nano-banana-2
Pass
Audited by Gen Agent Trust Hub on Sep 2, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill is a documentation-only resource providing instructions for interacting with the inference.sh image generation platform. It defines how an agent should use the 'belt' CLI tool to perform inference tasks.
- [EXTERNAL_DOWNLOADS]: The skill references configuration and installation files located at
https://raw.githubusercontent.com/inference-sh/skills/refs/heads/main/cli-install.md. This is a standard documentation source for the service. - [INDIRECT_PROMPT_INJECTION]: The skill defines an interface for processing untrusted user data, which is a common pattern for image generation skills.
- Ingestion points: User-provided
promptandimagesURLs are processed inSKILL.mdvia CLI and SDK examples. - Boundary markers: The instructions guide the agent to use JSON-formatted strings (e.g.,
--input '{"prompt": "..."}') to encapsulate data, which provides a clear boundary between command and data. - Capability inventory: The skill uses the
beltCLI tool (via Bash) and theinferenceshPython library to perform remote inference operations. - Sanitization: The skill relies on structured JSON input to prevent command injection at the CLI level.
Audit Metadata