DLP Specification
Part IX: Implementation

§27 Operation Catalog

244+ operations across 6 types. Five-gate authorization. Two-pass invariant enforcement.

§27 Operation Catalog

This specification section defines implementation contracts for the Decision Lineage Protocol substrate.

The Decision Lineage Protocol substrate exposes two hundred and forty-four named operations across the nineteen primitives and ten cross-cutting operation domains. Each operation is a typed, authorized action with defined preconditions, postconditions, and invariant implications. This section specifies the complete logical contract that governs what the SDK must support; transport protocol, serialization format, and client-library ergonomics are DESIGN SPACE. The operations act on the logical schema defined in §26.

This section defines six operation types (§27.1), specifies per-primitive operation contracts organized by tier (§27.2–§27.6), catalogs cross-cutting composite operations (§27.7), consolidates the authorization matrix (§27.8), maps invariant enforcement points per operation (§27.9), and formalizes SDK constraints (§27.10).


§27.1 Operation Architecture Overview

§27.1.1 Six Operation Types

Every operation in the catalog belongs to exactly one of six types. The type determines the operation's structural role, its relationship to the primitive lifecycle, and its invariant enforcement profile.

Table 27.1.1: Operation Type Taxonomy

TypeDefinitionInvariant ProfileExample
CreateInstantiate a new primitive record with all minimum viable record (MVR) fieldsPass 1 always fires; truth type assigned at creationintent.create, evidence.create
ReadRetrieve primitive records with truth type filtering and projection controlNo invariant enforcement; respects depth-gated visibility (§23)intent.read, authority.read
UpdateModify mutable fields on existing records, respecting truth type immutability rulesPass 1 fires on affected invariants; Authoritative records reject updatescommitment.update, constraint.update
DeleteSoft-delete only — transitions record_lifecycle_state to archivedNo hard deletes in the governance graph; audit trail preservedcontext.deactivate, account.close
TransitionAdvance a primitive through its state machine lifecyclePass 1 fires; state machine guards enforce valid transitionsintent.activate, work.complete
CompositeMulti-primitive operations that compose atomic operations within a single logical transactionAll constituent invariants fire; transaction is atomicsignal.create, authority.delegate

Create and Read are universal. Every primitive supports at minimum a Create and a Read operation. Tier 1 primitives support the full six-type range. Tier 2 primitives omit some Transition operations because their lifecycles are simpler.

Transition is the primary governed operation type. Most governance-significant actions are state machine transitions, not field updates. A Commitment does not become active by updating a status field — it becomes active through commitment.activate, which enforces B2 capacity verification, records the transition event, and triggers downstream invariant checks.

Composite operations are atomic. A composite operation either completes entirely or fails entirely. Partial completion is not a valid state. The SDK determines whether atomicity is achieved through database transactions, saga patterns, or another mechanism — this is DESIGN SPACE.

§27.1.2 Authorization Model

Every write operation — Create, Update, Delete, Transition, and Composite — requires authorization. Read operations respect depth-gated visibility (§23) but do not require explicit authorization checks beyond instance membership.

The authorization check comprises five sequential gates:

  1. Authority scope. The actor holds an Authority with scope covering the operation target.
  2. Authority traceability. The Authority is traceable to a root authority via the B5 delegation chain.
  3. RACIVG role permission. The actor's RACIVG role for the target context permits the operation (§22.3).
  4. Graduation gate. The actor's graduation stage meets the minimum required for the operation (§27.8).
  5. Profile gate. The profile tier makes the operation available — some operations exist only at EAS or BAS (§27.8).

All five gates pass or the operation is rejected. There is no partial authorization. The authorization check itself produces an Authoritative audit record regardless of outcome.

§27.1.3 Truth Type Rules

All operations that produce or modify records are governed by truth type rules (§6):

Authoritative records are append-only. No field may be modified after creation. Corrections require supersession — a new record with supersedes_ref pointing to the original. The original is preserved unchanged.

Declared records are versioned. Modifications create new versions. The fields id, created_at, created_by, and truth_type are immutable across versions. Temporal bounds (effective_from, effective_to) are adjustable.

Derived records are ephemeral. They are recomputable from source data. Manual editing is prohibited. When source records change, dependent Derived records are marked stale and queued for recomputation.

Opaque records are boundary-enforced. Content behind an inspection boundary that cannot be crossed (§6.1). Opaque is system-assigned at boundary detection — no actor writes Opaque directly. Opaque records cannot be promoted through the standard Derived → Declared → Authoritative path; they transition to another truth type only when the inspection boundary is removed (§6.1, Opaque → Computed transition).

AI actors produce Derived only. All AI-generated content enters the substrate as Derived regardless of the AI actor's confidence level. Promotion to Declared requires human review. Promotion to Authoritative requires a separate human Decision with traceable Authority (B5). Truth type promotion is monotonic — no demotion path exists.

§27.1.4 Invariant Enforcement

Every write operation triggers zero or more invariant checks from the ten behavioral invariants (§5). Enforcement follows the two-pass validation strategy:

Pass 1 (SHACL Core). Single-node property checks at O(1). Invariants B1, B2, B3, B4, B5-immediate, B6-binding, and B8-core. Always Blocking — no profile may downgrade below Blocking.

Pass 2 (SHACL-SPARQL). Graph traversal and transitive closure checks. Invariants B5-T, B6-U, B8-routing, and B9. Profile-configurable enforcement mode — defaults vary by invariant and profile tier.

Schema Pass. Shapes-on-shapes validation. Invariant B7 (all objects flaggable). Runs at deployment time and schema changes, not per-operation.

Operations declare which invariants they touch. The invariant enforcement engine validates after the operation writes but before the transaction commits. If any Blocking invariant fails, the transaction rolls back.


§27.2 Tier 1 Primitive Operations

Tier 1 contains the nine irreducible core primitives (§4.1). Each primitive supports the full operation type range and participates in at least one behavioral invariant.

§27.2.1 Intent

Intent represents a stated objective — the answer to "what should happen?" An Intent lifecycle progresses through proposal, activation, and terminal resolution.

Table 27.2.1: Intent Operations

OperationTypePreconditionsPostconditionsInvariants
intent.createCreateDescription non-empty; created_by resolves to active Entity; authority_scope references active AuthorityStatus = draft; MVR fields written; truth type assigned (human: Authoritative/Declared; AI: Derived)B5
intent.readReadTarget existsRecord returned at requested projection
intent.updateUpdateStatus not terminal (achieved, abandoned, superseded); truth type permits modificationVersion incremented; modified fields updatedB5
intent.proposeTransitionStatus = draft; description non-empty; created_by resolvesStatus → proposed
intent.activateTransitionStatus = proposed; activator holds activation authorityStatus → activeB5
intent.pauseTransitionStatus = active; reason providedStatus → paused
intent.resumeTransitionStatus = pausedStatus → active
intent.blockTransitionStatus = active; blocker recorded with escalation timeStatus → blockedB7, B8
intent.unblockTransitionStatus = blocked; no active blockers remainStatus → active
intent.achieveTransitionStatus = active; all weighted success criteria metStatus → achieved; child Intents cascadeB4
intent.abandonTransitionStatus ∈ {active, paused, blocked}; Decision recordedStatus → abandoned; child Intents cascadeB4
intent.supersedeTransitionStatus not terminal; replacement Intent linkedStatus → supersededB4

Lifecycle. draftproposedactivepaused | blockedachieved | abandoned | superseded.

Semantic boundary: Intent declares what should happen. Commitment (§27.2.2) binds parties to do it. Work (§27.2.4) executes it. Achievement of an Intent is a governance conclusion, not a work-tracking status.

§27.2.2 Commitment

Commitment binds two or more parties to an obligation with terms and capacity allocation. The lifecycle governs negotiation, acceptance, performance, and breach.

Table 27.2.2: Commitment Operations

OperationTypePreconditionsPostconditionsInvariants
commitment.createCreateParties ≥ 2; terms ≥ 1; capacity allocation specifiedStatus = proposed; B2 verifiedB2, B5
commitment.readReadTarget existsRecord returned
commitment.updateUpdateStatus not terminal; truth type permits modificationVersion incremented; capacity re-validatedB2
commitment.negotiateTransitionStatus = proposedStatus → negotiating
commitment.amendCompositeStatus = negotiating; all parties agree to amendmentTerms updated; parties notifiedB2
commitment.acceptTransitionStatus ∈ {proposed, negotiating}; all parties have acceptedStatus → acceptedB4
commitment.activateTransitionStatus = accepted; all parties confirmed; capacity verifiedStatus → activeB2, B4, B5
commitment.fulfill_termTransitionTerm exists; Evidence of fulfillment providedTerm marked fulfilled; if all terms complete, status → fulfilledB3, B4
commitment.breachTransitionTerm breach detected; Evidence recordedBreach recorded; material breach → status = breachedB3, B4
commitment.suspendTransitionStatus = activeStatus → suspendedB4
commitment.resumeTransitionStatus = suspendedStatus → active
commitment.cureTransitionStatus = breached; within cure period; no active breaches remainStatus → activeB4
commitment.terminateTransitionGoverning authority exercisedStatus → terminatedB4, B5
commitment.expireTransitioneffective_until < now()Status → expired
commitment.supersedeTransitionReplacement Commitment linkedStatus → supersededB4

Lifecycle. proposednegotiatingacceptedactivesuspendedfulfilled | breachedterminated | expired | superseded. Cure transitions breachedactive.

§27.2.3 Capacity

Capacity represents a quantified resource — time, skill, budget, authority scope — that can be allocated to Commitments. The B2 invariant ensures Commitments never exceed available Capacity.

Table 27.2.3: Capacity Operations

OperationTypePreconditionsPostconditionsInvariants
capacity.createCreatetotal_available ≥ 0; owner resolves; measurement unit specifiedStatus = available; allocated = 0B5
capacity.readReadTarget existsRecord returned with current allocation state
capacity.updateUpdateStatus not terminalVersion incremented; available recomputedB2
capacity.allocateTransitionRequested amount ≤ (available − reserved)Allocation recorded; available decrementedB2
capacity.releaseTransitionAllocation existsAllocated decremented; available incrementedB2
capacity.check_sufficientReadTarget existsReturns boolean sufficiency result
capacity.exhaustTransitionAvailable = 0 (system-triggered)Status → exhaustedB2
capacity.blockTransitionStatus ∈ {available, partially_available}Status → blocked
capacity.replenishTransitiontotal_available updatedStatus recalculated from new totalsB2

Lifecycle. availablepartially_availableexhausted | blocked | reserved | replenishing.

§27.2.4 Work

Work represents effort expended in service of a Commitment. B1 enforces the structural link — every Work instance must reference exactly one Commitment. Work Specifications (SM-4) govern how AI-composed work is confirmed and promoted.

Table 27.2.4: Work Operations

OperationTypePreconditionsPostconditionsInvariants
work.createCreatecommitment_id references active CommitmentStatus = defined; B1 link establishedB1, B5
work.readReadTarget existsRecord returned
work.updateUpdateStatus not terminalVersion incrementedB1
work.assignTransitionStatus = ready; assignee within delegation boundsStatus → assignedB5
work.startTransitionStatus = assignedStatus → in_progressB1
work.blockTransitionStatus = in_progress; blocker recordedStatus → blockedB7, B8
work.unblockTransitionStatus = blocked; no active blockersStatus → in_progress
work.submit_for_reviewTransitionStatus = in_progress; outputs presentStatus → review
work.acceptTransitionStatus = review; reviewer holds authorityStatus → acceptedB4, B5
work.rejectTransitionStatus = review; rationale providedStatus → rejectedB4
work.reworkTransitionStatus = rejectedStatus → in_progress
work.completeTransitionStatus = accepted; Evidence createdStatus → completedB3, B4
work.abandonTransitionDecision recordedStatus → abandonedB4
work.cancelTransitionGoverning authority exercisedStatus → cancelledB4, B5
work.supersedeTransitionReplacement Work linkedStatus → superseded

Lifecycle. definedreadyassignedin_progressblockedreviewaccepted | rejectedcompleted | abandoned | cancelled | superseded. Rejection loops back to in_progress via work.rework.

Table 27.2.5: Work Specification Operations (SM-4)

OperationTypePreconditionsPostconditionsInvariants
work_spec.composeCreateDelegation active; AI actor within boundsStatus = draft; truth type = DerivedB1, B3
work_spec.confirmTransitionHuman reviewer at Stage A/B, or elevated risk at Stage CStatus → confirmed; Decision + Account createdB4
work_spec.skip_confirmTransitionStage C only; consequence level = reversibleStatus → confirmed (bypasses human review)B4
work_spec.emitTransitionStatus = confirmedStatus → active; emitted to execution layer
work_spec.completeTransitionAll outputs presentStatus → completed; outputs = Derived EvidenceB3
work_spec.promoteTransitionHuman with traceable Authority; AI-Cannot-Be-PrincipalOutputs promoted from DerivedB3, B4, B5
work_spec.rejectTransitionHuman reviewer; rationale mandatoryStatus → rejectedB4

Lifecycle (SM-4). draftconfirmedactivecompletedpromoted | rejected. Stage C low-risk work may skip confirmation via work_spec.skip_confirm.

§27.2.5 Evidence

Evidence records factual artifacts — observations, measurements, documents, validation reports. Evidence is immutable after creation. The B3 invariant requires every Evidence instance to carry exactly one truth type.

Table 27.2.6: Evidence Operations

OperationTypePreconditionsPostconditionsInvariants
evidence.createCreateSource identified; truth type assigned; confidence_level ∈ [0,1]; human: Authoritative/Declared; AI: Derived onlyRecord created; immutable after creationB3, B5
evidence.readReadTarget existsRecord returned at fidelity level L0–L3
evidence.verifyCompositeEvidence exists; verification criteria appliedVerification record appended; state transitions through Unverified → VerificationPending → Verified/VerificationFailed/CannotVerifyB3
evidence.supersedeCompositeOld Evidence preserved; new record linked via supersedes_refNew record created; old unchangedB3, B4
evidence.promoteTransitionDerived → Declared or Declared → Authoritative (monotonic only); human with traceable Authority; AI-Cannot-Be-PrincipalTruth type promoted; Decision + Account createdB3, B4, B5

Immutability rule. Evidence records cannot be updated. Verification adds records — it does not modify the Evidence itself. Corrections require supersession: a new Evidence record that links to and replaces the original. The original is preserved.

§27.2.6 Decision

Decision records a governance choice — the answer to "what was chosen and why?" Every Decision creates an Account record (B4). AI actors may produce Decision proposals as Derived content; only humans may finalize governance Decisions.

Table 27.2.7: Decision Operations

OperationTypePreconditionsPostconditionsInvariants
decision.createCreatedecision_by resolves; authority_source traceable (B5); options_considered ≥ 2 for selection-type; rationale non-empty; human: Authoritative; AI: Derived only (proposals)Status = final; Account createdB4, B5
decision.readReadTarget existsRecord returned with deviation data
decision.reverseTransitionStatus = final; within reversal_window; reversible = trueStatus → reversed; new Decision created for the reversalB4
decision.supersedeTransitionReplacement Decision linkedStatus → supersededB4
decision.expireTransitionTemporal bound exceeded (system-triggered)Status → expired
decision.compute_deviationReadDecision exists; deviation metrics definedDeviation vector computed: ‖Δ‖ = √(Σ(wᵢ · Δᵢ²)) + penalty(Δ_C); escalation if above threshold

Lifecycle. draftfinalsuperseded | reversed | expired.

Decision zones govern AI participation. Zone 1 (Green): AI may decide autonomously at Stage C, logged, reversible, within delegation. Zone 2 (Yellow): AI prepares, human reviews within window. Zone 3 (Orange): human decides with AI support. Zone 4 (Red): human only — AI may not prepare.

§27.2.7 Authority

Authority represents the right to act — delegated or root. The B5 invariant requires every Authority to be traceable through a delegation chain terminating at a root authority. Authority is the foundation of the authorization model.

Table 27.2.8: Authority Operations

OperationTypePreconditionsPostconditionsInvariants
authority.createCreateRoot: explicit designation; Delegated: parent scope is superset; B5 chain validStatus = active; chain establishedB5
authority.readReadTarget existsRecord returned with chain metadata
authority.delegateCompositeDelegator holds sufficient scope; scope subset verified; creates Authority + Delegation + Decision + AccountNew Authority created; B5 chain extendedB4, B5
authority.revokeTransitionRevoking authority holds governing scopeStatus → revoked; cascade-revokes all downstream delegations; active signals re-routedB4, B5, B8
authority.suspendTransitionSuspending authority holds governing scopeStatus → suspended; in-flight work paused; active signals re-routedB4, B5, B8
authority.reinstateTransitionStatus = suspended; reinstating authority verifiedStatus → activeB4, B5
authority.reviewCompositeReview triggered (scheduled, expiry, audit)Review record created; may trigger revocation or renewalB4, B5
authority.expireTransitionTemporal bound exceeded (system-triggered)Status → expired
authority.can_exerciseReadActor and target specifiedReturns {authorized, reason, authority_ref}
authority.resolve_chainReadAuthority specifiedReturns complete delegation chain to root

Lifecycle. pendingactivesuspendedexpired | revoked | superseded.

Table 27.2.9: Delegation Operations (SM-5)

OperationTypePreconditionsPostconditionsInvariants
delegation.proposeCreateDelegator identified; scope specified; AI-Cannot-Be-PrincipalStatus = proposedB5
delegation.grantTransitionHuman approves; scope verified as subset of parentStatus → active; Decision + Account createdB4, B5
delegation.modifyTransitionHuman only; AI cannot modify own delegation; scope change verifiedStatus → modified; tighten-only cascade propagatedB4, B5
delegation.confirm_modificationTransitionAffected parties confirmModification finalizedB4
delegation.suspendTransitionSuspending authority holds governing scopeStatus → suspendedB4
delegation.reinstateTransitionStatus = suspendedStatus → activeB4
delegation.revokeTransitionRationale mandatory; governing authority exercisedStatus → revoked (terminal); signals routed through this delegation re-routedB4, B5, B8
delegation.expireTransitionTemporal bound exceeded; grace period for in-flight workStatus → expired (terminal); signals routed through this delegation re-routedB8

Lifecycle (SM-5). proposedactivemodified | suspendedrevoked | expired.

AI-Cannot-Be-Principal applies to all delegation writes. AI actors cannot propose, grant, modify, or revoke delegations. AI actors may query authority.can_exercise and authority.resolve_chain to inspect delegation state.

§27.2.8 Account

Account records per-event linkage between actors and their governance actions. Every Decision creates an Account entry (B4). Account records are always Authoritative and immutable.

Table 27.2.10: Account Operations

OperationTypePreconditionsPostconditionsInvariants
account.createCreateactor_id resolves; action_type specified; authority_chain capturedRecord created; truth type = Authoritative; immutableB4, B5
account.readReadTarget existsRecord returned as Authoritative
account.post_transactionCompositeAccount exists; transaction_type specified; evidence_ref linked; authorized_by verifiedTransaction appended; immutable; truth type = AuthoritativeB4
account.reconstruct_atReadAccount exists; target timestamp specifiedTransaction log replayed through target time
account.reconcileCompositeReconciliation initiated; all transactions reviewedReconciliation record created; reconciliation itself is a Decision (B4)B4
account.closeTransitionClosing authority exercisedNo new transactions acceptedB4, B5

Account is per-event linkage, not aggregation. Each Account record links one actor to one governance action. Temporal framing — "what was Actor X accountable for during Period Y?" — uses Cycle queries (§28). Account records are never modified or deleted.

§27.2.9 Constraint

Constraint represents a binding rule — policy, compliance requirement, physical limitation, or organizational boundary. B6 requires every Constraint to target at least one primitive instance and specify an enforcement mode.

Table 27.2.11: Constraint Operations

OperationTypePreconditionsPostconditionsInvariants
constraint.createCreateapplies_to non-empty; enforcement_mode specified; authority_source traceable; SHACL shape URI linkedStatus = activeB6, B5
constraint.readReadTarget existsRecord returned with current evaluation state
constraint.updateUpdateapplies_to cannot be emptied; enforcement_mode cannot be removedVersion incrementedB6
constraint.evaluateReadTarget Constraint and governed instances specifiedReturns: satisfied / violated / exception_applied / not_applicableB6
constraint.grant_exceptionCompositeOverride protocol (§5.5): overriding actor holds Authority governing the violated primitive; creates Decision + Evidence + AccountException recorded; override Decision = Declared (not Authoritative)B4, B5, B6
constraint.revoke_exceptionTransitionException exists; revoking authority verifiedConstraint re-enforced on targetB6
constraint.suspendTransitionStatus = activeStatus → suspendedB4
constraint.deprecateTransitionStatus ∈ {active, suspended}; optional link to replacementStatus → deprecatedB4

Lifecycle. activesuspendeddeprecated.

Override protocol (§5.5). Exceptions follow a four-step protocol: (1) the overriding actor must hold Authority governing the violated primitive, (2) the override is itself a Decision requiring an Account (B4), (3) Evidence is produced as Declared truth (not Authoritative), and (4) the original violation record is preserved — overrides are append-only.


§27.3 Tier 2 Infrastructure Operations

Tier 2 contains four unconditional infrastructure primitives (Identifier, Entity, Context, Namespace) that exist in every substrate deployment regardless of profile tier (§4.5), plus Actor lifecycle operations (SM-1). These primitives support the Tier 1 governance model but do not carry MVR fields.

§27.3.1 Identifier

Identifier provides unique naming and resolution services. System-generated UUIDs carry Authoritative truth type. Human-supplied external references carry Declared truth type.

Table 27.3.1: Identifier Operations

OperationTypePreconditionsPostconditionsInvariants
identifier.createCreateUnique within (instance, type, value)Record created; truth type assigned
identifier.readReadTarget existsRecord returned
identifier.resolveReadIdentifier value and type specifiedReturns bound primitive reference or null
identifier.versionTransitionIdentifier exists; new version justifiedVersion incremented; prior version preserved
identifier.bindCompositeTarget primitive exists; no conflicting bindingIdentifier bound to primitive instance

§27.3.2 Entity and Actor

Entity represents a participant — human, organization, role, or system. Actor is the governance identity that an Entity holds within a substrate instance. Actor lifecycle operations (SM-1) govern registration, activation, suspension, and deactivation.

Table 27.3.2: Entity Operations

OperationTypePreconditionsPostconditionsInvariants
entity.createCreateentity_type ∈ {human, organization, role, system}Record created; entity_type immutable after creationB5
entity.readReadTarget existsRecord returned
entity.updateUpdateentity_type immutable; other fields modifiableVersion incremented
entity.deactivateTransitionNo active sole-Accountable rolesStatus → inactiveB4
entity.archiveTransitionStatus = inactiveStatus → archived (terminal)

Table 27.3.3: Actor Operations (SM-1)

OperationTypePreconditionsPostconditionsInvariants
actor.registerCreateEntity exists; actor_type ∈ {human, ai_automation, ai_agentic, ai_assistive}Status = registered; Actor Context createdB5
actor.activateTransitionStatus = registered; activation authority verifiedStatus → active; Decision + Account createdB4, B5
actor.suspendTransitionStatus = active; suspending authority verifiedStatus → suspendedB4
actor.reinstateTransitionStatus = suspended; reinstating authority verifiedStatus → activeB4
actor.deactivateTransitionAI: all delegations revoked; Human: all roles transferredStatus → deactivatedB4
actor.archiveTransitionStatus = deactivatedStatus → archived (terminal)B4

Lifecycle (SM-1). registeredactivesuspendeddeactivatedarchived.

Actor Context operations. Each actor holds an Actor Context per instance they participate in (§22.1). Actor Context supports actor_context.create, actor_context.update, and actor_context.delete. Deletion is blocked if the actor is sole Accountable for active Decisions in the instance.

§27.3.3 Context

Context provides situational framing for governance actions. Six context types — temporal, spatial, relational, situational, interpretive, and composite — classify the environment in which primitives operate.

Table 27.3.4: Context Operations

OperationTypePreconditionsPostconditionsInvariants
context.createCreatecontext_type specified; relevant fields populatedRecord created
context.readReadTarget existsRecord returned
context.updateUpdateStatus = activeVersion incremented
context.deactivateTransitionNo dependent active primitives require this ContextStatus → inactive
context.scopeReadContext and target primitive specifiedReturns scoping applicability

§27.3.4 Namespace

Namespace organizes primitives into hierarchical scoping domains. Four scopes — core, domain, tenant, and sandbox — govern vocabulary and concept ownership. Sandbox namespaces carry mandatory expiration.

Table 27.3.5: Namespace Operations

OperationTypePreconditionsPostconditionsInvariants
namespace.createCreatescope ∈ {core, domain, tenant, sandbox}; sandbox requires expirationStatus = proposedB5
namespace.readReadTarget existsRecord returned
namespace.updateUpdatescope immutable; other fields modifiableVersion incremented
namespace.graduateTransitionGraduation Decision (B4); sandboxtenantdomaincore (monotonic)Status → graduated; scope upgradedB4
namespace.expireTransitionSandbox: expiration exceededStatus → expired
namespace.retireTransitionRetirement Decision recordedStatus → retiredB4
namespace.bindCompositeTarget primitive exists; namespace activePrimitive bound to namespace

Lifecycle. proposedunder_reviewactivegraduated | expired | retired.


§27.4 Tier 3 Governed Operations

Tier 3 primitives — Orientation, Learning, and Activation — support deliberate governance posture and institutional adaptation. They activate at graduation Stage B (§4.7). All Tier 3 write operations require elevated governance authority.

§27.4.1 Orientation

Orientation declares the pre-decisional framing — values, boundaries, and harm prohibitions — that governs an entity's participation. Only one Orientation may be active per owner. AI cannot create or modify Orientation (L4 gradient level — highest deliberation).

Table 27.4.1: Orientation Operations

OperationTypePreconditionsPostconditionsInvariants
orientation.createCreateOnly one active per owner; harm_boundary.absolute_prohibitions non-empty; human only (AI-Cannot-Be-Principal)Record created; previous Orientation superseded if existsB5
orientation.readReadTarget existsRecord returned
orientation.updateUpdateFormal governance process only (L4); human onlyVersion incrementedB4, B5
orientation.supersedeTransitionReplacement Orientation linked; governance Decision recordedStatus → supersededB4, B5
orientation.deactivateTransitionOwner deactivated or archivedStatus → inactiveB4

§27.4.2 Learning

Learning captures institutional adaptation signals — pattern observations, validations, threshold adjustments, and structural reconfigurations. Four governance levels (L1–L4) determine the deliberation required for application.

Table 27.4.2: Learning Operations

OperationTypePreconditionsPostconditionsInvariants
learning.createCreatelearning_type specified; governance_level ∈ {L1, L2, L3, L4}Status = detectedB3
learning.readReadTarget existsRecord returned
learning.analyzeTransitionStatus = detected; analysis criteria appliedStatus → analyzed
learning.applyTransitionStatus = analyzed; Decision required (B4); human only for L3–L4Status → appliedB4
learning.rejectTransitionStatus = analyzed; rationale mandatory; human onlyStatus → rejectedB4
learning.monitorTransitionStatus = appliedStatus → monitoring

Lifecycle. detectedanalyzedappliedmonitoring | rejected.

§27.4.3 Activation

Activation governs the six-stage governance pipeline (§12): discover, analyze, implement, evidence, verify, sustain. Each stage produces specific primitive outputs and enforces stage-appropriate invariants.

Table 27.4.3: Activation Operations

OperationTypePreconditionsPostconditionsInvariants
activation.createCreategovernance_stage ∈ {discover, analyze, implement, evidence, verify, sustain}Status = pendingB6
activation.readReadTarget existsRecord returned with stage progress
activation.readyTransitionStatus = pending; prerequisites metStatus → ready
activation.activateTransitionStatus = readyStatus → activeB6
activation.blockTransitionStatus = active; blocker recordedStatus → blockedB7, B8
activation.unblockTransitionStatus = blocked; no active blockersStatus → active
activation.completeTransitionStatus = active; stage outputs producedStatus → completeB4, B6
activation.advance_stageCompositeCurrent stage complete; A and G roles confirmed; next stage requirements projectedStage advanced; new Activation created for next stageB4, B5, B6

Status lifecycle. pendingreadyactiveblockedcomplete.

Six-stage pipeline (§12). discoveranalyzeimplementevidenceverifysustain. Stage advancement requires both Accountable (A) and Governor (G) role confirmation at EAS. BAS supports a subset of stages.


§27.5 Tier 4 AI-Native Operations

Tier 4 primitives — Interpretation and Environment Interface — extend the substrate with AI-specific governance capabilities. They activate at graduation Stage B and are available at EAS and BAS only (§4.7).

§27.5.1 Interpretation

Interpretation captures AI-generated meaning-making — classifications, extractions, recommendations, analyses, predictions, and gap identifications. All Interpretations enter as Derived (B3). Resolution requires human action.

Table 27.5.1: Interpretation Operations

OperationTypePreconditionsPostconditionsInvariants
interpretation.createCreateinterpretation_type specified (14 types); stakes × novelty routing evaluated; truth type = DerivedRecord created; red-zone guard applied for high-stakes × high-noveltyB3, B5
interpretation.readReadTarget existsRecord returned with provenance and routing metadata
interpretation.proposeTransitionStatus = pending; AI actor within delegationStatus → agent_proposedB3
interpretation.submit_for_reviewTransitionStatus = agent_proposedStatus → human_review
interpretation.resolveTransitionStatus = human_review; human only (AI-Cannot-Be-Principal)Status → resolved; Decision + Account createdB4, B5
interpretation.disputeTransitionStatus = resolved; dispute Evidence providedStatus → disputedB3, B4
interpretation.withdrawTransitionStatus ∈ {pending, agent_proposed}Status → withdrawn

Lifecycle. pendingagent_proposedhuman_reviewresolved | disputed | withdrawn.

Stakes × novelty routing matrix (§4.6.3). Low × Low: Citation Fidelity (verification). Low × High: Pattern Anomaly Flag (escalation). High × Low: Reasoning Integrity (logic review). High × High: Judgment Alignment (expert interpretation) — red zone. Red-zone guard rejects promotions without evidence of substantive review (rubber-stamp detection).

§27.5.2 Environment Interface

Environment Interface connects the substrate to external systems — inbound signals, outbound actions, and feedback loops. Inbound signals enter as Derived. Outbound actions require Authority.

Table 27.5.2: Environment Interface Operations

OperationTypePreconditionsPostconditionsInvariants
env_interface.createCreateinterface_type ∈ {inbound_signal, outbound_action, feedback_loop}; governance authority verifiedStatus = activeB5
env_interface.readReadTarget existsRecord returned
env_interface.updateUpdateStatus = activeVersion incremented
env_interface.receive_signalCompositeInterface active; inbound signal receivedInterpretation created (Derived); routed via Ontology Alignment (§19)B3, B7, B8
env_interface.send_actionCompositeInterface active; Authority verified; outbound action authorizedAction dispatched; Evidence recordedB4, B5
env_interface.close_loopCompositeFeedback data received; both inbound and outbound components existLearning signal generated; loop record updatedB3
env_interface.dormantTransitionStatus = active; inactivity threshold exceededStatus → dormant
env_interface.closeTransitionStatus ∈ {active, dormant}; closing authority exercisedStatus → closed (terminal)B4

Lifecycle. activedormantclosed.


§27.6 Tier 5 Configuration Operations

Tier 5 contains a single primitive — Cycle — that provides temporal configuration for governance rhythms. Cycle is available at EAS and BAS only.

§27.6.1 Cycle

Cycle defines recurring governance cadences — execution pulses, activation sweeps, weekly reviews, learning integrations, orientation checks, and commitment reconciliations. Every cycle's decision step requires sense_maker = human.

Table 27.6.1: Cycle Operations

OperationTypePreconditionsPostconditionsInvariants
cycle.createCreatecycle_type ∈ 6 types; cadence configured; human onlyRecord createdB5
cycle.readReadTarget existsRecord returned with cadence and dependency data
cycle.configureUpdateDecision step must retain sense_maker = humanConfiguration updated; dependent cycles notifiedB4
cycle.startTransitionStatus = scheduled; prerequisites metStatus → running
cycle.completeTransitionStatus = running; all outputs producedStatus → completedB4
cycle.skipTransitionStatus = scheduled; skip Decision recordedStatus → skippedB4
cycle.failTransitionStatus = running; failure condition detectedStatus → failed; Signal createdB4, B7
cycle.query_boundariesReadCycle exists; temporal query specifiedReturns cycle boundaries for Account temporal framing

Lifecycle. scheduledrunningcompleted | skipped | failed.


§27.7 Cross-Cutting Operations

Ten cross-cutting operation domains define composite operations that span multiple primitives. Six primary domains — signals (§14–§15), IQs (§16), VPs (§17), governance activation (§12), orchestration (§A1–§A2), and KPIs — address governance concerns that no single primitive handles alone. Four additional domains — profile/portfolio management, graduation, agent behavioral cycles, and ontology alignment — provide lifecycle and integration operations.

§27.7.1 Signal Operations

Signals implement the algedonic channel — the mechanism for surfacing problems at any governance level. B7 guarantees that all nineteen primitive classes have a signal attachment surface. B8 guarantees that every signal routes to an authority on the governance chain.

Table 27.7.1: Signal Operations

OperationTypeComposesInvariants
signal.createCompositeSignal + Evidence (Authoritative) + Context (situational)B7
signal.create_reservationCompositeSignal with reservation nature (7 types: concern, uncertainty, disagreement, precedent/scope/timing/resource question); does not block governance actionB7
signal.routeCompositeResolves authority chain; routes to immediate governing authority or any ancestor at any depthB8
signal.rerouteCompositeMust stay on governance chain; routing history preservedB8
signal.escalateCompositeNext higher authority on chain; emergency signals bypass intermediate hierarchyB8
signal.resolveCompositeCreates resolution Decision + Account; may produce Work, Decision, Evidence, or IQB4, B8
signal.dismissCompositeCreates dismissal Decision with mandatory rationaleB4
signal.monitorTransitionAcknowledged; ongoing monitoring with periodic Evidence updates

Lifecycle (SM-7). createdroutedprocessingescalatedmonitoringresolved | dismissed.

Any actor may raise a signal. Resolution and dismissal require a Decision (B4) by an actor with authority over the flagged object.

§27.7.2 IQ Operations

Investigative Queries operationalize emergence — the process by which questions arising from governance activity become institutional knowledge. B9 guarantees that every resolved IQ produces a Decision or Work item.

Table 27.7.2: IQ Operations

OperationTypeComposesInvariants
iq.createCreate8 intent types (explore, validate, disambiguate, fill_gap, challenge, connect, branch_inventory, confidence_probe); 4 origin types (claim, signal, capture, workflow)
iq.create_from_signalCompositeSignal-spawned IQ inherits routing contextB8
iq.create_from_claimCompositeClaims review reveals uncertainty
iq.create_branch_inventoryCompositeSecond-order lineage extension: enumerates open decision branches
iq.create_confidence_probeCompositeSecond-order lineage extension: evaluates if accumulated evidence shifts original confidence
iq.assignTransitionAssignment methods: manual, auto-routed, inherited
iq.investigateCompositeCreates investigation Evidence records and research Work itemsB1, B3
iq.resolveCompositeMust produce Decision or Work item (B9); Decision requires Account (B4)B4, B9
iq.promoteCompositeResolved findings promoted to permanent knowledgeB9
iq.abandonCompositeDecision to not act is itself a Decision — satisfies B9B4, B9
iq.supersedeTransitionReplacement link established
iq.revertTransitionresolvedopen; new evidence contradicts resolution; new resolution eventually requiredB9

Lifecycle (SM-8). openassignedinvestigatingresolvedpromoted | abandoned | superseded. Reversion: resolvedopen.

Zero friction, globally accessible. IQ creation has no authorization gate beyond instance membership. The cost of a question is always lower than the cost of an unasked question.

§27.7.3 VP Operations

Validated Projections materialize derived views from source primitives — plan baselines, condition assessments, and predictions. All VPs carry truth type Derived (B3). No human approval is required for materialization.

Table 27.7.3: VP Operations

OperationTypeComposesInvariants
vp.materializeCompositeSource primitives queried; derivation method applied; confidence computed; truth type = DerivedB3
vp.materialize_plan_baselineComposite4 modes: explicit, inferred, normative, comparative; TTL varies by modeB3
vp.materialize_condition_assessmentCompositeRequires current Plan Baseline; compares current vs. expected across 7 dimensionsB3
vp.materialize_predictionCompositeRequires current Condition Assessment; model_ref for reproducibilityB3
vp.mark_staleTransitionEvent-based (source changed) or time-based (valid_until exceeded)
vp.recomputeCompositeMust not change source data; must produce identical output given identical inputs; must update provenanceB3
vp.invalidateTransitionRecomputation fails; cascade to dependent VPs
vp.access_staleCompositeDemand-based synchronous recomputation triggeredB3

Cache states (SM-9). currentstalerecomputingcurrent | invalidated.

Invalidation cascade. Plan Baseline invalidation immediately invalidates dependent Condition Assessments, which immediately invalidates dependent Predictions. Staleness propagates differently — stale records may be recomputed; invalidated records cannot.

VP disposition routing. When a Condition Assessment reveals deviation, the disposition determines the governance response: investigate (creates Work), re-evaluate plan (creates Decision + recomputation), compensate (creates Constraint exception), acknowledge (creates Evidence + scheduled review), escalate (routes Decision by Authority), accept risk (creates Learning + Decision), remediate (creates corrective Work), defer (schedules future assessment), or no action (audit log entry only).

§27.7.4 Governance Activation Operations

Governance Activation implements the six-stage pipeline (§12) that discovers, projects, and closes governance gaps. Each stage is a composite operation producing specific primitive outputs.

Table 27.7.4: Governance Activation Operations

OperationStageProducesInvariants
gov_activation.discover_profile1Evaluates ~200 pattern signatures against organizational attributes
gov_activation.activate_sources2Evaluates trigger conditions (ANY-OF, ALL-OF Boolean); tier-aware gating: Stage A = Tiers 1–2, Stage B adds Tier 3, Stage C adds Tier 4B6
gov_activation.project_requirements3Projects verification types (V-EXIST through V-RATIFIED); criticality ordering: conservation-critical > invariant-enforcing > standard > advisoryB6
gov_activation.project_gaps4Compares required vs. existing verification types; gap inventory with criticality ordering
gov_activation.create_closure_roadmap5Creates Work items, Commitments, verifies Capacity; closure stages: Awareness → Enablement → Management → GovernanceB1, B2, B6
gov_activation.propagate_changes6Governance posture updated; Learning signals for institutional adaptationB6

§27.7.5 Orchestration Operations

The seven-step orchestration engine (§A1, §A2) classifies cognitive work, activates constraints, routes to actors, selects within delegation bounds, composes Work Specifications, confirms with human oversight, and emits to the execution layer.

Table 27.7.5: Orchestration Operations

OperationStepActionInvariants
orchestration.classify1AICAR decomposition: A-Citation, B-Reasoning, C-Context, D-CommunicationB4
orchestration.activate_constraints2Five-dimensional context: AICAR type × Decision type × Consequence level × Operating posture × Graduation stageB6
orchestration.route3Filters actor registry; matches capabilities within delegation boundsB5
orchestration.select4Verifies delegation active (SM-5); autonomy level sufficientB4, B5
orchestration.compose5Assembles Work Specification with composition strategy (sequential, parallel, iterative, conditional, delegated, federated)B1
orchestration.confirm6Required at Stage A/B (always), Stage C when elevated risk; AI-Cannot-Be-PrincipalB4
orchestration.emit7Emitted to execution layer

AICAR classification operations. aicar.classify_work classifies cognitive work into A/B/C/D components, supporting composite classifications (A+B+C). aicar.extend_type registers new derivative types under human approval; derivatives inherit parent properties and cannot modify them (taxonomy closure).

Delegation-as-Composition. delegation.compose is a composite operation that assembles Intent + Authority + Commitment + Constraint + Work + Context into a coherent delegation envelope. AI-Cannot-Be-Principal. B1/B2/B5 enforced.

§27.7.6 KPI Operations

KPI operations govern the introduction, retirement, and monitoring of governance health indicators. Three KPI classes — Invariant Signal (IS), Control KPI (CKPI), and Intent Progress Indicator (IPI) — serve distinct governance rhythms.

Table 27.7.6: KPI Operations

OperationTypeActionInvariants
kpi.introduceCompositeCreates KPI with class (IS/CKPI/IPI), trigger type (Intent-driven, Constraint-driven, Emergent); Decision requiredB4
kpi.retireComposite4 reasons: intent advanced, phase complete, constraint resolved, replaced; learning_summary mandatory; no KPI may silently disappearB4
kpi.monitor_invariant_signalsRead8 monitoring dimensions: cycle time distribution, quality/rework rate, variance/predictability, escalation frequency, judgment load, evidence completeness, semantic integrity, change failure rate
kpi.review_implicationsCompositeMandatory when Intent changes; reviews downstream KPI relevanceB4
kpi.validate_patternCompositePattern must be promoted Derived → Declared before influencing routingB3

§27.7.7 Profile and Portfolio Operations

Profile operations govern instance creation and graduation across the three profile tiers (§21, §23). Portfolio operations manage recursive parent-child instance relationships with tighten-only constraint monotonicity.

Table 27.7.7: Profile and Portfolio Operations

OperationTypeActionInvariants
profile.create_instanceCompositeCreates substrate instance with profile type (EAS/BAS/PAS); content packages loaded; RACIVG scope set (EAS = 6 roles, BAS = 4, PAS = 2); tier activation determinedB5, B6
profile.trigger_graduationCompositeTriggers: revenue, recurring customers, investment, employment, unbounded drift, type-specific thresholds
profile.evaluate_graduationCompositeHuman-initiated evaluation; evidence gatheredB4
profile.decide_graduationCompositeGO/NO-GO Decision; BAS → EAS requires parent concurrenceB4, B5
profile.execute_graduationCompositeEvidence carries forward with original truth types and lineage preserved; source instance archives; new instance establishes own RACIVG authority structure at genesis — BAS → EAS adds OVERSIGHT (V) and ADVISORY (G) authority scopesB5, B9
portfolio.spawn_instanceCompositeChild constraints = intersection of all ancestors (tighten-only); graduation ceiling = parent's stage; AI-Cannot-Be-PrincipalB5, B6
portfolio.tighten_constraintCompositeChild tightens parent position (always permitted)B6
portfolio.relax_constraintCompositeFive-step evidence-driven process; surfaces as Signal to constraining authority; recursive upward if authority itself constrainedB5, B6, B8
portfolio.depth_gated_readReadOwn data: full; direct children: full; deeper: aggregate only; siblings/cousins: none; parent operational: none

§27.7.8 Graduation Operations

Graduation operations manage the per-scope AI governance maturity model (§8, §A1, SM-2). Graduation is per-scope — each AICAR type × Decision type × domain namespace has an independent graduation stage.

Table 27.7.8: Graduation Operations

OperationTypeActionInvariants
graduation.advanceCompositeStage criteria evaluated (interaction counts, pattern validation, confirmation rates, sustained accuracy); no stage skipping; Decision + Account createdB4
graduation.auto_regressCompositeSystem-triggered on: rejection rate spike, escalation rate spike, routing accuracy degradation, deviation detection failure; drops to immediate predecessor onlyB4
graduation.human_regressCompositeHuman-initiated; may drop to any lower stage; Decision + Account createdB4

Stage criteria thresholds. Stage A → A+: 50+ interactions, 30+ days, 5+ contexts, 50+ human reviews. Stage A+ → B: 100+ interactions, validated patterns promoted to Declared, 2+ actors. Stage B → C: 85%+ confirmation rate, 90+ days sustained. Stage C maintenance: sustained accuracy, low rejection rate, stable escalation frequency.

§27.7.9 Agent Behavioral Cycle Operations

The agent behavioral cycle (SM-6) governs how AI actors operate within the substrate. Three tracks — monitoring (Stage A), navigation (Stage A+/B), and execution (Stage C) — are gated by graduation stage.

Table 27.7.9: Agent Behavioral Cycle Operations

OperationTypeTrackInvariants
agent.context_loadCompositeAll tracksLoads Actor Context, delegation, Constraints, Work Specifications; B5/B6 validated
agent.navigateCompositeNavigation (A+/B)Explores decision space; produces recommendations as Derived
agent.monitorCompositeMonitoring (all stages)Evaluates Constraints; checks 80% threshold proximity; blocking violation → ALERTING (preempts all other tracks)
agent.executeCompositeExecution (C only)Requires autonomy ≥ 2 and active Work Specification; all outputs = Derived
agent.alertCompositeMonitoring → AlertingSurfaces deviation to human; creates Signal; returns to LISTENING

Cycle states (SM-6). LISTENING → CONTEXT_LOADING → NAVIGATING | MONITORING | EXECUTING → PROPOSING | ALERTING | COMPLETING → LISTENING. The cycle is non-terminating — agents return to LISTENING after completing each pass.

§27.7.10 Ontology Alignment Operations

Ontology alignment operations (§18, §19) manage the Knowledge Integration pipeline that maps external domain concepts to the substrate's nineteen-primitive model.

Table 27.7.10: Ontology Alignment Operations

OperationTypeActionInvariants
ontology.preloadCompositeDeploys concept signatures, bridges, and source registry entries at deployment time; truth type = Declared; bypasses graduation
ontology.lazy_discoverCompositeRuntime discovery; sandbox namespace entry; graduation path: sandbox → tenant → domain → coreB3
ontology.alignCompositeEight-stage KI pipeline: External Field Scan → Lexical Matcher → Structural Matcher → Semantic Matcher → Confidence Aggregator → Claim Writer → Review Queue → Graduation Engine; three independent matchers for consensusB3
ontology.bridgeCompositeBridge types: Equivalent, Subsumes, Specializes, Overlaps, Complements; confidence decay per typeB3
ontology.graduate_conceptCompositeTwo-stage graduation: Stage A (normalization gates) + Stage B (multi-dimensional scoring); tier-weighted thresholdsB4

§27.8 Authorization Matrix

The authorization matrix maps every write operation to actor type, minimum RACIVG role, minimum graduation stage, and profile tier availability. This section provides the consolidated summary.

§27.8.1 Actor Type Permissions

Four actor types (§22.2) participate in operations with distinct permission boundaries. Permission values: P = Permitted, X = Prohibited, C = Conditional (conditions specified per operation).

Table 27.8.1: Actor Type Permission Summary

Operation CategoryHumanAI-AutomationAI-AgenticAI-Assistive
Create root AuthorityPXXX
Delegate/revoke AuthorityPXXX
Create/modify OrientationPXXX
Create/modify ConstraintsPXXX
Grant Constraint exceptionsPXXX
Accept/activate CommitmentsPXXX
Finalize Decisions (Zone 3–4)PXXX
Accept/complete Work (final)PXXX
Resolve InterpretationsPXXX
Promote truth typesPXXX
Hold Accountable (A) rolePXXX
Create Evidence (Derived)PP (system events)C (within delegation)C (within delegation)
Create Zone 1 DecisionsPXC (Stage C, reversible)X
Execute WorkPC (deterministic)C (Stage C, autonomy ≥ 2)C (interactive)
Allocate CapacityPXC (Stage C, reversible)X
Read any primitivePPPP
Resolve identifiers/chainsPPPP
Evaluate ConstraintsPPPP
Create SignalsPPPP
Create IQsPPPP
Detect Learning signalsPPPX

§27.8.2 RACIVG Role Requirements

Operations that modify governance state require specific RACIVG roles. The role requirement varies by profile tier: EAS uses all six roles (RACIVG), BAS uses four (RACI), PAS uses two (Owner/Collaborator mapped to R and A).

Table 27.8.2: RACIVG Minimum Role by Operation Category

Operation CategoryMinimum RoleEASBASPAS
Create/modify OrientationA + GA + GAOwner
Create root AuthorityAAAOwner
Delegate AuthorityAAAN/A
Activate CommitmentsAAAOwner
Finalize DecisionsAAAOwner
Complete WorkRRROwner/Collaborator
Advance Activation stageA + GA + GAN/A
Grant Constraint exceptionsA + GA + GAN/A
Resolve InterpretationsAAAN/A
Create SignalsAnyAnyAnyAny
Create IQsAnyAnyAnyAny

Exactly one A per context. No governance context may have zero or multiple Accountable parties. The A role is always human (§22.3).

G role for policy implications. When an operation implicates Constraints or Orientation, the Governor (G) role must participate. G is EAS-only; BAS and PAS approximate through the A role.

V role for evidence requirements. When an operation requires Evidence evaluation, the Verifier (V) role must participate. V is EAS-only.

§27.8.3 Graduation Gates

Some operations require a minimum graduation stage. Graduation is per-scope: AICAR type × Decision type × domain namespace (§8, §A1).

Table 27.8.3: Graduation-Gated Operations

OperationMinimum StageRationale
work_spec.skip_confirmCOnly fully graduated AI may bypass human confirmation
decision.create (AI Zone 1)CAutonomous AI decisions require maximum graduation
capacity.allocate (AI)CAI capacity allocation requires demonstrated reliability
agent.navigate (SM-6)A+Navigation track requires observed pattern reliability
agent.execute (SM-6)CExecution track requires full graduation
interpretation.createBAI interpretation requires Tier 3+ governance capability
activation.advance_stageA+Stage advancement requires demonstrated governance maturity
work.assign (to AI)BAI work assignment requires Tier 3 governance

No stage skipping. Graduation advances one stage at a time: A → A+ → B → C. Auto-regression may drop to the immediate predecessor stage only. Human-initiated regression may drop to any lower stage.

§27.8.4 Profile Tier Availability

Some operations are available only at specific profile tiers. PAS is the most restricted; EAS is the most permissive.

Table 27.8.4: Profile-Restricted Operations

Operation CategoryEASBASPAS
Tier 1–2 primitives (all operations)AvailableAvailableAvailable
Tier 3 primitives (Orientation, Learning, Activation)AvailableAvailable (subset of stages)Not available
Tier 4 primitives (Interpretation, Environment Interface)AvailableAvailableNot available
Tier 5 primitives (Cycle)AvailableAvailableNot available
Work Specifications (SM-4)AvailableAvailableNot available
Delegation lifecycle (SM-5)AvailableAvailableNot available
Constraint exception managementAvailableAvailableNot available
Namespace graduationAvailableAvailableNot available
RACIVG V and G rolesAvailableNot availableNot available

§27.9 Invariant Enforcement Points

Each behavioral invariant (§5) is triggered by a specific set of operations. This section maps invariants to their triggering operations, validation pass, enforcement timing, and override availability.

The trigger tables below list the operations whose primary purpose intersects the invariant — the operations where the invariant check is structurally essential to the operation's postconditions. Any operation that produces a Decision also triggers B4; the B4 table lists the operation categories rather than enumerating every Decision-producing operation individually. The per-primitive operation tables in §27.2–§27.6 are authoritative for the complete invariant profile of each operation.

§27.9.1 B1 — Work Requires Commitment

Every Work instance must link to exactly one Commitment. Prevents shadow work.

Shape: dlps:B1-WorkRequiresCommitment Pass: 1 (Core). Always Blocking. Override: None. No override mechanism exists for B1.

Table 27.9.1: B1 Triggering Operations

OperationTrigger Condition
work.createValidates commitment_id references active Commitment
work.updateValidates commitment_id not removed
work_spec.composeValidates parent Work has valid Commitment link
orchestration.composeValidates composed Work Specification links to Commitment
gov_activation.create_closure_roadmapValidates all generated Work items link to Commitments
iq.resolveValidates any generated Work items link to Commitments

§27.9.2 B2 — Commitment Requires Capacity

Every Commitment must link to at least one Capacity allocation. Prevents impossible promises.

Shape: dlps:B2-CommitmentRequiresCapacity Pass: 1 (Core). Always Blocking. Override: None.

Table 27.9.2: B2 Triggering Operations

OperationTrigger Condition
commitment.createValidates Capacity allocation exists and is sufficient
commitment.updateValidates updated terms do not exceed Capacity
commitment.activateRe-verifies Capacity sufficiency at activation time
capacity.releaseValidates release does not violate dependent Commitments
capacity.exhaustValidates no active Commitments depend on exhausted Capacity
gov_activation.create_closure_roadmapValidates generated Commitments have Capacity

§27.9.3 B3 — Evidence Requires Truth Type

Every Evidence instance must carry exactly one truth type from the four-value vocabulary (Authoritative, Declared, Derived, Opaque — §6.1). All AI-generated content enters as Derived. Opaque is system-assigned at boundary detection. Prevents epistemic corruption.

Shape: dlps:B3-EvidenceRequiresTruthType Pass: 1 (Core). Always Blocking. Override: None.

Table 27.9.3: B3 Triggering Operations

OperationTrigger Condition
evidence.createValidates truth type assigned; AI actor → Derived only
evidence.promoteValidates monotonic promotion (Derived → Declared → Authoritative)
work_spec.completeValidates all outputs carry truth type = Derived
work_spec.promoteValidates promotion follows truth type rules
agent.executeValidates all outputs carry truth type = Derived
interpretation.createValidates truth type = Derived for AI-generated interpretations
vp.materializeValidates truth type = Derived for all materialized projections
signal.createValidates Signal Evidence carries truth type = Authoritative
All AI actor write operationsEnforces outputTruthType: "derived" at protocol level

§27.9.4 B4 — Decision Requires Account

Every Decision must link to exactly one Account. Prevents context-free decisions.

Shape: dlps:B4-DecisionRequiresAccount Pass: 1 (Core). Always Blocking. Override: None.

Table 27.9.4: B4 Triggering Operations

OperationTrigger Condition
decision.createAccount created and linked
decision.reverseReversal Decision creates new Account
work_spec.confirm, work_spec.promote, work_spec.rejectConfirmation/promotion/rejection Decision requires Account
delegation.grant, delegation.modify, delegation.revokeDelegation governance actions create Decisions with Accounts
graduation.advance, graduation.auto_regress, graduation.human_regressGraduation changes create Decisions with Accounts
iq.resolve, iq.abandonResolution and abandonment are Decisions requiring Accounts
signal.resolve, signal.dismissSignal resolution/dismissal creates Decisions with Accounts
constraint.grant_exceptionOverride Decision requires Account
orchestration.classify, orchestration.select, orchestration.confirmClassification, selection, and confirmation create Decisions
kpi.introduce, kpi.retireKPI lifecycle events are Decisions requiring Accounts
actor.activate, actor.suspend, actor.reinstate, actor.deactivate, actor.archiveActor lifecycle transitions create Decisions with Accounts

§27.9.5 B5 — Authority Must Be Traceable

Every Authority must be root or link through a delegation chain terminating at root. Prevents untraceable authority.

Two shapes with different enforcement profiles:

B5-immediate (Pass 1, Always Blocking). Shape dlps:B5-AuthorityTraceable. Checks: isRootAuthority = true OR delegated_from present. O(1).

B5-T (Pass 2, Default Blocking). Shape dlps:B5-AuthorityChainReachesRoot. Checks: transitive closure to root; detects circular delegation. O(chain depth). May be downgraded to Warning during bootstrapping only — not below Logging.

Table 27.9.5: B5 Triggering Operations

OperationTrigger Condition
authority.createValidates root designation or parent chain
authority.delegateValidates scope subset and chain extension
delegation.grant, delegation.modifyValidates delegation does not break chain
authority.revokeValidates cascade does not orphan downstream authorities
decision.createValidates authority_source is traceable
constraint.grant_exceptionValidates overriding authority is traceable
portfolio.spawn_instanceValidates child instance authority derives from parent
portfolio.relax_constraintValidates relaxation authority is traceable
evidence.promoteValidates promoting authority is traceable

Profile variability. EAS and BAS: B5-T is Blocking (Warning during bootstrap only). PAS: B5-T is Warning.

§27.9.6 B6 — Constraint Binds Primitives

Every Constraint must target at least one primitive instance and specify an enforcement mode.

Two shapes:

B6-binding (Pass 1, Always Blocking). Shape dlps:B6-ConstraintBindsPrimitives. Checks: applies_to non-empty AND enforcement_mode in vocabulary. O(1).

B6-U (Pass 2, Default Warning). Shape dlps:B6-ConstraintUniversality. Checks: universal Constraint targets all instances of governed class. O(instances of target class). Override available via §5.5.

Table 27.9.6: B6 Triggering Operations

OperationTrigger Condition
constraint.createValidates binding and enforcement mode
constraint.updateValidates binding not removed
constraint.grant_exceptionValidates exception does not violate universality requirements
gov_activation.activate_sourcesValidates activated sources bind to primitives
gov_activation.propagate_changesValidates propagated changes maintain bindings
portfolio.spawn_instanceValidates child constraints bind to child primitives
portfolio.tighten_constraintValidates tightened constraint maintains binding
orchestration.activate_constraintsValidates activated constraints bind to work context

Profile variability. B6-binding: Blocking at all tiers. B6-U: EAS = Warning (may upgrade for regulatory domains), BAS = Warning, PAS = Logging.

§27.9.7 B7 — All Objects Flaggable

Every primitive class (all nineteen) must have a signal attachment surface. Prevents invisible problems.

Shape: dlps:B7-SignalSurfaceCoverage Pass: Schema Pass (deployment-time). O(19). Override: None. Deployment is halted until corrected.

Table 27.9.7: B7 Triggering Events

EventTrigger Condition
Schema deploymentValidates all nineteen primitive classes have signal attachment surface
Schema modificationRe-validates signal surface coverage after any class or property change
signal.createValidates target primitive class is in the flaggable set
Tier activation (§4.7)Validates newly activated tier classes have signal surfaces
env_interface.receive_signalValidates inbound signal targets a flaggable primitive class
intent.block, work.block, activation.blockValidates blocking event creates signal on the blocked primitive

Profile variability. Blocking at all profiles. EAS: all nineteen classes. BAS: Tiers 1–3 minimum. PAS: Tiers 1–2 minimum.

§27.9.8 B8 — Signals Route to Authority

Every Signal must route to an authority on the governance chain for the flagged object.

Two components:

B8-core (Pass 1, Always Blocking). Shape dlps:B8-SignalsRouteToAuthority (core). Checks: routed_to is not null and resolves to Authority. O(1).

B8-routing (Pass 2, Default Warning). Shape dlps:B8-SignalsRouteToAuthority (SPARQL). Checks: target is on the flagged object's authority chain at any depth. O(chain depth). Routing outside the chain is a violation; routing to a higher level on the correct chain is valid.

Table 27.9.8: B8 Triggering Operations

OperationTrigger Condition
signal.routeValidates target is on governance chain
signal.rerouteValidates new target stays on governance chain
signal.escalateValidates escalation target is next higher on chain
agent.alertValidates alert routes to authority
authority.revoke, authority.suspendValidates active signals are re-routed
delegation.revoke, delegation.expireValidates signals routed through affected delegation are re-routed

Profile variability. B8-core: Blocking at all tiers. B8-routing: EAS = Warning (may upgrade for regulated domains), BAS = Warning, PAS = Logging.

§27.9.9 B9 — IQ Resolution Creates Decision

Every resolved IQ must produce a Decision or Work item. Prevents inert emergence.

Shape: dlps:B9-IQResolutionCreatesDecision Pass: 2 (SPARQL). Default Warning. O(1) per resolved IQ. Override: Available via §5.5, rarely needed since abandonment (a Decision to not act) satisfies B9.

Table 27.9.9: B9 Triggering Operations

OperationTrigger Condition
iq.resolveValidates resolution produces Decision or Work
iq.abandonValidates abandonment Decision exists (abandonment is a Decision)
iq.promoteValidates promoted findings link to governance action
iq.revertReverted IQ must eventually re-resolve

Profile variability. EAS = Warning (may upgrade to Blocking for high-governance domains), BAS = Warning, PAS = Logging.

§27.9.10 Enforcement Summary

Table 27.9.10: Invariant Enforcement Profile Matrix

InvariantPassEASBASPASOverride
B11BlockingBlockingBlockingNone
B21BlockingBlockingBlockingNone
B31BlockingBlockingBlockingNone
B41BlockingBlockingBlockingNone
B5-immediate1BlockingBlockingBlockingNone
B5-T2BlockingBlockingWarning§5.5 (bootstrap only)
B6-binding1BlockingBlockingBlockingNone
B6-U2WarningWarningLogging§5.5
B7SchemaBlockingBlockingBlockingNone
B8-core1BlockingBlockingBlockingNone
B8-routing2WarningWarningLogging§5.5
B92WarningWarningLogging§5.5

Reading this table. Pass 1 shapes are universally Blocking — no profile may downgrade them. Pass 2 shapes decrease in strictness from EAS through PAS. No invariant may be downgraded below Logging for any profile. The Schema Pass (B7) runs at deployment time and is universally Blocking.


§27.10 SDK Constraints

Table 27.10.1: §27 SDK Constraints

IDConstraintTypeRationale
§27-C1Every write operation MUST check all five authorization gates (§27.1.2) before executionMUSTAuthorization is the foundation of governance integrity; partial checks create exploitable gaps
§27-C2Every operation that produces a record MUST assign truth type at creation timeMUSTB3 requires truth type on all Evidence; truth type tracking on other primitives enables provenance auditing
§27-C3All AI actor write operations MUST produce records with truth type = DerivedMUSTAI-Cannot-Be-Principal; truth type boundary prevents epistemic corruption (§6)
§27-C4Truth type promotion MUST follow monotonic order: Derived → Declared → AuthoritativeMUSTDemotion would invalidate governance decisions made on the basis of the higher truth type
§27-C5Composite operations MUST be atomic — complete entirely or fail entirelyMUSTPartial completion creates structurally invalid governance state
§27-C6Authoritative records MUST NOT be modified after creationMUST NOTAppend-only invariant preserves audit trail integrity
§27-C7Derived records MUST NOT be manually editedMUST NOTDerived content must be recomputable from sources; manual edits break reproducibility
§27-C8Operations MUST NOT hard-delete records from the governance graphMUST NOTSoft-delete (status → archived) preserves governance lineage; hard deletes destroy audit trail
§27-C9AI actors MUST NOT hold the Accountable (A) role, create root Authority, delegate Authority, promote truth types, or resolve InterpretationsMUST NOTAI-Cannot-Be-Principal (§22.3); these operations require human judgment and accountability
§27-C10AI actors MUST NOT modify their own delegation scopeMUST NOTSelf-modification of delegation boundaries violates principal-agent separation (SM-5)
§27-C11Pass 1 invariant shapes MUST NOT be downgraded below Blocking for any profileMUST NOTPass 1 shapes enforce structural properties that cannot be deferred without governance failure
§27-C12Invariant enforcement MUST NOT be downgraded below Logging for any profileMUST NOTEven the lightest enforcement mode must preserve audit visibility
§27-C13B7 schema validation MUST pass before any instance data is writtenMUSTSchema shapes validate the signal surface; instance data without signal surfaces violates B7
§27-C14The two-pass validation strategy MUST execute after write but before transaction commitMUSTPre-commit validation prevents structurally invalid state from becoming persistent
§27-C15Specific transport protocol (REST, gRPC, GraphQL, message queue) for operation invocationDESIGN SPACEProtocol choice depends on deployment topology, latency requirements, and client ecosystem
§27-C16Client SDK language bindings and ergonomic API surfaceDESIGN SPACELanguage choice and API style depend on implementation team and deployment environment
§27-C17Atomicity mechanism for composite operations (database transactions, saga pattern, event sourcing)DESIGN SPACEMultiple valid approaches exist; choice depends on storage layer and consistency requirements
§27-C18Concurrency control strategy for conflicting write operationsDESIGN SPACEOptimistic vs. pessimistic locking, conflict resolution policy, and retry strategy are implementation decisions
§27-C19Authorization cache strategy for the five-gate checkDESIGN SPACECaching delegation chains and role mappings improves performance but introduces staleness risk
§27-C20Pass 2 invariant evaluation scheduling (synchronous, asynchronous, batch)DESIGN SPACESchedule depends on profile tier, operation volume, and acceptable latency for governance feedback
§27-C21Event notification mechanism for operation completion and invariant violationsDESIGN SPACEWebhooks, message queues, server-sent events, and polling are all valid approaches
§27-C22Audit log storage format and retention policyDESIGN SPACEStorage format must preserve all fields from Account and invariant violation records; retention policy is deployment-specific

§27.11 Terminology

TermDefinition
MVR (Minimum Viable Record)The minimal set of fields required to create a valid primitive instance; defined per-primitive in §26
Operation TypeOne of six structural categories — Create, Read, Update, Delete, Transition, Composite — that classifies every operation in the catalog (§27.1.1)
Truth TypeClassification of epistemic status — Authoritative, Declared, Derived, or Opaque — assigned at record creation (or system-assigned for Opaque) and governed by monotonic promotion rules (§6). Opaque is boundary-enforced and does not participate in the standard promotion path.
Authorization GateOne of five sequential checks — Authority scope, Authority traceability, RACIVG role, Graduation gate, Profile gate — that every write operation must pass (§27.1.2)
Composite OperationA multi-primitive operation that executes atomically within a single logical transaction; partial completion is not a valid state
Pass 1 / Pass 2The two-pass SHACL validation strategy: Pass 1 (Core, single-node, O(1), always Blocking) and Pass 2 (SPARQL, graph traversal, profile-configurable)
Schema PassShapes-on-shapes validation that runs at deployment time; enforces B7 signal surface coverage across all nineteen primitive classes
Graduation StagePer-scope AI maturity level — A, A+, B, C — that gates operation availability; advances monotonically one stage at a time (§8, §A1)
Tighten-Only CascadeThe constraint monotonicity rule: child instances may only tighten inherited parent constraints, never relax them
AI-Cannot-Be-PrincipalThe governance boundary (§22.3) prohibiting AI actors from holding Accountable (A) role, creating root Authority, delegating Authority, promoting truth types, or resolving Interpretations
Soft-DeleteTransition of record_lifecycle_state to archived; no records are hard-deleted from the governance graph
SupersessionThe mechanism for correcting immutable records: a new record with supersedes_ref pointing to the original; the original is preserved unchanged

§27.12 Cross-References

SectionRelationship
§4 Primitive TaxonomyDefines the nineteen primitives and five-tier architecture that §27 operations act upon
§5 Behavioral InvariantsDefines B1–B10 invariants enforced by §27.9
§6 Truth Type ModelGoverns truth type assignment, immutability, and promotion rules applied across all operations
§8 Graduation ModelDefines per-scope graduation stages that gate operation availability (§27.8.3)
§12 Governance Activation PipelineDefines the six-stage pipeline operationalized by §27.7.4
§14–§15 Signal ArchitectureDefines signal semantics operationalized by §27.7.1
§16 IQ ArchitectureDefines investigative query semantics operationalized by §27.7.2
§17 VP ArchitectureDefines validated projection semantics operationalized by §27.7.3
§18–§19 Knowledge IntegrationDefines ontology alignment pipeline operationalized by §27.7.10
§21 Substrate ProfilesDefines EAS/BAS/PAS profile tiers referenced in §27.8.4
§22 Actor ModelDefines actor types, RACIVG roles, and AI-Cannot-Be-Principal boundary
§23 Portfolio PatternsDefines depth-gated visibility and tighten-only constraint cascade
§24 Entity LicensingDefines license structure governing SDK implementation
§25 License TermsDefines license-as-constraint architecture and compliance verification
§26 Implementation SchemaDefines the logical data model that §27 operations act upon
§A1–§A2 Orchestration AppendixDefines the seven-step orchestration engine operationalized by §27.7.5

Scope

Scope limited to logical operation contracts. Transport protocol, serialization format, and implementation languages are DESIGN SPACE.

Implementation Requirements

SDK implementations MUST respect all invariant shapes, enforce truth type boundaries, and validate all constraints at specified passes. Specification is immutable during lock period. SHACL two-pass validation is mandatory. Truth type boundaries are preserved through all operations. Session isolation is required for all writes.

On this page

§27 Operation Catalog§27.1 Operation Architecture Overview§27.1.1 Six Operation TypesTable 27.1.1: Operation Type Taxonomy§27.1.2 Authorization Model§27.1.3 Truth Type Rules§27.1.4 Invariant Enforcement§27.2 Tier 1 Primitive Operations§27.2.1 IntentTable 27.2.1: Intent Operations§27.2.2 CommitmentTable 27.2.2: Commitment Operations§27.2.3 CapacityTable 27.2.3: Capacity Operations§27.2.4 WorkTable 27.2.4: Work OperationsTable 27.2.5: Work Specification Operations (SM-4)§27.2.5 EvidenceTable 27.2.6: Evidence Operations§27.2.6 DecisionTable 27.2.7: Decision Operations§27.2.7 AuthorityTable 27.2.8: Authority OperationsTable 27.2.9: Delegation Operations (SM-5)§27.2.8 AccountTable 27.2.10: Account Operations§27.2.9 ConstraintTable 27.2.11: Constraint Operations§27.3 Tier 2 Infrastructure Operations§27.3.1 IdentifierTable 27.3.1: Identifier Operations§27.3.2 Entity and ActorTable 27.3.2: Entity OperationsTable 27.3.3: Actor Operations (SM-1)§27.3.3 ContextTable 27.3.4: Context Operations§27.3.4 NamespaceTable 27.3.5: Namespace Operations§27.4 Tier 3 Governed Operations§27.4.1 OrientationTable 27.4.1: Orientation Operations§27.4.2 LearningTable 27.4.2: Learning Operations§27.4.3 ActivationTable 27.4.3: Activation Operations§27.5 Tier 4 AI-Native Operations§27.5.1 InterpretationTable 27.5.1: Interpretation Operations§27.5.2 Environment InterfaceTable 27.5.2: Environment Interface Operations§27.6 Tier 5 Configuration Operations§27.6.1 CycleTable 27.6.1: Cycle Operations§27.7 Cross-Cutting Operations§27.7.1 Signal OperationsTable 27.7.1: Signal Operations§27.7.2 IQ OperationsTable 27.7.2: IQ Operations§27.7.3 VP OperationsTable 27.7.3: VP Operations§27.7.4 Governance Activation OperationsTable 27.7.4: Governance Activation Operations§27.7.5 Orchestration OperationsTable 27.7.5: Orchestration Operations§27.7.6 KPI OperationsTable 27.7.6: KPI Operations§27.7.7 Profile and Portfolio OperationsTable 27.7.7: Profile and Portfolio Operations§27.7.8 Graduation OperationsTable 27.7.8: Graduation Operations§27.7.9 Agent Behavioral Cycle OperationsTable 27.7.9: Agent Behavioral Cycle Operations§27.7.10 Ontology Alignment OperationsTable 27.7.10: Ontology Alignment Operations§27.8 Authorization Matrix§27.8.1 Actor Type PermissionsTable 27.8.1: Actor Type Permission Summary§27.8.2 RACIVG Role RequirementsTable 27.8.2: RACIVG Minimum Role by Operation Category§27.8.3 Graduation GatesTable 27.8.3: Graduation-Gated Operations§27.8.4 Profile Tier AvailabilityTable 27.8.4: Profile-Restricted Operations§27.9 Invariant Enforcement Points§27.9.1 B1 — Work Requires CommitmentTable 27.9.1: B1 Triggering Operations§27.9.2 B2 — Commitment Requires CapacityTable 27.9.2: B2 Triggering Operations§27.9.3 B3 — Evidence Requires Truth TypeTable 27.9.3: B3 Triggering Operations§27.9.4 B4 — Decision Requires AccountTable 27.9.4: B4 Triggering Operations§27.9.5 B5 — Authority Must Be TraceableTable 27.9.5: B5 Triggering Operations§27.9.6 B6 — Constraint Binds PrimitivesTable 27.9.6: B6 Triggering Operations§27.9.7 B7 — All Objects FlaggableTable 27.9.7: B7 Triggering Events§27.9.8 B8 — Signals Route to AuthorityTable 27.9.8: B8 Triggering Operations§27.9.9 B9 — IQ Resolution Creates DecisionTable 27.9.9: B9 Triggering Operations§27.9.10 Enforcement SummaryTable 27.9.10: Invariant Enforcement Profile Matrix§27.10 SDK ConstraintsTable 27.10.1: §27 SDK Constraints§27.11 Terminology§27.12 Cross-ReferencesScopeImplementation Requirements