Validation Pack: AOR-5, flat concept conditions

Unit: applied-ontology-readiness Plan section: Phase B Decisions: ADR-A89 items 1, 2 and 4 (SPARQL), ADR-A87, both Proposed

Invariant

A condition matching by ExactMatch or SetMembership compiles through the shared IR, and its SPARQL artefact decides every question as the ADR-A87 table says. It never denies for absent evidence (ADR-A24).

Test cases

ID	Given / When / Then	Level	+/-
AOR5-01	a set condition with three required concepts, one also excluded / compiled / plan holds both sets, sorted	L1	+
AOR5-02	the condition stripped of concepts / compiled / refused	L1	-
AOR5-03	a non-default wildcard policy / compiled / refused	L1	-
AOR5-04	an interval strategy / compiled as a concept plan / refused	L1	-
AOR5-05	candidate in the required set / SPARQL executed / `Permitted`	L1	+
AOR5-06	candidate outside every required concept / executed / `Denied`	L1	-
AOR5-07	candidate both required and excluded / executed / `Denied` (L10)	L1	-
AOR5-08	question with no candidate / executed / `Undetermined`	L1	-
AOR5-09	question with two candidates / executed / `Undetermined`	L1	-
AOR5-10	artefact / inspected / plan typed `exe:ConceptMatchPlan`, `exe:derivedFromVocabularyNode` names every concept	L1	+
AOR5-11	one plan compiled twice / compared / isomorphic	L2	+
AOR5-12	`condition-taxonomy.ttl`’s exact condition plus two questions / executed / `Permitted` and `Denied`	L1	+

Every decision case runs the generated query with rdflib’s SPARQL engine.

One command

mise run check:mork-compilers

Pass: 27 passed (the 15 existing compiler tests and the 12 above). Also run mise run check:ontology-catalog, which includes the Eligibility example tests.

Artefacts to inspect

elg:candidateConcept in ontology/eligibility/README.md §5 and the spec, and the README §4 sentence and table row on candidates. A question with more than one candidate is Undetermined. That reading is new and flagged in the status record.
elg:UndeterminedWhenNoCandidateInput in shapes/rules.ttl now recognises elg:candidateConcept. Without this, the migrated hierarchical-match.ttl received a second decision value and broke L6. AOR2-01 caught it.
exe:ConceptMatchPlan and exe:derivedFromVocabularyNode in ontology/mork/spec/Executable.ttl.
Versions: Eligibility is MINOR, so the Eligibility cascade moves to 0.4.0 against the committed 0.3.0, superseding AOR-2’s interim 0.3.1. Executable moves 0.2.0 → 0.3.0. Nothing imports Executable.

Adversarial probe (run by the agent)

Dropping the exclusion clause failed AOR5-07. Letting two candidates through (< 1 for != 1) failed AOR5-09.

Deviations from the plan

exe:ConceptMatchPlan lands here, not in AOR-7, because the SPARQL artefact’s provenance needs it. exe:derivedFromVocabularyNode is new, since concepts are neither Eligibility nor Quantification nodes.
Exclusion-only flat conditions (L12) need the bound scheme’s members, so they are refused here and handled with scheme resolution in AOR-6.
Conformance-corpus cases compare profile-level decisions. They land with profile aggregation in AOR-8.

Deliberate non-coverage

Hierarchical match and scheme resolution (AOR-6).
SHACL and SWRL for concept plans (AOR-7).
Diagnostic codes for Undetermined (AOR-7).