Feature Request: Support model selection in agent tool

[superwg1984]

Feature Request: Support model selection in agent tool

Problem

         The `agent` tool only accepts a `prompt` parameter and doesn't allow selecting a different model for sub-

agents. This limits flexibility when using different models for different tasks.

Proposed Solution

                   Add an optional `model` parameter to the `agent` tool, similar to:

{
  "name": "agent",
  "parameters": {
    "properties": {
      "prompt": {"type": "string"},
      "model": {"type": "string"}  // Optional, falls back to default
    }
  }
}

This would allow calling sub-agents with specific models like:

• Different reasoning levels (fast vs deep)
• Different context window sizes
• Specialized models for specific tasks

### Workaround

Currently need to change global configuration or use MCP servers to achieve similar results, which is less convenient.