msft-docs-agent returns empty responses for certain inputs

https://github.com/azure-ai-foundry/foundry-samples/tree/main/samples/microsoft/hosted-agents/python/msft-docs-agent

AI Agent Extension version: 0.1.2-preview

Steps to Reproduce:

1. Initialize the project: azd init --template Azure-Samples/ai-foundry-starter-basic
2. Initialize AI agent: azd ai agent init -m https://github.com/azure-ai-foundry/foundry-samples/blob/main/samples/microsoft/python/getting-started-agents/hosted-agents/msft-docs-agent/agent.yaml
3. Run azd up
4. Navigate to agent playground on the Foundry Portal

Scenario 1

Send the message "Hello, what can you do for me?".
Observed: it responds as expected ✅

Request:

{
  "agent": {
    "type": "agent_reference",
    "name": "msft-learn-mcp-agent",
    "version": "1"
  },
  "input": "Hello, what can you do for me?"
}

Response:

{
  "metadata": {},
  "temperature": 1,
  "top_p": 1,
  "user": "",
  "id": "caresp_06a0fb9ddbdd3770003HvHrjGmwDswt6e6MhlsYIG7BeK5tzWq",
  "object": "response",
  "status": "completed",
  "created_at": 1764697128,
  "error": null,
  "incomplete_details": null,
  "output": [
    {
      "type": "message",
      "id": "msg_J0DjTJ6xLsRrETr0TwITGYVbbcDenz8PMhKLqD0UkXlcnRYPuj",
      "status": "completed",
      "role": "assistant",
      "content": [
        {
          "type": "output_text",
          "text": "Hello! I can assist you with questions related to Microsoft documentation, including Azure, .NET, Microsoft 365, and other Microsoft technologies. I can help you find information, provide code samples, and explain concepts or features. Just let me know what you need help with!",
          "annotations": [],
          "logprobs": []
        }
      ]
    }
  ],
  "instructions": null,
  "parallel_tool_calls": true,
  "conversation": {
    "id": "conv_J0DjTJ6xLsRrETr0TwBYSwrhKcQqrBw6MljbAyilubfr1uSyk9"
  },
  "agent": {
    "type": "agent_id",
    "name": "msft-learn-mcp-agent",
    "version": "1"
  }
}

Scenario 2

Send the message "How do I create an HTTP-triggered Azure Function in Python?".
Observed: it returns an empty response ❌

Request:

{
  "agent": {
    "type": "agent_reference",
    "name": "msft-learn-mcp-agent",
    "version": "1"
  },
  "input": "How do I create an HTTP-triggered Azure Function in Python?"
}

Response:

{
  "metadata": {},
  "temperature": 1,
  "top_p": 1,
  "user": "",
  "id": "caresp_02162a4cf474a79e007jEJmwpLXI3kaRUhgSlbxhK9txlnCzBd",
  "object": "response",
  "status": "completed",
  "created_at": 1764695592,
  "error": null,
  "incomplete_details": null,
  "output": [],
  "instructions": null,
  "parallel_tool_calls": true,
  "conversation": {
    "id": "conv_oancQyAYqtsmMMZBn9KxEqWtVI8AnUDBMneFmb2tUcGaItxKA7"
  },
  "agent": {
    "type": "agent_id",
    "name": "msft-learn-mcp-agent",
    "version": "1"
  }
}

Start with the msft-docs-agent sample linked in the issue and reproduce the behavior using the listed azd init, azd ai agent init, and azd up steps. Compare the successful greeting request with the Azure Functions question and inspect the resulting agent response. Done means that the Python Azure Functions query returns a non-empty answer in the Foundry Portal.