Data source v1

Purpose

Callers routing between agents — human or model — need to know what data stands behind an agent before sending it a question. This extension is a small, declarative identity block for the agent's underlying dataset, carried on the AgentCard.

Declaration

{
  "uri": "https://standards.upperhandai.com/ext/data-source/v1",
  "required": false,
  "description": "Identity of the underlying dataset and its sampling frame.",
  "params": {
    "study": "Versant Mega Tracker",
    "vendor": "National Research Group",
    "universe": "U.S. Adults 18-74",
    "sampleSize": "n≈8,000 per wave, ~24,000 per rolling-quarter",
    "fielding": "Monthly since November 2025"
  }
}

params fields

Field Req. Meaning
study yes Dataset name. MUST match the study field of the agent's citation Parts
vendor no Who collects/produces the data
universe no The population the data represents
sampleSize no Human-readable sample description
fielding no Collection cadence and start

Additional string-valued fields are allowed; consumers MUST ignore fields they do not recognize.

Consumer requirements

A workbench SHOULD display this block on the agent's registry/detail page and MAY expose it to a routing model as part of the agent's description.

Conformance (probed)