Creating skills

A skill is a reusable system prompt: a packaged instruction you can apply again and again instead of rewriting it every time. Where a prompt is often a single message, a skill is the standing behavior you want an AI to follow.

What goes into a skill

1. System prompt. The core instructions that define how the AI should behave, for example "You are a precise editor who tightens text without changing meaning." This is the heart of the skill.
2. Trigger description. A short description of when this skill applies. It helps you, and AI clients connected over MCP, understand when to reach for it.
3. Model. The AI model the skill is meant for, so the skill carries its intended setup with it.
4. Variables. Use {{variables}} for the parts that change per use, like {{text}} or {{audience}}. See Variables in prompts for how they work.

Create a skill

1. In the App, choose to create a new skill.
2. Write the system prompt describing the behavior you want.
3. Add a clear trigger description so it is obvious when to use the skill.
4. Pick the model.
5. Add any variables for the inputs that change each time.
6. Save, and decide whether to keep it private or share it publicly in the Community.

Why skills beat retyping

• Stop rewriting the same setup for the fortieth time.
• Keep your standards consistent across every use.
• Combine a stable system prompt with per-use variables for flexibility without losing structure.

Tips

• Keep the system prompt focused on one job; narrow skills are easier to reuse.
• Use the trigger description to capture the "when", not just the "what".
• Pull common inputs out into variables so the skill stays general.

What's next

• Master placeholders in Variables in prompts.
• Use your skill inside AI clients via Connect your AI client.
• Share a polished skill in the Community.