text-to-cad-harness
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The installation process involves cloning a repository from an untrusted source (https://github.com/earthtojake/text-to-cad.git).
- [DYNAMIC_EXECUTION]: The skill's primary function is to have an AI agent generate Python code from user prompts and execute that code locally to produce CAD geometry (e.g., executing models/bracket/bracket.py).
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted natural language user prompts and interpolates them into a workflow that generates and executes code without explicit boundary markers or sanitization, creating an attack surface for potential prompt injection.
- [COMMAND_EXECUTION]: The workflow involves multiple shell commands for environment setup and execution, including environment preparation (python -m venv), package installation (pip install, npm install), and script execution.
Audit Metadata