Skills & Hats
Understand how skills bundle tools with instructions, and how hats group skills into roles that control what AI agents can do.
Skills
Instead of giving an AI agent access to every tool on every MCP server, you package the right tools together into a skill with clear guidance. The agent discovers and uses that skill as a single, well-defined capability.
Why Skills Matter
- Organization — Group related tools under a meaningful name so agents (and humans) can find what they need.
- Reusability — Build a skill once and assign it to any number of hats. Update the skill, and every consumer gets the change.
- Governance — Control exactly which tools an agent can call. Grant specific skills with clear boundaries instead of broad access.
Skill Components
A skill can contain more than just tool references:
| Component | Format | Purpose |
|---|---|---|
| References | Markdown | Instruction documents, SOPs, domain knowledge |
| Scripts | Python, Bash, JS, TS | Executable code with Script SDK access |
| Assets | Any file (up to 10 MB) | Templates, images, PDFs, reference data |
Execution Modes
Every skill runs in one of three modes:
| Mode | Behavior | Best For |
|---|---|---|
| LIST (default) | Exposes all tools directly to the agent | Small tool sets (fewer than 10 tools) |
| OPTIMIZED | Agent searches for tools, then calls them by name | Large tool sets (10+), cost optimization |
| SMART | Delegates to a sub-agent with its own AI settings | Specialized domains, complex workflows |
See Skill Modes for a detailed guide.
AI-Powered Skill Generation
Describe what you want a skill to do in natural language, and Skilder suggests a name, description, and relevant tools from your installed MCP servers. Refine the result in the editor.
Hats
A hat is a role for your AI agent. When an agent "wears" a hat, it gets access to all the skills in that hat — and only those skills. Think of hats as job roles: "Marketing", "Engineering", "Customer Support."
Why Use Hats
- Organization — Group related skills under a descriptive role name instead of managing them individually.
- Access control — Assign hats to API keys. A support agent gets "Customer Support"; a dev agent gets "Engineering." Each sees only the skills it needs.
- Quick switching — Change an agent's entire capability set by switching its hat.
Examples
| Hat | Skills |
|---|---|
| Marketing | Social Media Posting, Content Drafting, Campaign Analytics |
| Engineering | Code Review, CI/CD Monitor, Incident Response |
| Customer Support | Ticket Triage, Account Lookup, Knowledge Base Search |
| Data Analysis | SQL Query Runner, Chart Generator, Report Formatter |
Organization Hierarchy
Hats are organized into teams that form a hierarchy (up to 10 levels deep). Model your org structure — e.g., Engineering > Backend > API Team. Manage this from the Organization Settings button on the Hats page.
Enable and Disable
Disable a hat to temporarily hide it from agents without deleting it. Disabled hats appear grayed out and can be re-enabled anytime.
How They Fit Together
API Key → Workspace → Hats → Skills → Tools- Install MCP servers to bring tools into your workspace.
- Create skills that bundle the right tools with instructions.
- Organize skills into hats that represent roles.
- Connect your agent via MCP — it discovers exactly the capabilities defined by its hat.
Next Steps
Create Your First Skill
Step-by-step guide from naming a skill to selecting tools and verifying it works.
Tutorial: Manually Compose a Skill
Full hands-on walkthrough — create a hat, write instructions, add tools, scripts, assets, and test end to end.
Skill Editor
The full-screen editor for code, instructions, assets, and settings.
Skill Modes
Understand LIST, OPTIMIZED, and SMART execution modes.
Writing Scripts
Use the Script SDK to call tools, spawn sub-agents, and chain workflows.
Assets & References
Attach instruction docs, binary files, and downloadable assets.
Organize Skills with Hats
Create hats, assign skills, and manage your organization hierarchy.