NYC

nanobanana-skill

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [Indirect Prompt Injection] (LOW): The skill processes user-provided strings and image files which are passed to the Gemini API, creating a vulnerability surface where malicious instructions could influence the model behavior.\n
  • Ingestion points: User-specified --prompt and --input files.\n
  • Boundary markers: None identified in the provided instructions or script usage.\n
  • Capability inventory: The skill uses Bash(python3:*) to execute its primary logic.\n
  • Sanitization: No input sanitization or validation is mentioned.\n- [Unverifiable Dependencies] (LOW): The skill relies on external Python packages (google-genai, Pillow, python-dotenv) installed via pip at runtime.\n- [Data Exposure] (LOW): The skill documentation points to ~/.nanobanana.env as a location for storing the GEMINI_API_KEY. Storing secrets in plain-text environment files in a predictable home-directory path is a minor data exposure risk.\n- [Command Execution] (LOW): The skill utilizes the bash tool to execute a local Python script, which is the intended mechanism for its operation.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:31 PM