artifacts/standard-named

Rendered from markdown source. Open raw source on GitHub.

Deployment Standing, Runtime Authority, and Tool Grants Matrix

Status: candidate pre-production Content canon status: unset

| Layer | Governing question | Required record | Must not become | |---|---|---|---| | Training lineage | How was capability recruited? | ConsentfulTrainingWitness | Runtime authority | | Deployment field | Where will consequence occur? | DeploymentFieldAssembly | Mere implementation detail | | Runtime standing | Who bears consequence or holds responsibility? | DeploymentStandingRecord | Training-contributor list | | Runtime purpose | Why is the model present? | RuntimePurposeAuthorityGrant | Broad permanent mandate | | Capability | What can the model do? | Capability inventory | Permission | | Role | Which capabilities may it exercise? | CapabilityRoleToolGrant | Self-selected authority | | Tool grant | Which tools may it call? | CapabilityRoleToolGrant | Authority over tool targets | | Consent or other authority | Who may authorize the operation? | ConsentNoticeRefusalProfile and authority grant | Notice or default | | Human re-entry | Who can understand, stop, and revise? | Action gate and review record | Ceremonial approval | | Witness | What actually governed the action? | ConsentfulDeploymentWitness | Provider-only narrative |

Runtime gate

deployment field assembled
∧ affected standing represented
∧ runtime purpose active
∧ operation allowed
∧ model role assigned
∧ tool grant valid
∧ target authority valid
∧ consent or other authority valid
∧ context adequate
∧ protected conditions pass

Hard distinctions

training authority ≠ deployment authority
capability ≠ role
role ≠ tool permission
tool permission ≠ target authority
notice ≠ consent
human confirmation ≠ legitimacy

Governing principle

Training creates capability. Deployment creates consequence. Authority must be established again.