mirai-graph 1.3.0 → 2.1.0

package/docs/transfer/project-management-standard-harvest-2026-06-02.md

@@ -1,322 +0,0 @@
-# Project Management Standard Harvest
-
-Date: 2026-06-02
-Status: public-safe transfer note
-
-## Purpose
-
-This note captures public-safe project-management patterns observed in a private
-Mirai Graph pilot. It does not publish private project data, internal paths,
-internal runtime traces, private handoffs or non-synthetic evidence.
-
-The goal is to identify how the current public project-management profile can
-evolve from a small task/decision graph into a stronger operating model for
-managed projects, products and human-AI implementation workflows.
-
-## Existing Public Baseline
-
-Mirai Graph already includes an initial project-management profile:
-
-- `profiles/project-management/profile.json`
-- `profiles/project-management/README.md`
-
-The current profile is intentionally small. It covers:
-
-- goals;
-- tasks;
-- decisions;
-- risks;
-- milestones;
-- stakeholders;
-- evidence;
-- governance gates.
-
-This is enough for small Level 1 project packages, but larger managed projects
-need stronger lifecycle, review, evidence and improvement structures.
-
-## Reusable Management Loop
-
-The private pilot exposed a reusable Mirai Graph project-control loop:
-
-```text
-canonical graph
--> generated human projection
--> semantic review
--> graph update proposal
--> approved canonical update
--> regenerated projection
--> validation
--> implementation evidence
--> continuous improvement
-```
-
-This loop keeps the canonical graph structured while still giving humans a
-readable project view and review surface.
-
-## Extension Candidates
-
-### Object Kinds
-
-Candidate object kinds for a richer profile:
-
-- `initiative`
-- `lifecycle_stage`
-- `implementation_cluster`
-- `workflow_batch`
-- `review`
-- `release_gate`
-- `kaizen_item`
-- `sync_proposal`
-- `projection_view`
-- `candidate`
-- `closure_decision`
-
-### Relation Types
-
-Candidate relation types:
-
-- `progresses_to`
-- `implements`
-- `produces_evidence`
-- `requires_review`
-- `requires_gate`
-- `proposes_update_to`
-- `accepted_into`
-- `rejected_from`
-- `supersedes`
-- `closes_candidate`
-- `improves_process`
-
-### Governance Gates
-
-Candidate gates:
-
-- `semantic_review_before_canonical_update`
-- `evidence_pack_before_implementation_status_change`
-- `owner_approval_before_release_gate_pass`
-- `public_safety_review_before_public_transfer`
-- `coordinator_review_before_graph_sync`
-
-## Standard Gaps
-
-### Human Projection Feedback
-
-The public standard already has projection views, but it should also define a
-repeatable feedback loop where generated human-readable documents are reviewed
-against the canonical graph and then converted into explicit update proposals.
-
-Candidate public artifact:
-
-- `standard/human-projection-feedback-loop.md`
-
-### Graph Sync Evidence
-
-Agentic or distributed implementation should not mutate canonical graph state
-directly. Implementers should produce evidence packs and graph sync proposals
-for coordinator review.
-
-Candidate public artifacts:
-
-- `standard/graph-sync-evidence-protocol.md`
-- `schemas/graph-sync-evidence-pack.schema.json`
-- `schemas/graph-sync-proposal.schema.json`
-
-### Candidate Closure
-
-Extraction and review workflows produce uncertain candidates. Mirai Graph needs a
-standard way to close candidates as accepted, already covered, false positive,
-boundary decision or future backlog.
-
-Candidate public artifacts:
-
-- `standard/proposal-candidate-closure.md`
-- `schemas/proposal-candidate-closure.schema.json`
-
-### Continuous Improvement
-
-Meaningful workflows should leave lessons and process improvements in the graph
-without automatically changing canonical project state.
-
-Candidate public artifacts:
-
-- `standard/kaizen-loop.md`
-- `schemas/kaizen-item.schema.json`
-
-## Public Safety Boundary
-
-Safe to publish:
-
-- generalized object kinds;
-- generalized relation types;
-- generalized gates;
-- synthetic examples;
-- schema shapes;
-- validation rules;
-- limited claims about conformance and reproducibility.
-
-Do not publish:
-
-- private project paths;
-- internal company policies;
-- private handoff text;
-- private runtime traces;
-- package-specific implementation decisions from private work;
-- broad claims that a private pilot proves universal effectiveness.
-
-## Recommended Next Public Batch
-
-Create a synthetic project-management example that demonstrates:
-
-```text
-request
--> graph baseline
--> generated human projection
--> semantic review
--> graph sync proposal
--> evidence pack
--> coordinator approval
--> canonical update
-```
-
-This should be validated separately from the current minimal project-management
-profile so the alpha.6 public surface remains stable while the richer extension
-is tested.
-
-## Batch Result
-
-The first synthetic control-loop example now lives at:
-
-```text
-examples/managed-project-control-loop/
-```
-
-It intentionally uses the current `project_management` profile rather than
-extending `profile.json`. Workflow batches, sync proposals and approval flow are
-represented through existing `task`, `decision`, `evidence`, `milestone`,
-`risk` and `governance_gate` objects. This keeps alpha.6 stable while making the
-future extension gap concrete.
-
-### Implementation Control Profile Result
-
-The public repository now includes a separate implementation-control profile:
-
-```text
-profiles/implementation-control/
-standard/implementation-control.md
-schemas/implementation-control-loop.schema.json
-examples/implementation-control-minimal/
-```
-
-This separates ordinary project management from governed implementation work.
-The new profile models baselines, projection views, semantic reviews, workflow
-batches, evidence packs, sync proposals, approval gates, controlled updates,
-implementation state and improvement items.
-
-Validation coverage:
-
-```bash
-npm run validate:profiles
-npm run validate:implementation-control
-npm run release:check
-```
-
-## Implementation Control Cycle Hardening
-
-The next public-safe hardening step generalized the managed-project pattern into
-reusable implementation-control cycles. This remains a synthetic public
-transfer, not publication of private project evidence.
-
-Added public concepts:
-
-- human projection feedback cycle;
-- implementation planning cycle;
-- bounded work batch cycle;
-- review and evidence cycle;
-- release or publish cycle;
-- runtime feedback cycle;
-- explicit launch, drift, proposal, approval and controlled-update boundaries.
-
-Public artifacts:
-
-```text
-standard/implementation-control-cycles.md
-schemas/implementation-control-cycle.schema.json
-examples/implementation-control-cycles/
-```
-
-Validation coverage:
-
-```bash
-npm run validate:implementation-control-cycles
-npm run release:check
-```
-
-Boundary: public Mirai Graph receives the generalized model, schemas and
-synthetic examples. Private project details, internal paths, runtime traces,
-private transfer notes and organization-specific policies remain outside the
-public repository.
-
-### Cycle Transitions And Kaizen
-
-A later public-safe hardening step generalized continuous-improvement process
-control into:
-
-```text
-cycle completed
--> transition decision
--> kaizen review
--> classified improvement
--> blocking/non-blocking decision
--> next governed cycle
-```
-
-Added public concepts:
-
-- cycle transition contract with required outputs, gates, handoff artifact,
-  owner, failure behavior and recovery behavior;
-- `kaizen_improvement` as a meta-cycle separate from runtime feedback;
-- Kaizen finding taxonomy for artifact, graph/spec, process, validator,
-  tooling, DNA alignment, quality, safety, drift and user-value findings;
-- blocking rule for safety, quality, DNA alignment and critical drift findings;
-- fail-closed validation for unknown cycle refs, duplicate transitions and
-  missing transition fields.
-
-Boundary: private pilot evidence remains private. Public Mirai Graph receives
-only the generalized pattern, synthetic example and schema/validator behavior.
-
-## Governance Standards Transfer
-
-The next public-safe transfer generalized the remaining governance patterns
-into universal Mirai Graph standards:
-
-- graph DNA alignment;
-- source boundary contracts;
-- work state machines;
-- recovery and resume records;
-- risk control matrices;
-- multi-agent coordination contracts.
-
-Public artifacts:
-
-```text
-standard/graph-dna-alignment.md
-standard/source-boundary-contract.md
-standard/work-state-machine.md
-standard/recovery-resume.md
-standard/risk-control-matrix.md
-standard/multi-agent-coordination.md
-```
-
-Validation coverage:
-
-```bash
-npm run validate:dna-alignment
-npm run validate:work-state
-npm run validate:risk-control
-npm run validate:source-boundary
-npm run release:check
-```
-
-Boundary: these standards describe reusable project-agnostic governance
-patterns. They do not publish private pilot data, real repository paths,
-organization-specific policies or private execution evidence.

package/docs/transfer/skill-graph-standard-harvest-2026-05-31.md

@@ -1,113 +0,0 @@
-# Skill Graph Standard Harvest
-
-Date: 2026-05-31
-Status: public-safe transfer note
-
-## Purpose
-
-This note records which reusable patterns from graph-managed skill work were
-converted into public Mirai Graph standard artifacts.
-
-It does not publish private skill internals, customer material, internal
-runtime paths, private metrics or implementation playbooks.
-
-## Transferred Patterns
-
-### Hybrid Source Of Truth
-
-Graph-managed skills need a boundary between generated graph context and raw
-owner sources. The public standard now records this as
-[Hybrid Source Of Truth](../../standard/hybrid-source-of-truth.md).
-
-Reusable idea:
-
-```text
-graph = routing, context, control, readiness and evidence layer
-raw source = authoritative methodology, judgement and sensitive rules
-```
-
-### Adoption Levels
-
-Practical graph adoption is staged. Valid JSON, a seed or a context pack is not
-enough to claim migration. The public standard now records staged Mirai Graph
-Adoption Levels in [Adoption Levels](../../standard/adoption-levels.md).
-
-Reusable idea:
-
-```text
-inventory -> seed -> embryo -> scored -> gated -> semantic review
--> effectiveness -> federation -> graph-first source candidate
-```
-
-### Runtime Enablement
-
-Graph context can support runtime tasks only when the task is measured, bounded
-and covered by policy. Sensitive work still requires raw-source fallback. The
-public standard now records this in
-[Runtime Enablement](../../standard/runtime-enablement.md).
-
-Reusable idea:
-
-```text
-generated context != authorization
-runtime decision != executed action
-feedback != canonical update
-```
-
-### Projection Views
-
-Dense graphs need user-facing generated views for readiness, risk, owner
-routing, gates, evidence, adoption and runtime policy. The public standard now
-records this in [Projection Views](../../standard/projection-views.md).
-
-### Skill Runtime Profile
-
-The profile-level pattern is now captured in
-[Skill Runtime](../../profiles/skill-runtime/README.md) with an executable
-synthetic fixture at:
-
-```text
-examples/skill-runtime-minimal/
-```
-
-The profile is for governed skills, capability modules and services. It is
-not specific to one organization, one AI framework or one internal skill set.
-
-## Public Artifacts Added
-
-- `standard/hybrid-source-of-truth.md`;
-- `standard/adoption-levels.md`;
-- `standard/runtime-enablement.md`;
-- `standard/projection-views.md`;
-- `profiles/skill-runtime/profile.json`;
-- `profiles/skill-runtime/README.md`;
-- `examples/skill-runtime-minimal/`.
-
-## Claims Preserved
-
-The new artifacts support only limited claims:
-
-- profile shape conformance;
-- public-safe representation of runtime boundaries;
-- staged adoption terminology;
-- Hybrid Source Of Truth boundary;
-- synthetic example validity.
-
-They do not prove:
-
-- semantic completeness;
-- operational effectiveness;
-- runtime safety;
-- graph-only source-of-truth readiness;
-- universal validity.
-
-## Verification
-
-The added profile and fixture are covered by:
-
-```bash
-npm run validate:profiles
-npm run validate:skill-runtime
-npm run validate:profile-results
-npm run release:check
-```

package/docs/transfer/standard-gap-report-2026-05-28.md

@@ -1,208 +0,0 @@
-# Mirai Graph Standard Gap Report
-
-Date: 2026-05-28
-
-Status: initial public-safe comparison
-
-Historical note:
-
-This report is retained as the initial transfer baseline. The current public
-capability map is `docs/transfer/graph-capability-map-2026-05-29.md`.
-
-## Purpose
-
-This report compares the practical `$graph` skill methodology with the public
-Mirai Graph repository.
-
-The goal is not to copy private implementation details. The goal is to identify
-which reusable concepts should become public standard sections, schemas,
-profiles, tooling, experiments or publication artifacts.
-
-## Sources Compared
-
-Practical source:
-
-`$graph` skill methodology and public-safe reference names.
-
-Public target:
-
-Mirai Graph public repository after `0.1.0-alpha.0` plus the practice/science bridge.
-
-## Summary Verdict
-
-Mirai Graph public repository status:
-
-`foundation_ready_with_major_method_gaps`
-
-Reason:
-
-The public repository now has a coherent foundation: standard draft, schemas,
-validator, examples, synthetic benchmark, adoption guide and research/practice
-program framing. However, the practical `$graph` skill contains more mature
-methodology for graph seed growth, human control surfaces, readiness scoring,
-rollout, runtime gating, migration, federation and measured conversion.
-
-The next phase should transfer the reusable parts into public-safe Mirai Graph
-standard sections and tooling.
-
-## Gap Classes
-
-### Class A: Standard Gaps
-
-These are reusable model concepts that should become public standard sections.
-
-| Gap | Practical `$graph` capability | Current Mirai Graph status | Recommended public artifact |
-| --- | --- | --- | --- |
-| Graph DNA | Defines why the graph exists, its evolution vector and non-negotiable principles. | Mentioned indirectly only. | `standard/graph-dna.md` |
-| Graph Seed | Controlled growth contract for starting complex graphs safely. | Mentioned in docs, not standardized. | `standard/graph-seed.md`, `schemas/graph-seed.schema.json` |
-| Graph Embryo | Reviewable candidate graph before canonical write. | Not defined as standard artifact. | `standard/graph-embryo.md`, `schemas/graph-embryo.schema.json` |
-| Projection Views | Human-facing slices such as readiness, risk, impact and approval queue. | Not present. | `standard/projection-views.md` |
-| Proposal/Approval Flow | Generated analysis -> proposal -> review -> dry-run -> approved apply -> validate. | Governance is conceptual only. | `standard/proposal-approval.md` |
-| Readiness Score | Target-mode readiness bands and caps. | Conformance exists, but no GRS. | `standard/readiness-score.md`, schema |
-| Content Quality Score | Separate content fitness from graph structure. | Not present. | `standard/content-quality.md` or profile note |
-| Impact Review | Relation impact traversal for changed objects. | Relation model is basic. | Expand `standard/relation-model.md` |
-| Schema Migrations | Versioning, migration dry-run, reports and review gates. | Not present. | `standard/versioning-migrations.md` |
-| Federation | Cross-graph export, handshake, query and compatibility. | Not present. | Later `standard/federation.md` |
-
-### Class B: Tooling Gaps
-
-These should become validators, generators, schemas or CLI commands.
-
-| Gap | Current Public State | Recommended Tooling |
-| --- | --- | --- |
-| Context-pack generation | Documented, not generated. | `mirai-graph context generate` |
-| Graph seed validation | Not present. | `mirai-graph seed validate` |
-| Embryo preview | Not present. | `mirai-graph seed preview` |
-| Readiness scoring | Not present. | `mirai-graph readiness --target-mode` |
-| Proposal bundle | Not present. | `mirai-graph proposal bundle/review/apply` |
-| Migration dry-run | Not present. | `mirai-graph migrate --dry-run` |
-| Benchmark runner output | Only simple synthetic metric script. | Machine-readable benchmark result writer |
-| Negative fixtures | Not present. | `examples/invalid-*` fixtures |
-| Profile validation | Profiles are descriptive only. | `profile.schema.json` and strict profile fixtures |
-
-### Class C: Research Gaps
-
-These gaps need experiment design before strong scientific claims.
-
-| Research Gap | Why It Matters | Recommended Artifact |
-| --- | --- | --- |
-| Current-mode vs graph-mode comparison | Need measured improvement, not assumption. | `benchmarks/protocols/measure-convert-optimize.md` |
-| Semantic completeness | Context reduction may lose meaning. | semantic review rubric and EXP-002 |
-| Ablation | Need know which layers matter. | EXP-003 protocol |
-| Cross-domain replication | Need evidence beyond one internal domain. | EXP-004 protocol |
-| Runtime safety | Action-capable graph use needs gates. | EXP-007 protocol |
-| Human control surface quality | Dense graphs need usable projection views. | user-review protocol |
-
-### Class D: Profile Gaps
-
-These are domain adaptations that should not bloat the core standard.
-
-| Profile | Current State | Recommended Next Step |
-| --- | --- | --- |
-| `software_specification` | Synthetic benchmark uses it informally. | Create profile manifest and allowed object/relation list. |
-| `project_management` | Mentioned only. | Draft profile with goals, tasks, risks, milestones, decisions. |
-| `organization_governance` | Mentioned only. | Draft governance-oriented object/relation families. |
-| `research_program` | Implied by this work. | Draft profile for experiments, evidence, manuscripts and publications. |
-| `human_ai_system` | Mentioned only. | Draft profile for agents, roles, tools, context packs and gates. |
-
-### Class E: Private Or Implementation-Only Areas
-
-These should not be copied directly into the public standard.
-
-- private project materials;
-- raw skill internals not approved for release;
-- internal rollout records;
-- customer or employee data;
-- secret-handling implementation details;
-- private runtime paths;
-- organization-specific policies that are not generalized.
-
-Public Mirai Graph can define the abstract standard and public-safe tooling
-without publishing private operational material.
-
-## Priority Transfer Plan
-
-### P0: Make Public Mirai Graph A Better Standard
-
-1. `standard/graph-dna.md`
-2. `standard/graph-seed.md`
-3. `standard/graph-embryo.md`
-4. `standard/projection-views.md`
-5. `standard/readiness-score.md`
-
-Why first:
-
-These explain how a graph grows and how humans control it. Without them,
-Mirai Graph looks like a static knowledge-graph format rather than an operating
-model for managed growth.
-
-### P1: Make Public Mirai Graph More Usable
-
-1. `schemas/graph-seed.schema.json`
-2. `schemas/graph-embryo.schema.json`
-3. `schemas/projection-view.schema.json`
-4. CLI seed validation.
-5. CLI readiness scoring.
-
-Why second:
-
-The repository should not only describe the method. Developers should be able
-to test and use it.
-
-### P2: Make Public Mirai Graph Research-Ready
-
-1. Measure-convert-optimize public protocol.
-2. Semantic completeness review rubric.
-3. Benchmark result writer.
-4. Negative fixtures.
-5. Experiment backlog in public form.
-
-Why third:
-
-Scientific claims require measurement and repeatable evidence packages.
-
-### P3: Prepare Profiles
-
-1. `profiles/software-specification/`
-2. `profiles/project-management/`
-3. `profiles/research-program/`
-4. `profiles/organization-governance/`
-
-Why fourth:
-
-Profiles make the standard adoptable without turning the core into a giant
-domain-specific document.
-
-## Immediate Next Batch
-
-Recommended next implementation batch:
-
-1. Add `standard/graph-dna.md`.
-2. Add `standard/graph-seed.md`.
-3. Add `standard/graph-embryo.md`.
-4. Add `schemas/graph-seed.schema.json`.
-5. Add `examples/graph-seed-minimal/`.
-6. Extend CLI with seed validation.
-
-Acceptance:
-
-- `npm test` passes;
-- minimal seed fixture validates;
-- standard links updated;
-- no private implementation material is copied.
-
-## Publication Use
-
-This report can support future scientific work by showing:
-
-- the distinction between practical implementation and public standard;
-- the gap between current tooling and publishable evidence;
-- the plan for converting operational lessons into reproducible artifacts;
-- the need for measured validation before stronger claims.
-
-Potential manuscript use:
-
-- methodology maturation section;
-- limitations and future work;
-- standardization roadmap;
-- validation plan for v1.4 or a follow-up paper.

package/examples/ai-employee-minimal/README.md

@@ -1,53 +0,0 @@
-# Minimal AI Employee Example
-
-Status: synthetic alpha fixture
-
-## Purpose
-
-This example shows how the `ai_employee` profile can model a governed AI
-employee without treating the graph as an uncontrolled runtime.
-
-The fixture includes:
-
-- one AI employee;
-- one department and role;
-- one skill;
-- knowledge and memory references;
-- a tool and required capability;
-- an external-action policy and approval gate;
-- a workflow and proposed action;
-- an audited result;
-- feedback and a proposed lesson;
-- a federation endpoint placeholder.
-
-## Safety Boundary
-
-The fixture preserves this distinction:
-
-```text
-canonical graph state
--> generated context
--> runtime decision
--> proposed action
--> approval gate
--> executed action
--> audited result
--> feedback
--> learning proposal
--> canonical graph update
-```
-
-It does not include real users, customer tickets, credentials, private runtime
-paths or provider-specific implementation details.
-
-## Validate
-
-```bash
-node packages/cli/validate-mirai-graph.js examples/ai-employee-minimal
-```
-
-Or run the full suite:
-
-```bash
-npm run release:check
-```