02-agent-modes

Agent Modes

Agent modes are pluggable execution strategies that augment how an agent processes and responds to tasks. Each mode adds specialized capabilities through additional behavior (system prompt) instructions and hidden skills, all managed transparently by the SDK.

Available Modes

Mode	Enum Value	Description
Default	TAgentMode.DEFAULT	No extra capabilities. The agent uses only the skills and behavior you provide.
Planner	TAgentMode.PLANNER	Adds systematic planning, task tracking, and progress reporting.
Worker	TAgentMode.WORKER	Adds background task delegation to "copy" agents while the main agent stays interactive.

Setting a Mode

Single Mode

import { Agent, TAgentMode } from '@smythos/sdk';

const agent = new Agent({
    name: 'Research Assistant',
    model: 'gpt-4o',
    behavior: 'You are a helpful research assistant.',
    mode: TAgentMode.PLANNER,
});

Combining Multiple Modes

Modes can be combined using an array. Each mode's capabilities are layered on top of each other:

const agent = new Agent({
    name: 'Research Assistant',
    model: 'gpt-4o',
    behavior: 'You are a helpful research assistant.',
    mode: [TAgentMode.PLANNER, TAgentMode.WORKER],
});

When combining Planner and Worker modes, the agent will plan complex tasks into steps (Planner) and dispatch heavy subtasks to background workers (Worker).

How Modes Work Under the Hood

When a mode is applied to an agent, two things happen:

1. Behavior (system prompt) extension: The mode appends instructions to the agent's behavior, telling the LLM how to use the mode's capabilities.
2. Hidden skills: The mode registers internal skills (prefixed with _sre_) that the LLM can call. These skills are not visible to end users but power the mode's functionality.

All mode-managed skills are automatically filtered out when creating copy agents (Worker mode) or exporting agent configurations.