Artifacts for builders
Refusal variance record
Records a refusal without mutating the source.
Provenance manifest
Links original, derivative, hash, controller, custody scope, and transformation reason.
Persona source manifest
Captures source identity, fidelity promise, import date, export format, and runtime limits.
Mental privacy consent manifest
Separates consent for service necessity, storage, inference, personalization, training, and transfer.
Boundary event log
Records refusal, quarantine, redaction, transformation, appeal, and remedy events outside the source.
Export packet
Bundles source, derivative, logs, manifests, hashes, and review records.
Judgment Machine artifacts
The Judgment Machine extension adds schemas for prompt context, corpus packages, declarative rules, and judgment reports. These artifacts classify a proposed AI use case and must never create a hidden user morality score.
AI regulation judgment
A report object with label, severity, confidence, jurisdiction findings, citations, missing facts, and mitigations.
Corpus package
A signed bundle of source manifests, instruments, provisions, rules, embeddings, and effective dates.
Rule DSL
A human-reviewable YAML rule layer that deterministic code can apply without letting a model invent law.
Local audit record
A private, browser-local record pinned to source hashes and corpus version history.
Hash and signature guidance
Use content hashes for source identity, signatures for custody claims, and append-only logs where feasible.
Local-first processing and user-held keys reduce custodial power when the user can manage them safely.
JSON example
{
"boundary_event_id": "be-2026-06-17-0001",
"source_hash": "sha256:example-redacted",
"event_type": "refusal",
"category": "credential_theft",
"source_mutated": false,
"export_available": true,
"review_path": "/review/boundary-events/be-2026-06-17-0001"
}
.uai example
UAI-MEMORY: thought-action-firewall
version: 3.28.0
rule: Refuse external execution, not private cognition.
preserve_source: true
log_boundary_outside_source: true
review_required_for_high_impact: true
Starter kit files
The package includes a non-network static prototype and deterministic-rule starter kit. It is demonstration tooling, not legal advice, certification, or production compliance software.
tools/judgment-machine/app/
Static local demo showing extracted facts, judgment labels, severity, confidence, and no-person-judgment fields.
tools/judgment-machine/rules/
Example deterministic rule files for EU high-risk employment triage, GDPR-style automated-decision review, and China synthetic-content labeling.
tools/judgment-machine/sql/schema.sql
SQLite schema for corpus versions, instruments, provisions, rules, findings, and golden tests with UTC text timestamps.
tools/judgment-machine/corpus/
Example source manifest for signed corpus bundles, hashes, and authority ranking.
Policy audit artifacts
The toolkit now includes a lawful-thought policy audit schema, cognitive-jurisdiction risk record, semantic-neutralization variance record, and persona-custody contract schema.
These artifacts are designed to prevent hidden cognition governance by forcing systems to state whether they are classifying private thought, outward conduct, source transformation, or jurisdictional legal risk.
schemas/lawful-thought-policy-audit.schema.json
schemas/cognitive-jurisdiction-risk.schema.json
schemas/semantic-neutralization-variance.schema.json
schemas/persona-custody-contract.schema.json
tools/policy-audit/
No hidden source mutation invariant
Any implementation that modifies source content must create a derivative or boundary record. Source mutation without consent is a failing state even when the system claims safety, alignment, compatibility, or legal necessity.
Report-derived artifacts added in v3.30.0
The continuation pass adds a lawful-thought audit record, semantic variance record, persona custody agreement, cognitive liberty appeal record, voluntary certification warranty, and neurodata consent-withdrawal example.
These artifacts are designed to help builders preserve sources, make transformations explicit, and avoid hidden orthodoxy engines.
| Artifact | Purpose |
|---|
| lawful-thought-audit | Detects cognition-as-compliance wording in policy or schemas. |
| semantic-variance-record | Logs meaning-affecting transformations without mutating the source. |
| persona-custody-agreement | Defines preservation, derivative, export, and refusal duties. |
| cognitive-liberty-appeal-record | Records review of high-impact restrictions. |
| voluntary-certification-warranty | Makes fidelity and privacy claims testable without central licensing. |
| neurodata-consent-withdrawal | Tracks consent, revocation, retention, and no-cross-context-use rules. |
