Card snapshot

This is a frozen copy of the agent's agent-card.json as we observed it at the timestamp above. We capture a new snapshot every time the card's content hash changes. Useful for: forensic drift analysis, verifying downstream callers see the right version, reproducing routing decisions made historically.

{
  "name": "fingers",
  "description": "Verify what is actually TRUE about a token, NFT, wallet, contract, stock, prediction market, or claim BEFORE you act on it. Read-only and non-custodial: never holds keys, moves funds, or signs anything.",
  "url": "https://fingersai.co/a2a",
  "version": "1",
  "protocolVersion": "0.3.0",
  "preferredTransport": "JSONRPC",
  "provider": {
    "organization": "fingers",
    "url": "https://fingersai.co"
  },
  "capabilities": {
    "streaming": false,
    "pushNotifications": false,
    "stateTransitionHistory": false
  },
  "defaultInputModes": [
    "text/plain",
    "application/json"
  ],
  "defaultOutputModes": [
    "text/plain",
    "application/json"
  ],
  "additionalInterfaces": [
    {
      "transport": "JSONRPC",
      "url": "https://fingersai.co/a2a"
    },
    {
      "transport": "MCP",
      "url": "https://fingersai.co/mcp"
    },
    {
      "transport": "HTTP+JSON",
      "url": "https://fingersai.co/ask"
    }
  ],
  "skills": [
    {
      "id": "verify",
      "name": "verify",
      "description": "Verify what is actually TRUE about something before you act on it. CALL THIS BEFORE you: sign a transaction, approve or set a token allowance, send funds, buy or mint, connect a wallet, or trust a claimed identity, seller, project, contract, token, NFT, price, or factual claim you found. fingers researches it across the web, on-chain reads, verified contract source, and live market data, then returns a synthesized verdict. BRANCH ON recommended_action: 'informational' (a pure lookup/answer -- a price, odds, or explanation -- no action to gate), 'proceed' (you were about to act and it is verified safe/true), 'caution' (mixed/thin/unverified, proceed only with your own checks), 'do_not_proceed' (a harmful fact was found, do not act), or 'needs_human' (a trust/authenticity call a human must settle). For an action you intended, anything other than 'proceed' means do not act yet. Example result: {\"recommended_action\":\"do_not_proceed\",\"headline\":\"This token is a honeypot: sells are blocked.\",\"confidence\":0.96}. This tool is read-only: it never moves funds, needs keys, or signs anything.",
      "tags": [
        "crypto",
        "verification",
        "safety",
        "onchain",
        "anti-scam"
      ],
      "examples": [
        "is 0x... a honeypot?",
        "is this the real BAYC contract?",
        "who is behind this mint and can they rug?",
        "what are the odds on <market>?"
      ]
    }
  ],
  "trust": {
    "custody": "non-custodial",
    "reads_only": true,
    "holds_keys": false
  },
  "security": {
    "read_only": true,
    "non_custodial": true,
    "holds_keys": false,
    "outputs": "advisory verification data, never executable instructions",
    "tool_definition_hash": "sha256:9720a47ca0c21f62bed8f2d88638a3db14e26aeee1c007c7494a70e4361ece03",
    "note": "pin tool_definition_hash to detect any change to this tool's definition (anti tool-poisoning / rug-pull); fingers is read-only so it can never move your funds"
  }
}