sadd:launch-sub-agent

Automatic model selection for sub-agent tasks evaluates task complexity to assign optimal models like opus, sonnet, or haiku. It dispatches a sub-agent with the selected model, ensuring isolated context and traceable decision-making for development, research, or testing workflows.

The supervisor orchestrator pattern in multi-agent workflows isolates context and modularizes task execution. It launches specialized sub-agents for individual tasks, applying zero-shot Chain-of-Thought reasoning at the start and enforcing self-critique verification at the end.

To launch a sub-agent with zero-shot chain-of-thought reasoning, provide a task description to the sub-agent launcher. It automatically selects a model, applies CoT reasoning at the start, and enforces mandatory self-critique verification before completing the execution.

Multi-agent orchestration supports documentation, architecture, research, development, and testing tasks. It automatically selects the appropriate model and optional specialized agents to execute these tasks with deterministic sub-agent execution and self-critique verification.

The sub-agent framework enforces mandatory self-critique verification automatically as part of its workflow. This ensures deterministic sub-agent execution, validating the zero-shot CoT reasoning applied at the start of the dispatched task.

Automatic model selection in sub-agent orchestration is limited to choosing between opus, sonnet, and haiku models. Task complexity evaluation determines the model dispatch, confining execution to these specific configurations without supporting external or custom model integrations.