Nano banana 2

• Outcome: Pass
• Latest audit: May 29, 2026, 8:21 PM
• Version:
  1.0.6

Overview

This is a coherent Pixwith image-generation skill, but users should know that prompts and selected local images are sent to Pixwith and upload storage.

Install only if you are comfortable using a Pixwith API key and sending selected prompts and images to Pixwith and its upload storage provider. Use a dedicated revocable API key, monitor credit use, avoid sensitive personal or proprietary images unless Pixwith's policies are acceptable, and ask the agent to disable prompt optimization when prompt translation or rewriting is not desired.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code
• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep

Findings (3)

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: The skill instructs the agent to upload user-provided local images to external services, including a presigned S3 endpoint, without an explicit warning that the files leave the local environment and are transferred to third-party infrastructure. In the context of user photos and local files, this creates a real privacy and data-transfer risk because users may not understand that sensitive images are being sent off-device.

Natural-Language Policy Violations

Medium

• Confidence: 88% confidence
• Finding: The skill enables automatic prompt translation to English by default, which means user input may be transformed and transmitted in altered form without explicit opt-in. This can expose sensitive prompt content to additional processing paths and may change user intent, which is especially relevant when prompts contain private or regulated information.

Natural-Language Policy Violations

Medium

• Confidence: 89% confidence
• Finding: The examples and defaults normalize automatic English translation without presenting any language choice or consent step. In a skill that forwards prompts to an external image API, this increases privacy and integrity risk because user-supplied text may be silently rewritten before processing.

VirusTotal

51/51 vendors flagged this skill as clean.