Use when an IFC file is schema-valid but you must judge whether it is actually useful, audit an IFC model for information completeness, score model quality before handover, or answer why a valid IFC file is empty in facility management, quantity take-off, or analysis software. Prevents confusing schema validity with information completeness, declaring a model "good" because it passed validation, auditing project-specific demands that belong to IDS, and shipping a model whose elements carry no properties, materials, classifications, quantities, or georeferencing. Covers a cross-skill model-quality audit across seven completeness dimensions (spatial containment, property sets, geometry representation, materials, classifications, quantities, georeferencing), a deterministic weighted scoring rubric, the validity-versus-completeness-versus-IDS three-tier model, and routing each gap to the owning skill, across IFC2x3, IFC4 and IFC4.3. Keywords: IFC quality check, IFC model audit, information completeness, model qua
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Une commande directe contourne le prompt de vérification. Examinez la source avant de l'exécuter.
Use when an IFC file is schema-valid but you must judge whether it is actually useful, audit an IFC model for information completeness, score model quality before handover, or answer why a valid IFC file is empty in facility management, quantity take-off, or analysis software. Prevents confusing schema validity with information completeness, declaring a model "good" because it passed validation, auditing project-specific demands that belong to IDS, and shipping a model whose elements carry no properties, materials, classifications, quantities, or georeferencing. Covers a cross-skill model-quality audit across seven completeness dimensions (spatial containment, property sets, geometry representation, materials, classifications, quantities, georeferencing), a deterministic weighted scoring rubric, the validity-versus-completeness-versus-IDS three-tier model, and routing each gap to the owning skill, across IFC2x3, IFC4 and IFC4.3. Keywords: IFC quality check, IFC model audit, information completeness, model quality score, IFC quality report, valid but empty IFC, missing property sets, missing materials, missing classification, missing quantities, missing georeferencing, orphan elements, completeness coverage, pre-handover audit, "my IFC file is valid but useless", "nothing shows in facility management", "quantity take-off is empty", "how good is this IFC model", is my BIM model complete, what is wrong with this IFC model.
license
MIT
compatibility
Designed for Claude Code. Requires IFC IFC2x3, IFC4, IFC4.3.
metadata
{"author":"OpenAEC-Foundation","version":"1.0"}
IFC Quality Checker (Orchestration)
Drive a complete, scored quality audit over an IFC file. This is an ORCHESTRATION
skill : it audits seven information-completeness dimensions, scores each with a
deterministic rubric, produces one quality report, and routes every gap to the
skill that holds the detailed fix.
This skill answers one question : the file is valid, but is it any USE. A model
can pass every buildingSMART validation stage and still be near-worthless
downstream : walls with no properties, no materials, no classification, no
quantities, sitting at the origin with no georeferencing. Validity is the floor,
not the goal.
This skill does NOT itself teach the fixes. It is the auditor ; the syntax, impl,
and errors skills are the remediation. It does NOT run the schema-conformance
pipeline either : that is ifc-agents-file-validator.
Quick Reference
The three-tier model : validity, completeness, requirements
ALWAYS place every check in exactly one tier:
Tier
Question
Owner
1. Schema validity
Does the file obey the IFC standard
ifc-agents-file-validator
2. Information completeness
Does the file carry the generic data any consumer needs
THIS skill
3. Project requirements
Does the file carry what a specific project demanded
IDS (out of package scope)
This skill owns tier 2 ONLY. NEVER audit tier-1 schema rules here : route them to
ifc-agents-file-validator. NEVER audit tier-3 project-specific demands here :
those are expressed in IDS, which this package does not cover. Tier 2 is the
GENERIC completeness every downstream use (facility management, quantity
take-off, energy and structural analysis, cost) needs regardless of project.
Validity gate : audit quality only on a valid file
ALWAYS confirm tier 1 first. Run ifc-agents-file-validator and require zero
stage 1, 2, or 3 errors before this audit. A quality score on an invalid file is
meaningless : a malformed file cannot be measured for coverage. If validation
reports errors, STOP and route to ifc-agents-file-validator.
Weights sum to 100. They are FIXED : ALWAYS use these exact weights so two audits
of the same file produce the same score.
Scoring rubric (deterministic)
For each dimension, compute coverage = elements satisfying the check / elements in scope, as a percentage. Then:
coverage = 100% : status COMPLETE.
1% <= coverage <= 99% : status PARTIAL.
coverage = 0% : status ABSENT.
dimension score = weight * coverage. overall score = sum of dimension scores,
0 to 100. Grade bands, applied ALWAYS:
Overall score
Grade
90 to 100
A
75 to 89
B
50 to 74
C
25 to 49
D
0 to 24
F
A dimension that is genuinely out of scope (for example precise georeferencing on
a coordination-only model) is marked not applicable : drop it and renormalise
the remaining weights to sum to 100. NEVER score a not-applicable dimension as 0.
Decision Trees
Is this a quality gap, a validity error, or a project requirement
The finding under review ...
├─ breaks an EXPRESS rule, a normative Gherkin rule, or the STEP syntax
│ -> tier 1, schema validity. Route to ifc-agents-file-validator.
├─ is generic data any consumer needs, missing or thin (no Pset, no material,
│ no class, no quantity, no georeferencing, an orphan element)
│ -> tier 2. This skill. Score it under the matching dimension.
└─ is a project-, client-, or nation-specific demand (a named Pset required by
this contract, a specific LOD)
-> tier 3. IDS territory. Out of this package's scope. Do NOT score it.
Which dimension does a symptom map to
Downstream symptom ...
├─ "element missing from the model tree / facility-management export"
│ -> Dimension 1, spatial containment.
├─ "schedules and parameter views are empty"
│ -> Dimension 2, property sets.
├─ "element has no shape / nothing draws in the viewer"
│ -> Dimension 3, geometry representation.
├─ "material take-off / material schedule is empty"
│ -> Dimension 4, materials.
├─ "cannot filter or cost by classification code"
│ -> Dimension 5, classifications.
├─ "quantity take-off returns nothing"
│ -> Dimension 6, quantities.
└─ "model sits at the origin, will not align with survey or GIS"
-> Dimension 7, georeferencing.
Georeferencing depth by version
What schema is the file ?
├─ IFC2x3
│ -> precise CRS georeferencing entities do not exist. Check IfcSite
│ RefLatitude, RefLongitude, RefElevation only. Present = COMPLETE for
│ this version ; absent = ABSENT.
└─ IFC4 or IFC4.3
-> check IfcMapConversion plus IfcProjectedCRS (precise). Present = COMPLETE.
Only IfcSite RefLatitude/RefLongitude present = PARTIAL (coarse only).
Neither = ABSENT.
Patterns
Pattern 1 : run the quality audit in order
ALWAYS run the audit as an ordered pass:
1. Validity gate : run ifc-agents-file-validator. Require zero tier-1 errors.
If errors exist, STOP and route there. Do not score.
2. Determine the file schema (IFC2x3, IFC4, IFC4.3) and the declared MVD.
Schema decides which entities the audit checks (Pattern 5, version notes).
3. Build the element scope : the set of physical element occurrences. In
IFC2x3 and IFC4 these are IfcBuildingElement subtypes ; in IFC4.3 they are
IfcBuiltElement subtypes (the entity was renamed and made non-abstract).
4. Audit dimensions 1 to 7 (Patterns 2 to 5), each producing a coverage %.
5. Score and grade (Pattern 6). Emit one report.
For every physical element occurrence, check it is contained in exactly one
spatial level through IfcRelContainedInSpatialStructure, and that the spatial
tree is complete from IfcProject down.
coverage = contained elements / total physical elements.
An element with zero containment links is an orphan : it does not count as
covered.
ALSO confirm the tree itself : an IfcProject, at least one IfcSite, and an
unbroken IfcRelAggregates chain. A broken tree caps this dimension at PARTIAL
regardless of element coverage.
Route gaps to ifc-errors-spatial-structure and ifc-impl-spatial-decomposition.
Pattern 3 : audit property sets and quantities (dimensions 2 and 6)
Properties and quantities share the same attachment machinery : both are
IfcPropertySetDefinition subtypes attached through IfcRelDefinesByProperties
(on occurrences) or the type object's HasPropertySets (on types). Audit them as
two dimensions.
Dimension 2, property sets:
For each element, check it carries the standardised common property set for its
class : an IfcPropertySet named Pset_<Class>Common (for example
Pset_WallCommon on IfcWall), resolving type-level and occurrence-level
property sets and letting the occurrence win on a name collision.
coverage = elements with their Pset_*Common / total elements.
A custom property set whose Name uses the reserved Pset_ prefix is a
defect : flag it and route to ifc-errors-property-set-mistakes. It does NOT
count as coverage.
Dimension 6, quantities:
For each element, check it carries an IfcElementQuantity with the relevant
measured quantities. IFC4 and IFC4.3 standardise the Qto_ prefix (for example
Qto_WallBaseQuantities) ; IFC2x3 used a BaseQuantities-style convention, so
for IFC2x3 check for the presence of an IfcElementQuantity, NEVER for a
Qto_ name.
coverage = elements with measured quantities / total elements.
Route property gaps to ifc-errors-property-set-mistakes and
ifc-impl-data-enrichment ; quantity gaps to ifc-syntax-quantities and
ifc-impl-property-extraction.
Pattern 4 : audit materials and classifications (dimensions 4 and 5)
Dimension 4, materials:
For each element, check a material association exists : an
IfcRelAssociatesMaterial whose RelatedObjects includes the element (or its
type) and whose RelatingMaterial is an IfcMaterial, IfcMaterialLayerSet,
IfcMaterialProfileSet, IfcMaterialConstituentSet, or a usage of one.
coverage = elements with a material association / total elements.
Dimension 5, classifications:
For each element, check an IfcRelAssociatesClassification exists whose
RelatingClassification is an IfcClassificationReference (a code such as a
Uniclass or OmniClass notation) or an IfcClassification.
coverage = classified elements / total elements.
Route material gaps to ifc-syntax-materials and ifc-impl-data-enrichment ;
classification gaps to ifc-syntax-classifications and ifc-impl-data-enrichment.
Pattern 5 : audit geometry and georeferencing (dimensions 3 and 7)
Dimension 3, geometry representation:
For each element, check it has an IfcProductDefinitionShape with at least one
usable IfcShapeRepresentation (typically a Body representation) bound to a
valid representation context.
coverage = elements with a usable representation / total elements.
A degenerate or context-less representation does NOT count : route it to
ifc-errors-geometry-issues.
Dimension 7, georeferencing : audit by version per the decision tree.
IFC4, IFC4.3 : COMPLETE when IfcMapConversion plus IfcProjectedCRS are
present. PARTIAL when only IfcSiteRefLatitude and RefLongitude are set.
ABSENT when neither.
IFC2x3 : precise CRS entities do not exist. COMPLETE when IfcSiteRefLatitude, RefLongitude, RefElevation are set ; ABSENT otherwise.
Georeferencing is a single model-level dimension, not a per-element coverage :
score it directly as COMPLETE (100%), PARTIAL (50%), or ABSENT (0%). Route gaps
to ifc-syntax-georeferencing.
Pattern 6 : produce the scored quality report
Emit one report with a fixed structure. ALWAYS show every dimension, its weight,
coverage, status, and weighted score, then the overall score and grade.
ALWAYS state the grade with the score. ALWAYS list, per gap, the owning skill so
the reader can act. NEVER present a grade without the per-dimension breakdown :
an opaque score hides which data is missing.
Reference Links
references/methods.md : the seven dimensions in full, the check entities per
dimension, the scoring formula, version notes, the anti-pattern-to-dimension map.
references/examples.md : worked audits and complete sample reports.
references/anti-patterns.md : auditing mistakes and why each one is wrong.
Related skills
ifc-agents-file-validator : the tier-1 schema-conformance pipeline. ALWAYS run
first.
ifc-core-validation : the conceptual validation model.
ifc-impl-data-enrichment : how to ADD the missing properties, materials, and
classifications this audit reports.
ifc-errors-spatial-structure, ifc-errors-property-set-mistakes,
ifc-errors-geometry-issues : the detailed fixes for each gap class.
ifc-syntax-property-sets, ifc-syntax-quantities, ifc-syntax-materials,
ifc-syntax-classifications, ifc-syntax-georeferencing : the authoring
syntax for each dimension.
Verified sources
The check entities (IfcRelContainedInSpatialStructure, IfcPropertySet,
IfcRelDefinesByProperties, IfcElementQuantity, IfcRelAssociatesMaterial,
IfcRelAssociatesClassification, IfcClassificationReference,
IfcMapConversion, IfcProjectedCRS, IfcSite), the Pset_ and Qto_
conventions, and the version differences are verified 2026-05-20 against the
buildingSMART IFC4.3, IFC4 ADD2 TC1, and IFC2x3 specifications recorded in
SOURCES.md. The validity-pipeline boundary is sourced from the buildingSMART
validation service documentation.
