{
  "id": 20537248,
  "title": "Part 29: Agent cards in practice",
  "url": "https://urgent.news/2026/11/18/part-29-agent-cards-in-practice",
  "topic": "tech",
  "section": "Tech",
  "published": "2026-11-18T03:11:28.000Z",
  "source": {
    "name": "Dev.to",
    "slug": "dev-to",
    "url": "https://dev.to/thasha/part-29-agent-cards-in-practice-1oe9"
  },
  "original_language": "en",
  "account": "An agent card is a self-describing manifest for an agent, containing essential information such as its name, description, version, supported interfaces, capabilities, security schemes, security requirements, default input and output modes, skills, and optional elements like signatures and examples. Each field plays a crucial role in determining how the agent functions and interacts with others.\n\nFor the invoice exception agent, which runs on MuleSoft Agent Fabric, the card is concise and informative. It specifies that the agent is a read-only, capable of push notifications but not streaming. The security scheme ensures callers authenticate using their Entra agent identity, with a read scope limited to invoice status. The gateway URL points to Omni Gateway, abstracting the underlying Mule flow and process API. The only skill offered is to explain an invoice exception, clearly stating it never changes the invoice itself. No sensitive data or internal details are included, aligning with the specification's recommendation to avoid security risks.\n\nSimilarly, the supplier-communication agent on Azure Foundry advertises a skill to draft a reply to the supplier once a fix is approved, but the actual sending of the email is handled internally within the supplier agent's process. This separation of concerns prevents the card from suggesting actions that occur outside the agent's control, adhering to the rule that no write skills should be included on a card intended for external calls. Both cards exemplify best practices in designing agent cards, emphasizing clarity, security, and minimalism to facilitate effective communication between agents while safeguarding sensitive information.",
  "summary": "Part 29 · One Agent, Five Platforms An agent card is a CV for an agent. It tells other agents who it is, what it can do, where to reach it and how to prove who's calling. And like a CV, what you leave off matters as much as what you put on. Part 28 had the supplier-communication agent on Azure read the invoice agent's card before asking about invoice 4471. This part writes both cards. It's based…",
  "key_points": [
    "Agent card contains agent's name, description, version and capabilities.",
    "Invoice exception agent is read-only, uses Entra agent identity for authentication.",
    "Supplier-communication agent drafts reply skill, but doesn't send emails."
  ],
  "editors_take": "Standardizing agent cards through minimal, self-describing manifests enables secure interoperability by clearly defining capabilities and boundaries while preventing agents from overpromising actions beyond their technical control.",
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}