skills/tryopendata/skills/svg-design/Gen Agent Trust Hub

svg-design

Pass

Audited by Gen Agent Trust Hub on Sep 17, 2026

Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to use standard system commands including cp for asset management and open (macOS) or xdg-open (Linux) to launch a local HTML preview page in the user's default browser.
  • [DYNAMIC_EXECUTION]: The design workflow involves generating a local JavaScript data file (variants.js) containing project metadata and variant details, which is then dynamically loaded and executed by the preview.html scaffold to render the designs.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests user preferences via AskUserQuestion and processes existing SVG graphics. This untrusted data is interpolated into generated code and preview scripts, creating a potential injection surface.
  • Ingestion points: User design preferences (mood, focus, inspiration) and existing SVG files provided for editing.
  • Boundary markers: There are no explicit delimiters or specific safety instructions defined for interpolating user-provided text into the generated variants.js data structure.
  • Capability inventory: Writing local files (SVG, JavaScript) and executing system-level commands to manage files and launch browser processes.
  • Sanitization: The provided preview.html asset includes a basic HTML entity escaping function (esc) used for rendering variant names and descriptions.
  • [EXTERNAL_DOWNLOADS]: The skill recommends installing the well-known svgo optimization utility via the global npm package manager to minify SVG outputs.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 17, 2026, 03:58 PM
Security Audit — agent-trust-hub — svg-design