{"version":"2.1.30","anchor":"enhanced-custom-subagent-schema","canonical_anchor":"enhanced-custom-subagent-schema","heading":"Enhanced Custom Subagent Schema","tier":null,"area":null,"scope":null,"heads_up":null,"url":"https:\/\/changelogs.core-directive.com\/v\/2.1.30\/e\/enhanced-custom-subagent-schema","release_url":"https:\/\/changelogs.core-directive.com\/v\/2.1.30","markdown":"### Enhanced Custom Subagent Schema\n\nThe settings schema for defining custom subagents has been formalized with descriptive annotations on all fields. New configurable fields include:\n\n- `description`: Natural language description of when to use the agent\n- `tools`: Array of allowed tool names (inherits from parent if omitted)\n- `disallowedTools`: Array of tool names to explicitly disallow\n- `prompt`: The agent's system prompt\n- `model`: Model selection (`sonnet`, `opus`, `haiku`, or `inherit`)\n- `criticalSystemReminder_EXPERIMENTAL`: Critical reminder injected into system prompt\n- `skills`: Array of skill names to preload into the agent context\n- `maxTurns`: Maximum agentic turns before stopping\n\n**Evidence**\n\nSubagent definition schema (search for `\"Definition for a custom subagent that can be invoked via the Task tool\"`)"}