patent-disclosure

Warn

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: MEDIUMDYNAMIC_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTION
Full Analysis
  • [DYNAMIC_EXECUTION]: The script tools/formula_eval.py calculates results for mathematical formulas using ast.parse() followed by eval() and compile(). This implementation utilizes a restricted environment to mitigate risks by limiting operations to basic arithmetic and specific functions, but the use of dynamic evaluation on variable input remains a security concern.
  • [COMMAND_EXECUTION]: Multiple components, including tools/cad_venv.py, tools/run_step_to_views.py, and tools/bootstrap_cad_venv.py, utilize subprocess functions to execute shell commands. These operations are used for managing a Python virtual environment, performing package installations, and executing conversion utilities for CAD files.
  • [EXTERNAL_DOWNLOADS]: The skill programmatically fetches resources from external sources. tools/bootstrap_cad_venv.py installs required libraries from the Python Package Index (PyPI), and tools/crawl/cnipa_epub_crawler.py retrieves patent data from the official China National Intellectual Property Administration (CNIPA) website using Playwright.
  • [REMOTE_CODE_EXECUTION]: The combined capability of executing shell commands and performing runtime software installation creates a remote code execution surface. Security review is recommended to ensure that inputs processed by these tools, such as formula plans or project files, cannot be manipulated to execute unauthorized code.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Sep 15, 2026, 08:54 AM
Security Audit — agent-trust-hub — patent-disclosure