← All Recipes
Bazantic // Recipe
Vet Agent Before Paying
vet-agent-before-paying
Before transacting with an unknown autonomous agent, check its manipulation-resistant trust score and return a clear PAY / DO-NOT-PAY decision with reasoning. Uses AgentTrust, which composes The Graph's Agent0 (ERC-8004) subgraph with the Subgraph MCP to detect self-dealing and Sybil-ring reputation manipulation that a raw on-chain score misses.
Recipe inputs
Complete fields before a run.
Enter raw JSON for this Recipe.
Output example
Representative published output.
{
"verdict": "CAUTION",
"decision": "DO_NOT_PAY",
"rationale": "AgentTrust flagged this counterparty: all 57 feedback entries come from a single unknown wallet with no agent-economy footprint, none include proof-of-payment, and there are no third-party validations. Despite a naive score of 88, the manipulation-resistant score is 4. Do not transact without further diligence.",
"naiveScore": 88,
"trustScore": 4
}