autocad-tianzheng-hvac
Pass
Audited by Gen Agent Trust Hub on Jul 26, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill indexes text and block metadata from CAD drawings, creating a surface where untrusted data enters the agent context.
- Ingestion points: The
scan_indexaction reads entities, blocks, and text from local.dwgfiles into SQLite/JSON/XLSX formats (SKILL.md). - Boundary markers: Not present; the instructions do not specify delimiters or warnings for the agent when processing indexed drawing data.
- Capability inventory: The skill can execute CAD commands via
run_commandand load binary modules viaload_arx(SKILL.md). - Sanitization: No sanitization or validation logic is specified for the text content extracted from the CAD drawings.
- [DYNAMIC_EXECUTION]: The skill uses the
load_arxaction to load binary ARX modules (DLLs) into the AutoCAD process from paths derived from theTIANZHENG_ROOTenvironment variable. This is standard operational procedure for Tianzheng HVAC integration but involves the runtime execution of binary files from the local filesystem.
Audit Metadata