The settings schema for defining custom subagents has been formalized with descriptive annotations on all fields. New configurable fields include:
description: Natural language description of when to use the agenttools: Array of allowed tool names (inherits from parent if omitted)disallowedTools: Array of tool names to explicitly disallowprompt: The agent's system promptmodel: Model selection (sonnet,opus,haiku, orinherit)criticalSystemReminder_EXPERIMENTAL: Critical reminder injected into system promptskills: Array of skill names to preload into the agent contextmaxTurns: Maximum agentic turns before stopping
Evidence
Subagent definition schema (search for "Definition for a custom subagent that can be invoked via the Task tool")
Strings lifted out of the shipped bundle, so the claim above can be checked against them.