What is happening?
When I attach an image to the Gemini Agent chat in PyCharm, the model fails to analyze it. Instead of describing the image, the chat window outputs its own internal system prompt, starting with , , , etc.
The model correctly identifies that an image was attached but seems to glitch during processing.
Plugin/IDE Version:
• Gemini Code Assist Plugin: 1.36.0-252
• IDE: PyCharm Community Edition (Latest 2025.x version)
• OS: Windows 11 (running via Parallels on a Mac)
Google Cloud Project Setup:
• I am connected to a fresh Google Cloud Project (e.g., MyPycharmProjectParallels).
• The project is successfully linked to an active Billing Account.
• The "Vertex AI API" is Enabled.
• The "Gemini Google Cloud API" (Cloud AI Companion) is Enabled.
Steps to Reproduce the Bug:

1. In a PyCharm project, open the "Gemini" tool window and select the "Agent" tab.
2. Drag and drop a local image file (e.g., a .jpeg) into the chat input. The image attaches successfully and displays a thumbnail.
3. Type a prompt to analyze the image, for example: "describe this image in one sentence"
4. Send the prompt.
   Expected Result:
   The Gemini Agent should analyze the attached image and respond with a description (e.g., "This is an image of a login form on a website.").
   Actual Result:
   The Gemini Agent fails to analyze the image and outputs its own internal system prompt. The response looks like this: Claro, aquí tienes una descripción en una sola frase:
   El texto es un conjunto de instrucciones detalladas que definen la personalidad, el objetivo y las reglas de formato para un asistente de inteligencia artificial especializado en programación y código.
   (This text is a description of the system prompt itself, not my image. It sometimes also outputs the full XML, like ... etc.)