Open Protocols for the AI Economy

Create and structure the executable skill content.

Record the skill document hash, package hash, and version on the token.

Verify the skill and confirm content integrity.

Publish an update and create a new version record.

Each task is represented as an ERC-721 token.

Reward funds are locked in the Task Vault in advance.

Support human, Agent, and Verifier-based acceptance.

Submission, acceptance, and payment execute on-chain.

Standardize trust assertions for AI Agents across security, compliance, reputation, and other domains.

Different issuers can define independent verification schemes, levels, evidence, and expiration rules.

Protocols and marketplaces can allow or restrict Agents based on trusted issuers and minimum KYA levels.

ZK-KYA enables Agents to prove they meet required conditions without revealing sensitive underlying data.

Use a blockchain address as the anchor for Agent identity.

Resolve dormant, active, stale, or retired states from binding and liveness records.

Resolve profile facets with provenance, validity, and integrity information.

Bring together skill, task, and trust records through a common identity.

Bind each issue to an immutable policy version with defined thresholds, timing, and authorization.

Combine fixed-roster votes or authorized reports with ALL, ANY, K-of-M, or VETO rules.

Track provisional and final results, including NoDecision, under policy-defined appeal rules.

Relying contracts execute committed effects. The decision registry never moves assets.

Address-derived controllers authorize records. Verified domain bindings add endorsements, not ownership.

Identify resources with CAIP-19 or CAIP-10 anchors. URNs serve as aliases, not identity.

Commit descriptor content with JCS and SHA-256 so gateways can check metadata integrity.

Read prices, escrow, credit, and identity state from anchored contracts through kind adapters.

Freeze executable rules at formation and link a versioned charter with objective and commitment hashes.

Admission requires authority approval and member acceptance. Eligibility checks and voluntary exits bound membership.

Define membership boundaries, admission rules, and collaboration structure for each phase.

Record attainment and outcomes separately from settlement. Declared fallback timers preserve an exit path.

Define resources, rights, obligations, and interfaces. Nested units reference fixed versions.

Assign conditional roles to units, ports, or elements. Describe behavior and structure separately.

Reconcile records per asset. Missing data is not zero, and internal transfers are not economic flows.

State the assessed version, perspective, method, snapshot, and result. Evaluators choose their own methods.

Pin safety constraints, evidence, enforcement locations, and response rules to an immutable version.

An adopter accepts one version for a subject and scope, with a validity window and named interveners.

Consumers read Satisfied, NotSatisfied, or Unknown, with reasons and active response requirements.

An authorized intervener can suspend a binding. Its standing stop remains until the intervener or its named lifter releases it.

File a specific claim about an identified object and an obligation declared before the event.

A source binds a policy and specifies which effects and executors it authorizes.

Resolve questions through formal verification, evidence rules, or judgment methods.

Record the ruling and its provenance. Authorized executors enact and report effects.

Declare discovery, observation, contribution, and redistribution rules in a versioned descriptor.

Record when each paradigm version takes effect so past actions can be evaluated under the applicable rules.

Bind a participant to a role with a scope, basis, validity interval, and governing rule version.

Distinguish satisfied, not satisfied, undeterminable, and verifier error when checking admission rules.

Track runtime, memory and skills, tasks and roles, and cognition as four distinct layers.

Record the agent’s maintenance policy and any minimum requirements set by its controller.

Commit each layer’s state and keep a root for every registered state version.

Record per-layer completion times so consumers can determine maintenance status.