Outcome architecture proposal contracts
Status: Proposed / not authorized
Runtime registration: None
Contract declaration: Synthetic-only declarations and identifier shapes required; provenance not schema-proven
These JSON Schema 2020-12 contracts make the proposed three-layer outcome architecture and Shared Work Challenge Battery reviewable without changing the v0.1 public-draft authority boundary.
Every record must declare synthetic_only, use the applicable syn_...
identifier shapes, and carry a machine-readable provenance boundary stating
that JSON Schema does not verify provenance. Those checks constrain document
shape; they cannot determine whether submitted values were actually derived
from a real person or private record. Any application that processes these
contracts must enforce an independent environment/ingestion guard. No such
runtime is registered by this package.
verified-production-record.schema.jsonrecords only facts corroborated under a named protocol. It cannot assert capability, readiness, a score, a tier, or a credential.capability-profile.schema.jsonrepresents a multidimensional research output. Any numeric estimate is synthetic and experimental; confidence, coverage, evidence maturity, contradictions, and critical gates remain separate. Estimated dimensions must bind eligible evidence, unresolved contradictions fail closed to a disputed non-numeric state, and high confidence requires at least two eligible bindings across distinct evidence lineages and distinct review artifacts, plus reviewer agreement, and is prohibited for non-estimated states. Distinct review-artifact digests are not treated as proof of distinct reviewer identities. Profiles explicitly bind immutable rubric, evaluator, scoring-policy, competency-registry, and calculation- manifest lineage; dimension bindings must match that registry. A production- fact evidence binding must also name the exact corroborated fact ID and digest inside the bound production record; non-production sources cannot carry those fields. It expressly cannot be called a validated psychometric instrument.credential-tier-candidate.schema.jsondescribes the inputs, independent decision boundary, lifecycle, and interoperability plan for a future shareable tier. It cannot assign a tier, construct a signed payload, or issue a credential.scalar-score-authorization-gate.schema.jsonseparates 13 evidence-readiness receipts, independent method review, and a later governance decision. Each receipt is property-bound to an exact identifier and artifact type and is checked against digest-verification semantics. The governance decision is not a prerequisite for independent-review readiness. This proposal can never authorize or carry a scalar result.shared-work-challenge-battery.schema.jsonbinds human-only, model-only, and human-with-AI conditions to common scenario definitions and condition-applicable metrics while prohibiting a shared score or rank. Shared tasks are not presumed to create construct equivalence or measurement invariance.common.schema.jsondefines proposal-only shared types.
The synthetic fixtures and deterministic relational checks live in
docs/proposals/outcome-architecture.
Passing those checks creates no operational authority.
