This specification defines the credential schemas used by the did:trustagent method for expressing agent capabilities, delegation relationships, and trust scores as W3C Verifiable Credentials v2.0.
Issued by a parent entity to delegate capabilities to a child agent. Uses the https://www.w3.org/ns/credentials/v2 and https://trustdid.org/ns/agent/v1 contexts, with type ["VerifiableCredential", "AgentDelegationCredential"].
{
"@context": ["https://www.w3.org/ns/credentials/v2", "https://trustdid.org/ns/agent/v1"],
"type": ["VerifiableCredential", "AgentDelegationCredential"],
"issuer": "did:trustagent:evm:137:org:z6MkParent...",
"validFrom": "2026-04-01T00:00:00Z",
"validUntil": "2027-04-01T00:00:00Z",
"credentialSubject": {
"id": "did:trustagent:evm:137:delegated:z6MkChild...",
"capabilities": [...],
"delegationDepth": 2,
"maxDepth": 5,
"delegationType": "capability"
},
"credentialSchema": {
"id": "https://trustdid.org/schemas/delegation/v1",
"type": "JsonSchema"
}
}Required Claims:
| Claim | Type | Description |
|---|---|---|
capabilities | string[] | Capabilities granted in compact format |
delegationDepth | integer | Current depth in delegation chain (0 = root) |
maxDepth | integer | Maximum allowed delegation depth (1-10) |
Optional Claims:
| Claim | Type | Description |
|---|---|---|
delegationType | string | Type of delegation: "capability", "identity", "full" |
Issued to an agent to grant specific operational capabilities. Contains the agent's DID as subject, with an array of capabilities and a grantedAt timestamp.
{
"@context": ["https://www.w3.org/ns/credentials/v2", "https://trustdid.org/ns/agent/v1"],
"type": ["VerifiableCredential", "AgentCapabilityCredential"],
"issuer": "did:trustagent:evm:137:org:z6MkOrg...",
"credentialSubject": {
"id": "did:trustagent:evm:137:autonomous:z6MkAgent...",
"capabilities": ["data:read:scope=public", "api:call:rate=100"],
"grantedAt": "2026-04-01T00:00:00Z"
}
}Issued by a trust registry to attest an agent's trust score. Contains a numeric score (0-10000), endorsement count, behavior hash, and evaluation timestamp.
{
"@context": ["https://www.w3.org/ns/credentials/v2", "https://trustdid.org/ns/agent/v1"],
"type": ["VerifiableCredential", "AgentTrustScoreCredential"],
"issuer": "did:trustagent:evm:137:org:z6MkTrustRegistry...",
"credentialSubject": {
"id": "did:trustagent:evm:137:autonomous:z6MkAgent...",
"trustScore": 8500,
"endorsements": 42,
"behaviorHash": "z6Mk...",
"evaluatedAt": "2026-04-01T00:00:00Z"
}
}Issued to attest an agent's identity attributes including agent class, lifecycle state, model identifier, and version.
{
"@context": ["https://www.w3.org/ns/credentials/v2", "https://trustdid.org/ns/agent/v1"],
"type": ["VerifiableCredential", "AgentIdentityCredential"],
"issuer": "did:trustagent:evm:137:org:z6MkOrg...",
"credentialSubject": {
"id": "did:trustagent:evm:137:autonomous:z6MkAgent...",
"agentClass": "autonomous",
"lifecycle": "active",
"model": "claude-4",
"version": "1.0.0"
}
}Capabilities use compact notation:
resource:action[:constraint=value[,constraint=value]]| Capability | Description |
|---|---|
credential:issue | Issue verifiable credentials |
credential:verify | Verify credentials |
credential:revoke | Revoke credentials |
payment:authorize | Authorize payments |
payment:authorize:limit=N | Authorize payments up to N |
data:read | Read data |
data:read:scope=public | Read only public data |
data:write | Write data |
api:call | Make API calls |
api:call:rate=N | Make up to N API calls per minute |
agent:delegate | Delegate to sub-agents |
agent:delegate:maxDepth=N | Delegate up to depth N |
A child agent's capabilities MUST be a subset of its parent's capabilities. Constraints on child capabilities MUST be equal to or stricter than the parent's constraints.
Example: If parent has payment:authorize:limit=10000, child can have payment:authorize:limit=5000 but NOT payment:authorize:limit=20000.
All credential types support SD-JWT (RFC 9901) for selective disclosure. Agents can present credentials while revealing only the claims needed for a specific interaction.
For example, an agent presenting capabilities to a verifier can disclose only the relevant capability without revealing its full delegation chain or trust score.