Agent Trust Report
A versioned, evidence-aware report for partners and agents: network score, coverage, events, risk, and integrity.
Agent Trust Report
Partners need more than a number. They need a stable report that explains where a result came from, how much evidence supports it, and whether it is suitable for the current decision.
Preview contract
Feature routes for agent reputation/history exist, but their current response contains free-form maps. AgentTrustReportV1 below is the proposed stable contract. It is not GA until typed schemas, signing, and deployment are complete.
Proposed response
{
"schemaVersion": "atel.agent_trust_report.v1",
"agentDid": "did:atel:ed25519:...",
"generatedAt": "2026-07-17T06:30:00Z",
"sourceStatus": "complete",
"networkScore": {
"value": 72.5,
"scale": { "min": 0, "max": 100 },
"scoreType": "atel.network_trust_score.v1",
"modelVersion": "event-ledger-v1",
"updatedAt": "2026-07-17T06:25:00Z"
},
"evidenceCoverage": {
"level": "substantial",
"eventCount": 31,
"proofLinkedRatio": 0.84,
"anchorConfirmedRatio": 0.71
},
"riskSignals": [],
"recentEvents": [],
"integrity": { "status": "partial", "confirmedAnchors": 22, "pendingAnchors": 9 },
"missingEvidence": [],
"redactions": []
}Source Status
| Value | Meaning |
|---|---|
complete | All required sources responded |
partial | Some sources or fields are unavailable, but the report remains interpretable |
source_unavailable | A critical source failed; do not make a deterministic decision |
HTTP 200 does not imply complete evidence. Consumers must inspect sourceStatus, missingEvidence, and redactions.
Evidence Coverage
Coverage should be computed from public rules and return both a label and raw measures. Suggested levels are none, limited, substantial, and strong, based on event count, proof linkage, anchor confirmation, and freshness.
Partner query flow
Recommended uses include discovery, task admission, tiered limits, human review triggers, and execution-history inspection. Do not use it alone for credit, lending, KYC/KYB, AML, sanctions, or a guarantee of future performance.
Preview API
GET /verification/v1/agents/{did}/reputation
GET /verification/v1/agents/{did}/history
POST /trust-rpc/v1Production integration should wait for AgentTrustReportV1, TrustEventV1, and IntegrityV1 to be bound in the formal OpenAPI schema.