skills/openaec-foundation/computational-design-day-delft-march-2026/ifcos-impl-profiles/Gen Agent Trust Hub
ifcos-impl-profiles
Pass
Audited by Gen Agent Trust Hub on Mar 24, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: Indirect prompt injection surface. The skill provides implementation patterns that ingest untrusted data from external IFC files, which could potentially be manipulated to influence agent logic.
- Ingestion points: Usage of
ifcopenshell.open("existing_model.ifc")inreferences/examples.mdto load external BIM models. - Boundary markers: Absent from the provided code templates.
- Capability inventory: Examples demonstrate the ability to perform file system writes via
model.write()and modify model entities. - Sanitization: No sanitization or validation of model metadata is implemented in the examples.
- [COMMAND_EXECUTION]: The skill uses
ifcopenshell.api.run(), which is the standard programmatic dispatcher for the IfcOpenShell library. This pattern is internal to the library's architecture and does not represent arbitrary shell command execution. - [EXTERNAL_DOWNLOADS]: The skill documentation identifies dependencies on the
ifcopenshellandnumpyPython packages. These are industry-standard, well-known libraries for architectural engineering and numerical data processing.
Audit Metadata