Skip to content

PR12.6 — Generic Domain Evaluation Policy Specification v1

Purpose

PR12.6 defines the first generic declarative policy shape that can describe what semantic aspects a later governed evaluator must cover before a claim may be considered domain-sufficient.

It does not activate a policy, approve a policy, evaluate evidence, create ClaimEvaluation, derive PersonalCapabilityState, or grant progression/readiness/permission authority.

POLICY SPECIFICATION != ACTIVE POLICY
POLICY SPECIFICATION != CLAIM EVALUATION
POLICY SPECIFICATION != CLAIM TRUTH
POLICY SPECIFICATION != CAPABILITY STATE

Exact applicability

A DomainEvaluationPolicySpecification is bound to one exact:

EvaluationPolicyRef
CapabilityConceptRef revision
ClaimScope value

Applicability is exact structural equality only. There is no fuzzy scope matching, latest-revision matching, statement regex matching, or caller-controlled widening.

SAME CAPABILITY ID + DIFFERENT REVISION != SAME APPLICABILITY
SIMILAR CLAIM SCOPE != EXACT CLAIM SCOPE

Declarative requirements

Each DomainEvaluationPolicyRequirement contains only:

requirement_key
description
required_for_sufficiency

Requirement keys are canonical lowercase machine keys, unique within a specification, and stored in deterministic order. At least one requirement must be marked required_for_sufficiency=True.

The frozen v1 sufficiency semantics are identified as:

all_required_requirements_explicitly_covered

This means only that a later governed evaluation layer may establish sufficient semantic coverage when every required requirement key has been explicitly covered by governed evidence assessment. PR12.6 itself performs no evidence mapping and computes no coverage result.

REQUIREMENT != EVIDENCE
REQUIREMENT != SUPPORT
COVERAGE REQUIREMENT != RELIABILITY

No repetition or independence fiction

The v1 specification contains no observation-count, repetition, source-count, statistical-independence, evaluator-majority, confidence-weighting, or recency-weighting field.

TWO EvidenceRecord VALUES != TWO INDEPENDENT OBSERVATIONS
MULTIPLE SOURCES != INDEPENDENCE
REPETITION COUNT != REPLICATION AUTHORITY

Generic dependence/repetition governance remains future work.

Identity and serialization

domain_evaluation_policy_specification_sha256_v1(...) computes a domain-separated SHA-256 over the complete canonical declarative content:

  • exact EvaluationPolicyRef;
  • exact CapabilityConceptRef revision;
  • exact ClaimScope;
  • every canonical requirement;
  • frozen v1 sufficiency semantics.
POLICY REF != POLICY CONTENT HASH
SAME POLICY REF + DIFFERENT CONTENT != SAME SPECIFICATION
HASH != SIGNATURE
HASH != AUTHENTICATION
HASH != ACTIVATION

Schema-v1 dict/JSON serialization rejects unknown fields, missing fields, duplicate JSON keys, malformed nested objects, non-canonical requirement ordering, and changed sufficiency semantics.

Public hash/serialization paths strict-reconstruct nested typed content so post-construction runtime corruption fails closed.

Authority boundary

Production is isolated to:

capability_lab.evaluation_policy.specification
capability_lab.evaluation_policy.serialization

The generic implementation imports only shared epistemic/semantic value types and standard-library utilities. It imports no derivation, history, progression, Player Window, pilot, domain implementation, or HDE runtime authority.

The package-root capability_lab surface is unchanged.

Non-goals

PR12.6 != POLICY REGISTRY
PR12.6 != POLICY ACTIVATION
PR12.6 != POLICY APPROVAL
PR12.6 != HUMAN POLICY REVIEW
PR12.6 != ClaimEvaluation
PR12.6 != MULTI-EVIDENCE EVALUATOR
PR12.6 != CONFLICT RESOLUTION
PR12.6 != EVIDENCE INDEPENDENCE GOVERNANCE
PR12.6 != PersonalCapabilityState
PR12.6 != CLOSED-LOOP CAPABILITY UPDATE

The intended later chain remains separate:

PR12.6 specification
→ governed policy admission/registry
→ exact admitted policy application
→ directional ClaimEvaluation
→ existing complete-portfolio/state path