{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreibslyvcg3mifkohld6c4uwsasasjcumbjgxz5rrfa4fdl64c6t3xq",
    "uri": "at://did:plc:lk3jfj3zq4k4wxnk474axylu/app.bsky.feed.post/3mm6pzygk3p62"
  },
  "path": "/t/proposal-a-cryptographic-identity-infrastructure-for-autonomous-ai-agents-and-robotics/1367783#post_7",
  "publishedAt": "2026-05-19T05:02:36.000Z",
  "site": "https://community.openai.com",
  "textContent": "The part I keep seeing in day-to-day agent use is less agent identity as a full protocol and more the immediate credential boundary: an agent needs to complete a login, but the secret should never enter the model context or trace.\n\nI’m building a small local tool around that practical layer: AgentPass Local stores credentials in the OS secret store, checks allowed domains before use, and returns only the outcome to the agent. It’s not a replacement for agent identity standards, but it covers the painful moment where a workflow hits a login screen.\n\nIf interested, check agentpass(.)romeoapps(.)com",
  "title": "Proposal: A Cryptographic Identity Infrastructure for Autonomous AI Agents and Robotics"
}