ClawHub
Back to skill

Security audit

Text-to-Carousel

Security checks across malware telemetry and agentic risk

Overview

This skill appears to do what it claims: generate carousel images with Gemini, with normal privacy considerations for sending user content and optional images to a third-party API.

Install only if you are comfortable using a Gemini API key and sending carousel prompts, fetched article contents, URLs, and any selected product images to Google Gemini. Avoid using confidential drafts, unpublished business material, or private images unless that external processing is acceptable.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (3)

Vague Triggers

Medium
Confidence
83% confidence
Finding
The activation text is very broad, covering common phrases like creating slides, social media graphics, or converting articles, which can cause the skill to trigger in situations where the user did not intend external fetching or image generation. Over-broad routing increases the chance of unexpected processing of sensitive text, URLs, or brand assets.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The workflow instructs fetching article URLs, WordPress posts, and product images, then sending prompts and optionally inline image data to the Gemini API, but the skill description does not warn users that external retrieval and third-party transmission will occur. This is a real privacy and data-handling issue because user-supplied URLs, article contents, and images may contain confidential or proprietary information.

Natural-Language Policy Violations

Medium
Confidence
72% confidence
Finding
The instruction to always include a Chinese-text requirement creates an unconditional language constraint that can override user intent and introduce prompt-level policy drift. In a multilingual content-generation skill, hard-coding such requirements can cause incorrect outputs, confuse downstream model behavior, and increase the chance that user instructions are silently distorted.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.