EM-XCT-05 · Published partial scope

Enterprise Disclosure Review

Describe an exact package of single-object projections and preserve attributed reviews of what their combination may disclose.

Research and executed evidence

Claude and Grok completed independent studies, three separate prototype audits each, and three frozen full-package audit rounds. Both final R3 audits accepted the bounded reference with limits; concrete R1/R2 findings were corrected. Executed code/native checks are separate Codex evidence.

108 behavior checks; three new synthetic Vercy Dimensions; two independent objects and three record revisions each; 66 native negative checks overall. Downloaded-package verification is recorded separately. All examples are synthetic.

Reviewable draft, not a serving authorization or privacy guarantee. The reference compares metadata and attributed review applicability. A trusted host still resolves authentic actors, current context, rights, source pins and custody.

Separate identities and responsibilities

ContextPackageProposal pins exact source objects, scalar field declarations, schema, projection shape, classification, audience and purpose. JointDisclosureReview has its own identity, reviewer, validity and correction chain. Members are embedded values; the review result is an ephemeral diagnostic. Expiry, changed context, incomplete active sets and conflicting verdicts are explicit.

Corrections from independent review

R1/R2 prototype findings and full-package R1 findings were reconciled before the final R3 freeze. Corrections include type-safe storage authority, independently exercised time guards, distinctly timed synthetic snapshots, complete reference evidence and precise host storage/lifecycle limits. Earlier reviewed snapshots remain evidence; they are not relabeled as the published version.

Actual evidence

Provider responses preserve their original form. JSON/YAML files were compact-rendered with identical parsed values for audit; reviewers did not run code, recompute hashes or inspect external toolchain bytes. Exact frozen inputs and declared read scopes are retained.

Remaining broader scope

Production source/grant/classification/custody adapters; inference/privacy assessment beyond attributed opinions; retention holds and disposal execution; full-scale storage/rollover/CAS, transitive active-set governance and existing-Dimension migration. All inherited parent research holds remain.

Five predecessor specifications were compared with exact byte pins; their holds remain visible. Classification and retention research is not declared complete by this narrower publication.

Dossier checksums · Next research contour