[Feat]: Data labeling

### Is your feature request related to a problem? Please describe.

There is no standard way for agents to label their responses (text and other artifacts) to indicate whether it contains data that needs to be kept private (e.g., PII, NTK, user data, etc) or may alter the agent's behavior (e.g., untrusted data which may lead to prompt injection). Google workspace API, for example, can return metadata about documents and an agent could use that to populate such fields. This sort of metadata is useful to enforce policies on data flow (including propagating that information in longer-lived agentic memory) and possibly for logging too.

### Describe the solution you'd like

Given the the specs already support a metadata field , it may be useful to standardize a naming convention for the annotation, with the aim to make them interoperable across agents and to avoid name collisions. Support for custom labels would be useful too.
A simple idea could be to use the agent's domain as the key to the annotation: e.g., `company.com/a2a/annotations/` and define a list common keys under `a2a.dev/annotations/` like `a2a.dev/annotations/pii`

### Describe alternatives you've considered

_No response_

### Additional context

_No response_

### Code of Conduct

- [x] I agree to follow this project's Code of Conduct

The issue proposes a metadata labeling standard for agent responses. Start by reviewing the existing spec's metadata field structure, likely in protocol definition files. Look for existing annotation patterns or naming conventions. Determine how to define and integrate common keys like 'a2a.dev/annotations/pii' and support custom labels. Check for any related discussions or prior work in the repository.

Scritto dal modello di indicizzazione a partire dal testo della issue.