visualize

Pass

Audited by Gen Agent Trust Hub on Jul 27, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes user-provided dataset names and chart descriptions via the $ARGUMENTS variable, which are then used to construct file paths and script content. This presents a surface for indirect prompt injection or path traversal if the input is not sanitized.
  • Ingestion points: User-supplied dataset names and chart descriptions in $ARGUMENTS; CSV data files loaded from the input directory.
  • Boundary markers: None identified in the instructions to delimit user data from code templates.
  • Capability inventory: The skill utilizes 'Write' to create scripts, 'Bash' to execute them, and 'Read' to access local data.
  • Sanitization: No explicit sanitization or validation steps are defined for the user-supplied arguments before interpolation.
  • [COMMAND_EXECUTION]: The skill dynamically generates a Python visualization script at runtime and executes it using the Bash tool. While this is the primary purpose of the skill, the execution of dynamically assembled code is a security consideration.
  • The script utilizes standard libraries: matplotlib, seaborn, and pandas.
  • The process involves creating directories and writing files to the local filesystem.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 27, 2026, 08:50 AM
Security Audit — agent-trust-hub — visualize