Test Pass ID: CC-TEST-CRH-02

Version: 0.2

Date: 29 September 2026

Programme: Codex Commons

Standing: INTERNAL — EXECUTED HOSTILE EXTRACTION / KERNEL ADEQUACY TEST. No public publication or implementation authority.

Baseline: 🏛️CC-HLA-01 — Codex Commons Semantic Kernel High-Level Architecture v0.1

Purpose

Test whether Codex Resonance, Human Health, Personal Codex/CIE and federation work since the accepted Commons kernel exposes any semantic responsibility that cannot be expressed by:

RRAP + QMAP + Assessment Assertion + Non-Promotion + Lifecycle/Revalidation + external provenance/domain semantics.

Default hypothesis: NO KERNEL CHANGE.

A new profile is justified only if a material semantic responsibility remains non-reducible after hostile decomposition.

Test 1 — Representation Sufficiency

Question

Does the finding that an accurate representation may be insufficient for a different purpose/context require a new Commons profile?

Fixtures

• population aggregate → individual health inference;
• historical source state → current consequential reliance;
• semantic representation → stronger downstream question;
• version-correct representation → changed intended use.

Hostile conditions

FAIL if the current kernel cannot preserve:

• representation identity and role;
• declared purpose/intended use;
• applicability/scope;
• evidence/basis;
• a bounded assessment of sufficiency;
• requalification when purpose, source standing or version changes.

Decomposition

• RRAP identifies what role the representation is playing.
• QMAP can relate representation, target/question/context resources and preserve mapping purpose/applicability.
• Assessment Assertion can state SUFFICIENT / INSUFFICIENT / UNDETERMINED against an explicit basis without making sufficiency an intrinsic property of the resource.
• Non-Promotion prevents "accurate for P1" from becoming "sufficient for P2".
• Lifecycle/Revalidation handles source, version, purpose and standing change.
• External/domain semantics carry the substantive health, regulatory, operational or other domain meaning.

Result

STRONG PASS — NO NEW PROFILE.

Earned clarification

Representation sufficiency is purpose-, scope-, context- and evidence-relative. Sufficiency for one declared use does not confer sufficiency for another use without an explicit applicable basis.

This is a candidate Commons qualification invariant, not a fourth kernel profile.

Test 2 — Boundary Standing

Question

Does repeated standing-preservation across health, research, publication, customer, technology and federation boundaries require a standalone Standing Profile?

Fixtures

• population evidence → individual inference;
• CR finding → Arqua architecture;
• internal research → public publication;
• public Commons artefact → enterprise adoption;
• vendor assertion → customer implementation claim;
• customer evidence → Arqua learning;
• agent output → accepted/released artefact.

Hostile conditions

FAIL if transfer through Commons necessarily:

• increases standing;
• loses source standing;
• makes source standing automatically portable into consumer standing;
• requires Commons to own universal institutional-standing semantics.

Decomposition

The existing kernel preserves source role/provenance and allows explicit consumer-side assertions/mappings/assessments. The Non-Promotion Rule prevents silent elevation.

Source standing can be carried as attributable source context. Consumer standing requires an explicit applicable consumer/external basis.

Result

STRONG PASS — NO STANDING PROFILE.

Earned clarification

Boundary crossing does not confer standing. A source object may preserve attributable source standing, but standing in a receiving context must be separately grounded where required.

This strengthens existing Wave 1 standing-portability guidance; it does not change kernel composition.

Test 3 — Represented Authority

Question

Does Commons require an Authority ontology/profile in order to interoperate across medical, professional, Indigenous/custodial, legal, institutional or customer authority regimes?

Fixtures

• practitioner/clinical interpretation;
• personal choice and personal-data sovereignty;
• Indigenous/custodial authority;
• customer architecture/adoption authority;
• runtime authority represented by Arqua for qualification/admissibility purposes.

Hostile conditions

FAIL if Commons must:

• originate authority;
• decide which authority is legitimate;
• translate a source authority into Commons authority;
• infer authority from possession of data/system/role;
• universalise culturally, legally or institutionally specific authority semantics.

Decomposition

• external authority resources retain independent identity and source.
• QMAP may express a qualified relation between an assertion/resource and an authority reference for a declared purpose.
• Assessment Assertion may record an assessment against a declared authority basis.
• RRAP can distinguish authoritative source representation from observation, reference or derived material.
• provenance carries attribution.
• Non-Promotion blocks authority representation → authority.

Result

STRONG PASS — NO AUTHORITY ONTOLOGY/PROFILE.

Earned clarification

Commons may represent, reference and preserve externally constituted authority, but representation does not originate, transfer, expand or legitimise that authority.

Authority remains proposition-, scope-, jurisdiction- and institution-relative outside Commons core semantics.

Test 4 — Interpretation Assertion Gap Test

Question

Does Personal Codex/CIE expose a genuinely new semantic responsibility requiring an Interpretation Assertion Profile?

Fixtures

CIE progression:

Observation → Reference Comparison → Personal Pattern → Bounded Interpretation → Candidate Suggestion → Personal Choice → Outcome/Experience → Reflection/Revision.

Interpretation classes include descriptive comparison, personal pattern, contextual interpretation, explanatory hypothesis, research-informed interpretation, practitioner-attributed interpretation and clinical-boundary hold.

Hostile conditions

A new profile would be required if the kernel cannot preserve:

• that an output is an interpretation rather than source fact;
• the question/purpose/intended use;
• exact/recoverable input manifest;
• method/profile/version;
• evidence refs;
• limitations/uncertainty;
• alternative interpretations/disagreement;
• distinction between interpretation, hypothesis, suggestion, diagnosis, authority and action.

Decomposition

• RRAP carries the semantic role of the output: interpretation, hypothesis, suggestion, reference comparison, etc., using domain/profile role terms rather than universal Commons epistemic-state ontology.
• Assessment Assertion expresses the bounded interpreted finding against an explicit basis/method.
• QMAP expresses qualified relationships between inputs, reference profiles, findings and target concepts where needed.
• provenance binds exact inputs and method/profile.
• Non-Promotion prevents interpretation → fact, hypothesis → cause, confidence → clinical standing, suggestion → decision/authority.
• disagreement guidance preserves alternatives without forced consensus.
• lifecycle preserves later reinterpretation without overwriting earlier interpretation.

Counterexample attempt

Assume two CIE interpretations use the same observations but different declared purposes and methods. The required distinction is preserved by two independent Assessment Assertions with separate bases, purpose/method references and provenance. No kernel merge or special Interpretation object is required.

Assume a practitioner interpretation is imported. Its practitioner-attributed standing remains external/source-attributed; Commons does not convert it into clinical truth.

Assume a later model revises the interpretation. Lifecycle/versioning preserves both assertions and supersession/revalidation relationships.

Result

STRONG PASS — NO INTERPRETATION PROFILE.

Earned clarification

Interpretation is a role and/or basis-bound assertion, not automatically a distinct Commons object class. Commons must preserve its attribution, purpose, method, evidence, limitations and standing without promoting it into fact, cause, authority or permission.

A future profile should be reopened only if an unrelated domain demonstrates a non-reducible interpretation responsibility.

Test 5 — Recursive Learning / Lifecycle

Question

Does recursive learning require a new Commons Learning/Revision profile?

Fixtures

• Personal Codex: later interpretation must not overwrite earlier interpretation;
• customer → Arqua evidence-return loop;
• CR findings entering later architecture reconciliation;
• new operational evidence changing a prior assessment;
• longitudinal disagreement and supersession.

Hostile conditions

FAIL if current lifecycle guidance cannot preserve:

• historical identity/version;
• evidence that triggered reconsideration;
• new vs old standing;
• supersession without retroactive rewriting;
• independent institutional memory/authority across coupled loops;
• revalidation triggers.

Decomposition

• new evidence is represented/provenanced independently;
• new Assessment Assertions are made against the new basis;
• QMAP may relate old/new resources where qualified mapping is useful;
• lifecycle records effective period, supersession, applicability and revalidation;
• Non-Promotion prevents incoming evidence from automatically rewriting accepted standing;
• external institutional governance decides whether revision is accepted.

Result

STRONG PASS — NO LEARNING/REVISION PROFILE.

Earned clarification

Learning may trigger re-evaluation and new standing, but it does not rewrite historical standing. Revision is represented as governed succession/requalification with reconstructable provenance.

Short form:

Learning is recursive; standing remains governed; history remains reconstructable.

Cross-test result

Kernel decision

THE ACCEPTED COMMONS KERNEL SURVIVES THE CR + HEALTH + PERSONAL CODEX TEST PASS WITHOUT STRUCTURAL CHANGE.

Retain:

RRAP + QMAP + Assessment Assertion + Non-Promotion + Lifecycle/Revalidation.

Do not add:

• Representation Sufficiency Profile;
• Standing Profile;
• Authority Profile/Ontology;
• Interpretation Assertion Profile;
• Learning/Revision Profile.

Four earned Commons invariants

The test pass supports four cross-cutting candidate invariants/guidance statements:

1. Purpose-Relative Sufficiency
• Sufficiency for one declared use does not confer sufficiency for another.
2. Boundary Standing
• Boundary crossing does not confer standing.
3. Authority Preservation
• Representation of authority does not originate, transfer or expand authority.
4. Governed Revision
• Learning may change later standing but does not rewrite historical standing.

These are intentionally subordinate to / compositional with the existing No unlicensed semantic promotion rule.

Relationship to Non-Promotion

The emerging structure is:

No unlicensed semantic promotion

• standing A ≠ standing B without explicit applicable basis;
• sufficient for purpose A ≠ sufficient for purpose B;
• represented authority ≠ authority;
• interpretation ≠ fact/cause/permission;
• new evidence ≠ automatic historical rewrite.

This suggests that Non-Promotion is not merely one governance rule among several. It is the general control principle, with the four findings above as important specialisations/operational guidance.

Health significance

Codex Health remains a separate applied research domain. No health ontology, efficacy claim or clinical standing enters Commons through this test.

Its value to Commons is methodological: health provides unusually demanding fixtures for preserving distinctions among:

• evidence and inference;
• population and individual;
• observation and interpretation;
• association and cause;
• suggestion and decision;
• personal meaning and professional authority.

The kernel survives those fixtures without becoming health-specific.

Decision

CC-TEST-CRH-02 — COMPLETE / STRONG PASS / NO KERNEL CHANGE.

1. Preserve the accepted three-profile kernel.
2. Strengthen cross-cutting guidance with the four earned invariants.
3. Treat Interpretation as reducible to role + basis-bound assertion + qualified relation/provenance at current evidence.
4. Retain the rejection of universal Authority, Standing, Disagreement and Epistemic-State ontologies/profiles.
5. Keep machine-readable/GitHub implementation on HOLD.
6. No public publication is authorised by this test.
7. Reopen kernel design only when a future independent fixture demonstrates a non-reducible semantic responsibility.

Evidence / linked records

• No access
• No access
• No access
• No access
• No access
• No access