Skill Modes

Every skill runs in one of three execution modes that control how the consuming agent interacts with the skill's tools. You set the mode in the skill editor's settings panel.

---

LIST Mode (Default)

LIST mode exposes every tool in the skill directly to the agent. The agent sees all available tools and calls them as needed based on its own reasoning.

How It Works

When an agent loads a skill in LIST mode, it receives the full list of tools with their names, descriptions, and input schemas. The agent decides which tools to call, in what order, and with what parameters.

When to Use LIST Mode

• The skill contains a small number of tools (roughly 1--10).
• The agent needs full visibility into all available tools.
• You want the agent to have maximum flexibility in how it combines tools.
• Token usage is not a concern.

Example

A "Customer Lookup" skill with 3 tools -- search_customers, get_customer_details, get_order_history -- works well in LIST mode. The agent sees all three tools and uses them as the conversation requires.

---

OPTIMIZED Mode

OPTIMIZED mode reduces token usage by replacing the full tool list with two meta-tools: Search Tool and Call Tool. The agent first searches for relevant tools, then calls them by name.

How It Works

Instead of receiving all tool definitions upfront, the agent gets two tools:

1. Search Tool -- The agent describes what it needs, and the system returns matching tools from the skill.
2. Call Tool -- The agent calls a specific tool by name with the required parameters.

This two-step process avoids sending large tool lists in every request, significantly reducing token consumption.

When to Use OPTIMIZED Mode

• The skill contains many tools (10+).
• You want to reduce token usage and cost.
• The agent does not need to see all tools at once.
• Tools have clear, descriptive names that make search effective.

Example

A "DevOps Toolkit" skill with 30+ tools from multiple MCP servers benefits from OPTIMIZED mode. The agent searches for "deploy" and gets back the 3 relevant deployment tools instead of receiving all 30 tool definitions.

---

SMART Mode

SMART mode delegates tool execution to a sub-agent with its own AI configuration. Instead of the consuming agent calling tools directly, it delegates the entire task to a specialized agent that handles tool orchestration.

How It Works

When an agent triggers a skill in SMART mode:

1. The consuming agent sends the task to the skill.
2. A sub-agent is spun up with the skill's configured AI settings.
3. The sub-agent uses the skill's tools to complete the task.
4. The result is returned to the consuming agent.

Configurable AI Settings

SMART mode lets you customize the sub-agent's behavior:

When to Use SMART Mode

• You want a specialized agent optimized for a specific domain.
• The task requires multi-step tool orchestration that benefits from a focused system prompt.
• You want to isolate the skill's execution context from the consuming agent.
• You need different AI settings (model, temperature) for different skills.

Example

A "Financial Analysis" skill in SMART mode uses a sub-agent configured with a low temperature (0.1) and a system prompt emphasizing accuracy and numerical precision. The consuming agent simply says "analyze Q3 revenue trends" and receives a complete analysis back, without managing the individual tool calls itself.

---

Mode Comparison

---

Decision Guide

Use this table to pick the right mode:

You can change a skill's mode at any time in the skill editor without affecting the skill's tools, instructions, or other components.

---

Next Steps

• Create Your First Skill -- Build a skill and configure its mode.
• Skill Editor -- Learn where to set execution modes and other settings.