skills/datadrivenconstruction/ddc_skills_for_ai_agents_in_construction/ifc-data-extraction/Gen Agent Trust Hub
ifc-data-extraction
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external IFC (Industry Foundation Classes) files, which represents an attack surface for indirect prompt injection if the files contain malicious text in property sets or element descriptions.
- Ingestion points: User-provided IFC files are loaded in
SKILL.mdandinstructions.mdusingifcopenshell.open(ifc_path). - Boundary markers: The skill does not explicitly use delimiters or instruction-guarding prompts when presenting extracted metadata to the agent.
- Capability inventory: The skill has filesystem read/write access and the ability to export data to SQL databases via SQLAlchemy.
- Sanitization: No sanitization is performed on extracted element properties or strings before they are processed or exported.
- [SAFE]: The functionality for exporting to SQL databases (
sqlalchemy.create_engine) and local files is standard for data extraction tools and follows the vendor's documented methodology for BIM analysis. No obfuscation, unauthorized remote code execution, or credential exfiltration patterns were found.
Audit Metadata