skills/datadrivenconstruction/ddc_skills_for_ai_agents_in_construction/image-to-data/Gen Agent Trust Hub
image-to-data
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted construction images (site photos, scanned drawings, and documents) which could contain adversarial content or embedded instructions designed to influence the agent's logic.
- Ingestion points: Image files in various formats (.jpg, .png, .tiff, .bmp) provided to the agent as specified in
instructions.md. - Boundary markers: The instructions do not define clear boundaries or specific "ignore" rules for data extracted via OCR or vision analysis.
- Capability inventory: The skill requires
filesystempermissions to read local files andnetworkpermissions to communicate with external AI Vision APIs (Claude/OpenAI). - Sanitization: No explicit sanitization, validation, or filtering of the content extracted from images is described before it is integrated into the agent's context.
Audit Metadata