Skip to content

Execution plan: ground ontology alignment lessons into cskg.vnext and boundary profiles #69

@mdheller

Description

@mdheller

Purpose

Track the implementation plan for converting the Drive-folder ontology-alignment / KBP / grounded-reading lessons into repo artifacts and executable validation. This follows the inventory and lesson ledgers added on 2026-05-10 and links to the existing cskg.vnext and boundary-admission work.

Objective

Move from research grounding to implementation substrate:

paper lesson
-> typed record/schema
-> boundary watcher rule
-> golden fixture
-> negative fixture
-> implementation helper
-> validation receipt
-> cross-repo adoption plan

Tranche plan

T1 — Spec expansion

Target file:

  • spec/drafts/cskg_vnext_profile.md

Add first-class record definitions:

  • EvidenceContextRecord
  • AlignmentCandidateRecord
  • ConceptLinkCandidate
  • ModuleBoundaryRecord
  • MatcherRunReceipt
  • OutlierAuditRecord
  • PipelineRunReceipt
  • EDLReceipt
  • StructuredSourceMergeReceipt
  • AttributeValidationReceipt
  • DynamicsManifest
  • FeedbackReviewRecord

Acceptance criteria:

  • records have required/optional field classification;
  • record lifecycle is mapped to Beacon-A/B/C and BoundaryReceipt;
  • each record includes claim/evidence/admission state;
  • identity and privacy labels are explicit where applicable.

T2 — Annex X boundary semantics

Target file:

  • spec/drafts/annex_x_boundary_admission_profile.md

Add:

  • BoundaryWatcherDecision
  • BoundaryReceipt
  • BoundaryStateMachine
  • FailClosedCompositionPolicy
  • BoundaryNegativeTestVector

Acceptance criteria:

  • no path bypasses classification, validation, and receipt;
  • fallback cannot mean fail-open;
  • watcher timeouts deny/quarantine;
  • DoNotLearn and DoNotLink are defaulted for unknown identity risk;
  • proof-program and worldlet examples are included.

T3 — Golden and negative fixture set

Add fixture targets for:

Positive:

  • public ConceptNet-style IsA edge;
  • WordNet/ontology PartOf edge;
  • Socrates binary context evidence;
  • Socrates unary context evidence;
  • Socrates composite TOD evidence;
  • locality module with coverage receipt;
  • RTFM-style DynamicsManifest.

Negative:

  • embedding-near SameAs human identity link;
  • unary-derived contact relation;
  • uncertain concept link without review;
  • ontology module without coverage receipt;
  • worldlet action without dynamics manifest;
  • semantic-memory promotion without PromotionRule receipt;
  • vectorization/training on unclassified sentence text.

Acceptance criteria:

  • every negative fixture fails closed;
  • every positive fixture emits a receipt;
  • fixture names encode expected decision state.

T4 — Experimental reference implementation

Target package:

  • reference/experimental/tritrpc_requirements_impl_v4/

Add candidate modules:

  • boundary.py
  • cskg.py
  • memory.py
  • identity_prime.py

Acceptance criteria:

  • canonical encode/decode helpers for new records;
  • deterministic canonicalization for CSKG array/list fields;
  • watcher composition function;
  • fail-closed state transitions;
  • tests for fixtures.

T5 — Cross-repo adoption issues

Create downstream issues for:

  • SocioProphet/ontogenesis: schema generation and ontology/module objects;
  • SocioProphet/holmes: concept-link and formalization candidates;
  • SocioProphet/sherlock-search: evidence retrieval and context records;
  • SocioProphet/guardrail-fabric: watcher policy and fail-closed defaults;
  • SocioProphet/sociosphere: registry/admission dashboard and memory release ledger.

Acceptance criteria:

  • issues reference TriTRPC profile paths;
  • each downstream issue has acceptance tests;
  • no repo claims ownership of another repo’s boundary.

T6 — Proof/evidence doctrine hardening

Add a compact doctrine note connecting Gödelization, CSKG, FSMS, and boundary admission.

Acceptance criteria:

  • explicit distinction: candidate / evidence / admitted / promoted / theorem;
  • SameAs and entity resolution treated as proof obligations;
  • mathematical proof-program analogue included.

Estimated turn count

Chat-driven implementation estimate: 8 turns total, assuming connector write access remains available and no large CI failures block us.

  1. Spec expansion for records.
  2. Annex X boundary profile draft.
  3. Fixture design and JSON/YAML examples.
  4. Experimental reference implementation scaffolding.
  5. Test suite for fail-closed behavior.
  6. Cross-repo issue fan-out.
  7. Review pass and gap audit.
  8. Stabilization pass: tighten names, non-goals, acceptance criteria, and validation commands.

If we keep the work docs-only/spec-only, it can be done in about 4 turns. If we include reference code and tests, use 8 turns. If we require PRs, CI, and downstream cross-repo adoption, expect 10-12 turns.

Non-goals

  • Do not implement a production ontology matcher in TriTRPC.
  • Do not make CSKG relation semantics part of Braid243.
  • Do not admit identity/SameAs edges without Identity-Prime controls.
  • Do not permit direct semantic-memory promotion from parser/model output.
  • Do not replace stable v1.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions