Added supportedInput and supportedOutput fields

[marco-agile]

📝 Summary

This pull request introduces two new optional fields to the AgentDescriptor schema:
• supportedInput: defines which input types the agent can handle.
• supportedOutput: defines which output types the agent can produce.

Both fields accept a checklist of the following values:
• text
• images
• video
• files

📌 Motivation

These additions aim to make agent capabilities more explicit and machine-readable, enabling better discoverability, filtering, and compatibility validation across agentic systems.

🔧 Changes
• Added supportedInput and supportedOutput as arrays of strings with fixed enums to components.schemas.AgentDescriptor.

✅ Example

supportedInput: ["text", "images"]
supportedOutput: ["text", "files"]

📚 Documentation

The OpenAPI specification has been updated accordingly in the AgentDescriptor schema block.

[tmnd1991]

The proposal makes sense, I'm thinking a bit broader about this, shouldn't we rely on HTTP Content-Types for this kind of stuff? So that we don't "reinvent" the wheel? What's your position?

[tmnd1991]

Suggested change

	- `supportedInput` *(array)* – List of supported output formats or content types (e.g., text, images, video, files).
	- `supportedInput` *(array)* – List of supported input formats or content types (e.g., text, images, video, files).

[tmnd1991]

can you declare this as a type and reuse for input and output fields?

[marco-agile]

I agree that using standard MIME types makes sense in principle.
That said, I think relying only on them might be too rigid, especially for agents that handle broad categories like video or images, where formats can vary and evolve.
A possible compromise could be to define a category field (e.g. video, text, image) and optionally list specific MIME types when needed. What do you think about it?