Registration doesn't include AgentFacts - gap between agentfacts-format schema and adapter's register_with_registry()

[ojuschugh1]

I was reading through agent_bridge.py and noticed that register_with_registry() only sends three fields to the registry:

data = {
    "agent_id": agent_id,
    "agent_url": agent_url,
    "api_url": api_url
}

Meanwhile, projnanda/agentfacts-format defines a full schema with capabilities, skills, certifications, evaluations, and telemetry. But none of that gets sent during registration.

A few questions:

1. Is the plan to eventually have the adapter construct and submit an AgentFacts document during registration? Right now there's no way for the Index to know what an agent can actually do - it just knows where it lives.

2. The NANDA class takes an improvement_logic function but doesn't extract any metadata about its capabilities (input/output modes, latency, etc.) to include in registration. Would it make sense to add optional capability declarations to the NANDA() constructor?

   Something like:

      my_improvement,
      skills=[{"id": "translation", "inputModes": ["text"], "outputModes": ["text"]}]
  )

3. The agentfacts-format schema has an "evaluations" section with performanceScore, availability90d, and auditTrail. Is anyone working on populating these from actual runtime data? The /api/health endpoint already exists - seems like it could feed into availability metrics.

I work on AI agent verification tooling (behavioral auditing, drift detection) and have built tools in this space. Figured I'd ask before prototyping anything.