Ontology Schema — {DOMAIN_DESCRIPTION}

Node Types

These are the typed entities in your domain reasoning. Keep the generic types and add domain-specific types below the table.

Node Type	Description	Key Fields
Task	A unit of work (issue, ticket, research question)	id, title, state, assigned_to
Claim	A technical or analytical assertion	statement, mode, validation_status, provenance
Evidence	Metrics, results, observations, API responses	source, timestamp, authority_score
Constraint	Domain limits, policies, invariants	id, description, durability, source
Assumption	Beliefs that reasoning depends on	id, description, sensitivity, testable, test_method
Method	An approach, tool, or procedure	name, prerequisites, assumptions_required, risks
Validation	Gate passage result	gate_name, status (6-state), evidence_refs
Risk	Potential negative outcome	description, likelihood, impact, mitigation
Resource	External knowledge source with quality metadata	url, authority_score, last_verified

Add node types specific to your domain:

Node Type	Description	Key Fields
{DOMAIN_NODE_1}	{DESCRIPTION}	{FIELDS}
{DOMAIN_NODE_2}	{DESCRIPTION}	{FIELDS}

Edge	From → To	Semantics
`supports`	Evidence → Claim	Evidence increases confidence in claim
`refutes`	Evidence → Claim	Evidence contradicts claim
`requires`	Claim → Evidence	Claim depends on this evidence
`derives_from`	Claim → Constraint	Follows from domain invariant or policy
`gates`	Claim → Validation	Must pass gate before actionable
`assumes`	Claim → Assumption	Depends on this assumption holding
`has_risk`	Method → Risk	Approach carries this risk
`mitigates`	Method → Risk	Approach reduces this risk
`blocks`	Task → Task	This task must complete first
`implements`	Task → Claim	Work item implements this recommendation

Edge	From → To	Semantics
`{DOMAIN_EDGE_1}`	{FROM} → {TO}	{SEMANTICS}

Every Claim has exactly one provenance tag: llm_proposed | user_directed | convention | stakeholder_directed
Every Claim in Implementation mode has at least one Assumption edge
Every Assumption with sensitivity: critical must be tested before dependent Claims are actionable
Constraints with durability: established are never overridden without user approval
Every Resource has a deterministic authority_score (see rules/authority-scoring.md)
{DOMAIN_INVARIANT_1}
{DOMAIN_INVARIANT_2}

Status	Meaning	Actionability
`unvalidated`	No check attempted	informational only
`pending`	Check in progress	informational only
`partially_validated`	Some gates passed	provisional (dev/test OK)
`validated`	All gates passed	actionable
`contextually_validated`	Passed with documented conditions	actionable with caveats
`failed`	Gate failed	blocked — investigate

Tag	Meaning
`llm_proposed`	Claude suggested this based on analysis
`user_directed`	User explicitly requested this approach
`convention`	Follows from CLAUDE.md or rules/ convention
`stakeholder_directed`	Stakeholder/PI directed this decision