skills/datadrivenconstruction/ddc_skills_for_ai_agents_in_construction/batch-cad-converter/Gen Agent Trust Hub
batch-cad-converter
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses subprocess.run() to invoke external CAD converters like RvtExporter.exe and DwgExporter.exe. It correctly utilizes list-based arguments to prevent shell injection, though executing local binaries remains a significant capability that depends on the security of those external tools.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the filesystem, such as filenames and directory structures, creating a potential vector for indirect prompt injection if the agent attempts to process this metadata as instructions.
- Ingestion points: Filenames and paths are read from the input_folder in SKILL.md.
- Boundary markers: None identified to separate file metadata from processing logic.
- Capability inventory: The skill can execute commands via subprocess.run and write files via the generate_report function.
- Sanitization: Standard pathlib functions handle paths, but no specific filtering for prompt injection patterns in file metadata is implemented.
Audit Metadata