this to run one chain:

Take this YouTube URL. Pull the transcript.
Write a 7 slide carousel aimed at solo founders.
Generate the slide images with carousel-gen.
Save the images to ./out and list every file you made.

Notice the last line. Asking the agent to list its output files is how you verify a skill actually ran, instead of trusting a summary.

Video Cloning and Short-Form Video Skills Explained

The video cloning group takes footage you already have and re-renders it with your face or an avatar, without reshooting.

The repo includes longform-video-clone-edit for long-form videos with a HeyGen avatar, short-form-video-clone-edit for short clips with face swap, and wan-video-clone for the Wan model path. On the conversion side, youtube-to-heygen-longform and youtube-to-heygen-video take a YouTube link and produce avatar video, while youtube-to-viral-posts writes text posts from the same source.

That gives you three outputs from one recording: the original video, an avatar version for a different account, and a set of text posts. The cloning skills are where the risk lives. A face swap or avatar render needs review before it ships. Keep a human checkpoint between render and publish.

If your edits are animation rather than cloning, Kevin’s other open-source repo, OPEN VIDEO EDIT, takes a talking-head clip and adds captions, logos and b-roll cards, rendering locally with Python and ffmpeg. The Fireship style editing writeup covers that pipeline.

Publishing and Social Distribution Skills for AI Agents

The publish skills exist so the agent closes the loop, and they are the thinnest part of the repo by design.

carousel-publish sends carousels to Instagram or LinkedIn. infographic-publish sends infographics to social platforms. Everything else, every other format and every other network, needs a destination. This is where Creator OS fits in as a fact, not a guess: Creator OS ships a hosted MCP server, a CLI, and a REST API, so an agent that already holds your API key can publish to 11 platforms plus Skool, reply to comments, and read analytics.

Concretely, a skill can drop files into a folder and a follow-up step can post them:

creatoros media:upload ./out/slide-1.png
creatoros validate:post-length --text "<carousel caption>"
creatoros posts:create --text "Five things I learned shipping v1" \
  --platforms instagram,linkedin \
  --media ./out/slide-1.png \
  --scheduledAt 2026-03-02T14:00:00Z --timezone America/Toronto

The MCP side works the same way through tools such as create_post and upload_media_from_url. If you want the longer version of that idea, read AI agent skills for social media automation and the MCP docs. There is also an agent Chrome relay post for cases where the agent has to act inside a browser.

Claude Marketing Skills vs AI OS Skills: What Is the Difference

Claude skills and AI OS Skills describe the same idea in two different runtimes, and the difference is the host, not the content.

Claude skills load into Claude and Claude Code. Creator OS ships 14 of them through npx @creatoros/cli@latest init, covering post-everywhere, schedule-posts, respond-to-comments, analytics, ads and more, and they appear in Claude Code’s skill menu. AI OS Skills load into an AI Operating System built on Clawdbot. Both are natural-language instructions an agent follows. Neither is a plugin with a UI.

A practical split: generate with the 51, publish with the Creator OS side. The Claude Opus 5.5 plus social agents post covers the model side of that arrangement.

How to Install and Run AI OS Skills on Your Own Agent OS

Installation is a clone and a prompt, because the repo is a folder of skill definitions.

1. Clone github.com/kevinbadi/ai-os-skills.
2. Point your Clawdbot-based agent at the skills directory, following the instructions your runtime gives for loading skills.
3. Check which external accounts each skill needs. The video skills expect HeyGen, Klap, Gemini or Veo credentials depending on the folder.
4. Decide your publish path. Either let the publish skills do it, or hand the output to Creator OS.

Then test one skill end to end before touching the others. Ask the agent to run carousel-gen on a single post, then confirm the files exist. A skill you have not watched run is a skill you cannot trust. Watch the walkthroughs on the KevBuildsApps YouTube channel to see the repo demonstrated rather than described.

AI Reporting Agents and SEO: Where These Skills Fit in 2026

Reporting is the part agents do cheaply, and it is the part most creators skip.

The creation group gets the attention, but the real gain comes from a loop that reads results back. Creator OS exposes per-post analytics, daily metrics, follower stats and best time to post through MCP tools such as get_post_analytics, get_daily_metrics and get_best_time_to_post, plus the analytics:posts and analytics:best-time CLI commands. An SEO agent can pair those with blog_create_article to write and schedule articles on a WordPress or Shopify blog, then check what earned clicks through links_stats.

A weekly prompt that works:

Pull analytics for the last 7 days. Rank posts by engagement.
List the best three posting times per platform.
Draft one follow-up post for each top performer and leave them as drafts.

That is a reporting agent doing research the creator would otherwise do by hand. The fleet of AI agents content workflow post goes further on coordination, and Social Agents is Kevin’s open-source harness that runs your socials on Creator OS and reports back on a local dashboard.

FAQ

What are AI OS Skills?

They are 51 automation skills for AI Operating Systems, collected in an open-source repo by Kevin Badi. Each skill is a named set of instructions an agent loads and follows. The collection is built for Clawdbot, an agent runtime.

How many marketing skills are in the AI OS Skills repo?

The README lists 51. They are grouped into Content Creation, Video Cloning, and publishing skills, with names like carousel-gen, veo-video and carousel-publish.

Are Claude skills for marketing the same as AI OS Skills?

No. They share the format, instructions an agent loads on demand, but they run in different hosts. Creator OS ships 14 Claude Code skills, while AI OS Skills run on a Clawdbot-based AI Operating System.

What are the best AI agents for SEO and reporting?

Pick an agent that can both write and read data, not just draft text. On the reporting side, Creator OS gives any MCP-capable app per-post analytics, daily metrics, best time to post and click stats through its hosted MCP server and CLI.

Can AI OS Skills publish carousels and shorts automatically?

Carousels and infographics have dedicated publish skills in the repo. Short-form video generation is covered by skills like klap-generate-shorts and short-video, but distribution to specific networks depends on what publishing tool your agent can reach.

Get started with Creator OS

The fastest path is to let the 51 skills generate and let Creator OS publish. Create an account at creatoros.ca/sign-up and connect your accounts. Creator OS plans start at $14.99/month for 11 platforms plus Skool, with the API, the MCP server and the CLI included.

Start with the docs to get your API key, then watch the launch video and the rest of the KevBuildsApps channel for hands-on builds.

Get started for $0
Free trial for new subscribers. Cancel anytime.