# Human Stewardship Framework — Architecture **Source of truth:** `src/platform/stewardship/hsg-*.ts`. ## Module map ```text HumanStewardshipRegistry (hsg-registry.ts) ├─ StewardshipVolume[] (hsg-volume.ts) ├─ StewardshipDoctrine[] (hsg-doctrine.ts) │ ├─ volumeId │ ├─ principleIds[] │ ├─ implementationBoundary │ ├─ productApplicability (hsg-product-applicability.ts) │ └─ validationRules (six constitutional guarantees) └─ StewardshipPrinciple[] (hsg-principle.ts) Guardrails: - hsg-language-governance.ts (forbidden phrases + detector) - hsg-ai-governance.ts (assistive-only AI, view guard) - hsg-validation.ts (validatePrinciple / validateDoctrine / validateRegistry) - hsg-constitution-link.ts (declarative pointers, no runtime import) ``` ## Hierarchy ```text Human Stewardship Guidelines → Volumes → Doctrines → Principles → Implementation Guidance → Validation Tests → Product Applications ``` The hierarchy is enforced by types: a `StewardshipDoctrine` belongs to one `StewardshipVolume` and references only ratified `StewardshipPrinciple`s that exist in the registry. Products consume principles via `principlesFor(registry, target)` — they never mutate them. ## Extension rules - Volumes may be appended; existing entries may not be removed or silently mutated (`volumeAppendIsSafe`). - Ratified principles are frozen: `statement` and `purpose` cannot change; only `history` grows (`validatePrinciple`). - Doctrines must declare all six constitutional guarantees to ratify (`validateDoctrine`). - Applicability defaults to universal. Silence means universal; narrowing requires an explicit target list. No exception mechanism exists (`HSG_HAS_NO_EXCEPTION_MECHANISM`). ## Product applicability `HSGApplicabilityTarget` enumerates the eleven surfaces: PosterityOS, HeadsUp, LOX, Atlas, FutureProduct, DeveloperResource, AISystem, API, Documentation, InternalTool, ExternalExperience. Every product inherits HSG. Products may add expression notes; they may not carve out exceptions.