Runtype Skills Now Available in Skills SDK

You can now tell Claude Code, Cursor, Copilot, or any of 35+ AI coding agents to build something on Runtype and have it actually work.

npx skills add runtypelabs/skills

One command. Your agent learns Runtype — every surface type, every flow step, every SDK pattern, every CLI command. Then you describe what you want and it builds it correctly.

What this changes

Before skills, using an AI coding agent with Runtype was hit-or-miss. The agent could write code, but it didn't know the platform. It would guess at tool ID formats, skip validation steps, or wire things up in ways that looked right but broke at runtime.

Now the agent knows the platform the same way you do. It knows which surfaces to pick for your use case. It knows how to structure a product, configure an agent, embed a chat widget, or set up a scheduled workflow. It doesn't guess — it follows the same patterns an experienced Runtype developer would.

What's included

Six skills that cover every part of the platform:

Build products — Create agents, flows, tools, surfaces, and everything that connects them. The skill walks your agent through the right order of operations and validates as it goes.

Manage your account — Debug failing executions, inspect traces, check logs, and fix configuration issues. Read operations are unrestricted; mutations require confirmation.

Embed Persona — Set up the chat widget in your app. Handles all three integration paths (script tag, ESM, React), theming, client tokens, local browser tools, and fullscreen assistant layouts.

Create templates — Package your products as shareable templates with proper credential handling. Your agent catches common mistakes before they ship.

Use the SDK and CLI — TypeScript and Python SDK integration, CLI commands, and Marathon long-running tasks. All three SDK modes (stored, upsert-on-execute, virtual) with working examples.

Route automatically — An umbrella skill detects your intent and activates the right specialist. Ask about embedding a widget, you get the Persona skill. Ask about debugging, you get the admin skill. You don't manage this.

Works everywhere

Skills install through skills.sh, the open ecosystem for agent capabilities. That means they work in:

• Claude Code
• Cursor
• GitHub Copilot
• OpenAI Codex
• Windsurf
• Google Gemini
• Cline, AMP, Kiro, and 25+ more

Same skills, same knowledge, regardless of which agent you prefer.

Try it

Install:

npx skills add runtypelabs/skills

Or pick specific skills:

npx skills add runtypelabs/skills --list

Then open your agent and tell it what you want to build. "Set up a support agent with a Slack surface and a chat widget." "Create a flow that processes incoming webhooks and sends email summaries." "Debug why my scheduled agent isn't running."

The skills are open source and MIT licensed. If you find something missing, open a PR.