nano-banana
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references installation scripts and documentation from the inference.sh website and GitHub organization (github.com/inference-sh) within SKILL.md.
- [COMMAND_EXECUTION]: The skill uses the 'belt' CLI tool for authentication and application execution via the Bash tool in SKILL.md. The frontmatter configuration limits the scope of command execution to 'belt' commands only.
- [INDIRECT_PROMPT_INJECTION]: The skill accepts user-provided prompts and image URLs which are passed to the Gemini model, representing an attack surface for indirect prompt injection in SKILL.md. 1. Ingestion points: User input fields 'prompt' and 'images' in SKILL.md. 2. Boundary markers: Parameters are enclosed in JSON strings within CLI calls. 3. Capability inventory: Bash tool usage (restricted to 'belt' command) in SKILL.md. 4. Sanitization: No input sanitization is defined within the skill instructions.
Audit Metadata