For publishers

Publish a Skill, get paid for it

Publish your Skill once — no plugin API to port, no per-editor build. Then share it with the people who need it, set your own price, and get paid when they buy it.

What review covers

Every listing is checked as a package, not as a landing page. The four things that have to hold before a Skill appears in the catalogue:

The package

5 checks

  • A manifest that validates, with a version and a declared output
  • A SKILL.md your agent can follow without you in the room
  • Previews rendered from work the Skill actually produced
  • No secrets, and no network call the manifest doesn't declare
  • A protection declaration confirming server-side execution and creator ownership

your-skill/

manifest
output and version
SKILL.md
the workflow
templates · scripts
only what the work needs

Nothing is compiled for an editor.

The folder you upload is the folder every agent runs.

Three ways to publish

Pick the route that matches how you already work. All three end at the same review queue.

01

Web upload

Drop a folder or a .zip into the browser. We validate the manifest, render the preview and publish it under your namespace.

02

Agent publish

Point your agent at the repository and ask it to publish. It writes the manifest from the Skill you already have and opens the listing for review.

03

CLI publish

npx skillhub publish — for the pipeline case. Auth by token, one command, exits non-zero when the manifest fails validation.

How your Skill is protected

Every Skill uploaded to Skillry runs server-side in a sandboxed runtime. The real package — SKILL.md, prompts, tool definitions, and model keys — never ships to the client. End users send inputs and receive only the finished result.

  • Skill source is encrypted at rest and decrypted only inside the runtime.
  • Model API keys stay in a secret vault; users never see them.
  • Rate limits and audit logs protect against scraping and abuse.
  • You keep ownership; Skillry acts as the execution layer, not the owner.

Read the full architecture and call flow →

Upload your Skill

Upload a .md or .zip package, attach optional preview images, and submit it to the review queue.

You'll be prompted to sign in if not logged in

My submissions

Your submitted Skills and their review status.

How a Skill is made here

The catalogue is in-house, and the process shows what a good package contains. It is also the bar a new listing is measured against.

01

Design the result

We art-direct the finished output first: layout, type, color and motion.

02

Write the workflow

Original instructions, templates and scripts, packaged as one SKILL.md.

03

Run it for real

Each Skill page shows a real result it made.

04

Keep it current

The same team updates each Skill as agents and models change.

Ready to extend your agent?

Connect once, then install any Skill from our studio straight from your agent.

Browse the catalogue

Try it now

Describe the outcome. We will find the Skill.

Pick an example or write your own. Enter or press the arrow to open it in the conversation platform, with this Skill already selected.

NeoSpark catalogueBest fit from 7 Skills
Click a suggestion to start

Start from an example

What happens when you paste it

  1. 1Your agent connectsIt reads the install guide and wires up the official NeoSpark CLI.
  2. 2It picks the SkillIt matches your sentence to the catalogue and installs the closest fit.
  3. 3You get the workThe Skill runs and returns the deliverable you described.

Works in Claude Code, Cursor, Codex and any agent that reads SKILL.md. Install guide

Prompt ready

Paste this into your agent

It installs the Skill first, then asks for the work you described. Works in Claude Code, Cursor or any agent that reads SKILL.md.

Best-fit Skill, chosen from the catalogue

0 characters · this page never sends anything; the prompt is yours to paste.

Install guide