Five chapters. Thirty minutes. Go from a fresh machine to running domain-expert AI skills in Antigravity, Claude, Cursor, Windsurf, or Zed — with your own data sources connected.
Get the SkillOS Labs Inc command-line tool globally in under 30 seconds.
SkillOS Labs Inc ships as a single npm package. You need Node.js ≥ 18 and npm. One command installs the CLI and makes skillos available globally.
shell
$# Install globally
$npm install -g skillos-cli
$# npm prints installation details for your environment
shell
$# Prints the version of the installed CLI package
$skillos --version
No global install? You can use npx for a one-off run: npx -y skillos-cli --version. For day-to-day use, the global install is faster.
The MCP runtime is bundled separately. You don't need to install it manually — the skillos install command ( Chapter 3) handles that for you.
Single package
skillos-cli wraps the full CLI surface: install, serve, list, validate.
No registry account needed
Browse and activate public skills without signing up.
Model-agnostic
Works with MCP-compatible clients, including the clients shown in the setup examples.
2
Chapter 2
Browse the Registry
Discover and preview skills across verticals without signing up.
This chapter shows the current public skill list used by skillos list. Search the returned entries or open one to see its registry details.
current public registry snapshot
$skillos list
Fetching skills from the public registry …
$# Loading the current public catalog…
Search the loaded public registry snapshot:
Loading public skills…
Inspect any skill in detail with skillos inspect <slug>— shows the full SDS manifest, tier requirements, and tool surface before you activate anything.
3
Chapter 3
Activate in Your AI Client
One command configures Claude, Antigravity, Cursor, Windsurf, Zed, or any MCP client.
SkillOS now uses a remote API that requires authentication. First, get your API key from founderskillos.xyz/dashboard/api-keys, then set the SKILLOS_API_KEY environment variable. Pick your client below to see the correct configuration.
run in your terminal
$# Run this on the device where your AI client is installed
$skillos install
This page does not detect or configure software on your device. The CLI prints the clients it finds and reports each configuration update. Prefer manual setup? Select your client below and paste the JSON into its config file.
URL: https://founderskillos.xyz/api/mcpAuthentication: Sign in when neededProvider: SkillOS Labs Inc
After pasting, fully quit and reopen your AI client. MCP servers are loaded at startup — a simple window close is not enough.
1
Run skillos install
Or paste the JSON config manually for your chosen client.
2
Restart the AI client
Antigravity, Claude Desktop, Cursor, Windsurf, and Zed all load MCP servers at startup.
3
Open a new conversation
In your AI client (e.g. Antigravity MCP config, or Claude Settings → MCP Servers), verify "SkillOS Labs Inc" is listed.
4
Try a skill
Ask: "Activate a skill from the registry" or invoke @skillos in your IDE.
4
Chapter 4
Connect External Tools
Link Notion, GitHub, Jira and more so skills can act on real data.
Connectors give skills access to external data sources. The list and connection state below come from the current platform catalog and your signed-in account, when available.
live connector catalog
$skillos connectors list
Fetching the live connector catalog …
$# Loading current connector metadata and account state…
Loading connector catalog…
When credential storage is configured, SkillOS encrypts saved connector secrets before persisting them. A saved credential is not proof that the provider accepted it. The CLI's test command reports whether a provider check ran or only a format check was possible.
5
Chapter 5
Build with Skill Studio
Author, validate, and publish your own skill to the registry.
Skill Studio is the browser-based authoring environment. Skills are described in the Skill Definition Standard (SDS). The CLI checks the local file structure, required metadata, component content, declared tier boundaries, and configured risk signatures before submission.
1
Open Studio
Navigate to /studio and sign in. Existing contributors see their draft queue.
2
Create a new skill
Click "New Skill". Choose a vertical and tier. The editor pre-fills the SDS scaffold.
3
Fill in the SDS fields
Name, slug, description, tier (0–4), tools, prompt template, and test assertions.
4
Run validation
Run `skillos validate <file>`. It checks YAML/frontmatter syntax, required metadata, the 13 SDS components, declared-tool tier rules, and known risk signatures. It does not run model-behavior tests or certify safety.
5
Submit for review
Submissions enter the human review queue. Review status and notes are recorded; publication depends on the artifact type and the completed review path.
shell — CLI alternative
$# Run local structural and static checks; the report is generated from this file
$skillos validate my-skill.sds.yml
$# Use --json for a machine-readable report. Passing local checks is not publication approval.
Skill visibility and access depend on the current submission and account configuration. The platform license parameters are pending legal approval; check the component license files before relying on licensing terms.
SDS Validator
Local structure and declared-tier checks catch missing, duplicate, hollow, or invalid fields. Human review remains required.
Tier rubric
Clear criteria for Tier 0–4. Reviewers apply the same rubric publicly.
Instant publish for eligible submissions
After human approval, eligible new database-backed skills appear in the registry without a redeploy. Updates to packaged registry skills still require a reviewed source change and deployment.
