emojigen-nano-banana

Pass

Audited by Gen Agent Trust Hub on Mar 9, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes user-defined configurations that are interpolated into AI prompts.
  • Ingestion points: The skill reads from assets/example-config.json and user-supplied images.
  • Boundary markers: The prompt building logic in scripts/emojigen.mjs lacks explicit delimiters or instructions to ignore instructions within user-provided fields like characterNotes.
  • Capability inventory: The script can perform file operations and execute system commands.
  • Sanitization: No comprehensive sanitization is performed on text interpolated into the prompts.
  • [COMMAND_EXECUTION]: The skill relies on external system binaries and official SDKs.
  • Evidence: It executes magick for image processing, pngpaste for clipboard tasks, and osascript for macOS-specific actions. It also utilizes the @google/genai library to interface with Google's AI services.
  • Risk: Commands are executed via spawn with hardcoded names, though the AppleScript for clipboard access is constructed by interpolating a file path into a script string.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 9, 2026, 12:44 AM