Research · IEEE-format preprint
Open Agent Specification: a decentralized identity standard with cryptographic human accountability
The formal treatment of OAS: the did:oas method, the Agent Lineage protocol, threshold human roots, and the security analysis behind them.
Preprint · Submitted for peer review · 2026
Open Agent Specification (OAS): A Decentralized Identity Standard for Autonomous Entities with Cryptographic Human Accountability
Abstract
The rapid deployment of autonomous software agents—systems capable of independent decision-making, tool invocation, and inter-agent collaboration—has exposed a critical infrastructure gap: no universal, cryptographically verifiable mechanism exists for identifying autonomous entities or establishing who is responsible for their behavior. We present the Open Agent Specification (OAS), a decentralized identity standard designed against the W3C Decentralized Identifiers (DID) framework that introduces Agent Lineage—a cryptographic chain of agent lineage proofs linking every autonomous entity to one or more human roots. OAS defines a purpose-built DID method (did:oas), a taxonomy of autonomous entity types, graduated conformance levels, and a key derivation protocol (HKDF-SHA256 with Ed25519 signatures) verified through typed authorizing APIs and verifier trust policy. The specification is vendor-agnostic, framework-agnostic, and platform-agnostic: any party may implement OAS and issue conforming identities. We examine how OAS addresses three structural failures in current agent ecosystems—absent publisher identity, absent transitive reputation, and absent independent verification—and analyze lineage security under key compromise and replay threat models. Personhood, uniqueness, and Sybil resistance are out of band: they require external assurance credentials under verifier-owned policy. OAS provides the identity foundation required for accountable autonomous systems at scale.
Contents
Contributions
A purpose-built DID method with developer-registrable namespaces and a typed taxonomy of autonomous entities.
HKDF-SHA256 key derivation with Ed25519 signatures, verified through typed authorizing APIs and verifier trust anchors — structural checks never authorize privilege.
FROST (RFC 9591) threshold signatures distribute root accountability across t-of-n humans without changing verifier behavior. frost-ed25519 3.x requires share regeneration from 2.x.
The lineage protocol analyzed under key compromise and replay threat models. Personhood and Sybil resistance remain external assurance concerns.
Cite as
J. Rice Sr., "Open Agent Specification (OAS): A Decentralized Identity Standard for Autonomous Entities with Cryptographic Human Accountability," L1fe Labs, Inc., preprint, 2026.