data-visualization
Pass
Audited by Gen Agent Trust Hub on Feb 25, 2026
Risk Level: SAFEREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [REMOTE_CODE_EXECUTION]: The documentation provides an installation command for the vendor's CLI using
curl -fsSL https://cli.inference.sh | sh. This is a standard vendor-provided bootstrap script used to deploy the CLI tool. - [COMMAND_EXECUTION]: The skill utilizes the
infshtool via the Bash capability to run remote applications, specificallyinfsh/python-executorandinfsh/html-to-image, which are intended for processing code and rendering visuals. - [EXTERNAL_DOWNLOADS]: The skill installation script fetches binary files from
dist.inference.sh, and the documentation suggests adding related vendor skills usingnpx, both of which are standard software distribution methods for the platform. - [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface by providing templates that interpolate data into executable Python or HTML code. (1) Ingestion points: The chart data and labels provided by the user. (2) Boundary markers: No explicit markers are used to separate user-provided data from the chart-generating logic in the templates. (3) Capability inventory: The skill uses
infsh/python-executor, which allows for arbitrary Python execution to create visualizations. (4) Sanitization: Input data is embedded directly into code strings without explicit validation or escaping within the provided examples.
Audit Metadata