A single human-root keyholder. The lineage terminus: every privileged chain in OAS terminates at a human root. That is cryptographic ancestry under verifier policy — not legal personhood or Sybil resistance by itself.
did:oas:<ns>:hmr:<id>Root entity. Exempt from the lineage requirement.
A threshold group of human-root keyholders (t-of-n) acting collectively as a lineage terminus. FROST threshold signatures keep the group key indistinguishable from a single signer.
did:oas:<ns>:mhr:<id>Root entity. Exempt from the lineage requirement.
A stable corporate identity governed by a rotating MHR, providing organizational continuity across governance transitions. Added as the twelfth kind in spec v1.1.0.
did:oas:<ns>:enr:<id>Root entity. Verifiers additionally check its governing MHR.
A group of agents, tools, and workflows operating as a coherent unit, with its own identity distinct from any single member.
did:oas:<ns>:ao:<id>Lineage required. Chain terminates at a human root.
An autonomous software entity capable of independent decision-making and action. The primary operational kind: it performs tasks, invokes tools, and collaborates with other agents.
did:oas:<ns>:agent:<id>Lineage required. AgentLineageProof2025 at every link.
A specific running instance of an agent, identified by UUID. Separates the long-lived agent definition from each concurrent execution and its session keys.
did:oas:<ns>:agent:instance:<uuid>Lineage required. Inherits its chain through the parent agent.
A capability that agents invoke to interact with external systems or perform computations. Typed identity makes every invocation attributable.
did:oas:<ns>:tool:<id>Lineage required. Chain terminates at a human root.
A composable unit of agent behavior — a reusable capability package that can be attached to agents and governed independently.
did:oas:<ns>:skill:<id>Lineage required. Chain terminates at a human root.
A defined process involving one or more agents executing a sequence of steps. Gives multi-agent orchestration its own verifiable identity.
did:oas:<ns>:workflow:<id>Lineage required. Chain terminates at a human root.
A machine learning model with verified provenance. Identity-bound models make it possible to prove which model acted and where it came from.
did:oas:<ns>:model:<id>Lineage required. Chain terminates at a human root.
A data collection with verified provenance and lineage. Datasets carry the same accountability chain as the agents and models that consume them.
did:oas:<ns>:dataset:<id>Lineage required. Chain terminates at a human root.
An infrastructure service — compute, storage, networking — with verified identity, so even the substrate agents run on is attributable.
did:oas:<ns>:service:<id>Lineage required. Chain terminates at a human root.
Lineage in practice
The multi-generation example from the specification. Verifying the scraper at generation 3 means walking four lineage proofs back to Alice — from bundled evidence via typed authorizing verification.
Every humanRootChain must terminate at an hmr, mhr, or enr DID. Chains that loop or end at a non-root entity are invalid.
An entity's generation equals the length of its humanRootChain minus one, so trust policies can reason about distance from the root.
Every chain converges to exactly one human root — a single, unambiguous accountability anchor for every autonomous entity.