Skill Editor

The editor uses a three-panel layout similar to code editors like GitHub or VS Code, giving you a file tree, a code editor, and a settings panel all in one view.

Access it by clicking on a skill and opening it in edit mode, or by navigating directly to the editor URL for a specific skill.

---

Layout Overview

The editor has three panels:

---

File Tree (Left Panel)

The file tree shows the skill name as the root node with expandable sections for each component type:

• References -- Instruction documents in markdown
• Scripts -- Code files in Python, Bash, JavaScript, or TypeScript
• Assets -- Binary files like images, PDFs, or documents

Working with Files

• Click a file to open it in the center editor.
• Right-click a file to open the context menu with options to rename or delete.
• Add new components using the context menu or the add button at the section level.

When creating a new component, you provide a name and description. For scripts, you also select the language (Python, Bash, JavaScript, or TypeScript).

---

Code and Markdown Editor (Center Panel)

The center panel is a full-featured editor with:

• Syntax highlighting for all supported languages (Python, Bash, JavaScript, TypeScript, Markdown)
• Line numbers for easy reference
• Search and replace for finding and updating content within a file

Editing References

References use markdown. Write instructions, SOPs, or domain knowledge that guide the agent on how to use the skill's tools.

Editing Scripts

Scripts contain executable code. Each script has:

Viewing Assets

When you select an asset in the file tree, the center panel shows the asset details. Assets are binary files (images, documents, etc.) stored as base64-encoded content with a maximum size of 10 MB.

---

Settings Panel (Right Panel)

The settings panel is toggleable and contains skill-level configuration:

Skill Metadata

• Name -- The skill's display name (used by agents for discovery)
• Description -- What the skill does and when to use it
• Enabled/Disabled -- Toggle skill visibility to agents. Disabled skills are hidden from agents.

Tools

Select which MCP tools this skill includes. You can add tools from any installed MCP server in your workspace.

Execution Mode

Choose between LIST, OPTIMIZED, and SMART modes. For SMART mode, configure additional AI settings (model, temperature, max tokens, system prompt).

See Skill Modes for details on each mode.

Metadata

Optional fields for license information, compatibility notes, and custom JSON metadata.

---

Creating Sub-Components

Adding a Reference

1. In the file tree, use the add action in the References section.
2. Enter a name (e.g., "Research Guidelines") and description.
3. Write your instructions in the markdown editor.

Adding a Script

1. In the file tree, use the add action in the Scripts section.
2. Enter a name and description.
3. Select the script type: Python, Bash, JavaScript, or TypeScript.
4. Choose the execution target: AGENT or EDGE.
5. Write your code in the editor.

Adding an Asset

1. In the file tree, use the add action in the Assets section.
2. Enter a name and description.
3. Upload the file (up to 10 MB). Supported formats include images, PDFs, and other documents.

---

Saving

The skill editor saves changes in real time. A saving indicator in the editor shows the current save status so you know when your changes are persisted.

---

Next Steps

• Create Your First Skill -- Use the editor to build your first skill end to end.
• Skill Modes -- Configure the execution mode in the settings panel.
• Organize Skills with Hats -- Assign your skill to a hat so agents can access it.