inkscape

Warn

Audited by Gen Agent Trust Hub on May 5, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill's operational model relies on the execution of shell commands using the cli-anything-inkscape harness and the Inkscape binary. It specifically utilizes template variables like {pdfPath} and {citationKey} directly within shell execution patterns. This creates a surface for command injection if these values are derived from untrusted user input or external metadata.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection attacks because it processes external vector files without sufficient isolation.
  • Ingestion points: Data enters the agent's context through the document open command which reads existing SVG files from the filesystem.
  • Boundary markers: The instructions lack delimiters or explicit warnings to the agent to disregard instructions that might be embedded in the SVG's XML structure or metadata.
  • Capability inventory: The skill possesses significant capabilities including shell command execution, file reading, and file writing (SVG, PNG, PDF).
  • Sanitization: There is no documentation or instruction for sanitizing SVG content (e.g., removing script tags or malicious metadata) before the agent processes the file objects.
Audit Metadata
Risk Level
MEDIUM
Analyzed
May 5, 2026, 05:36 PM