Gpt Image 2 Prompts Skill

• Outcome: Review
• Latest audit: Jul 22, 2026 · 1:47 AM UTC
• Version: 0.1.0

Overview

This appears to be a legitimate prompt-library skill, but it automatically runs setup/update code and downloads remote files, so users should review those behaviors before installing.

Install only if you are comfortable with the skill running a Node setup script during install, refreshing its local prompt database from HuggingFace, and fetching preview images during use. Avoid pasting confidential articles, scripts, notes, or unpublished business content unless your assistant environment's data handling is acceptable to you.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• MCP Least Privilege: Underdeclared Capability, Wildcard Permission, Missing Permission Declaration
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands

Findings (9)

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 91% confidence
• Finding: The skill requires network access by instructing the agent to fetch remote metadata and images, but it does not declare that capability in its metadata. Hidden or undeclared network behavior reduces user and platform visibility into what the skill can do, making remote access harder to audit and constrain.

Context-Inappropriate Capability

Medium

• Confidence: 94% confidence
• Finding: The skill tells the agent to execute shell commands to download remote content into a fixed local path and then delete it, even though the core function is prompt recommendation. This expands the skill from content lookup into file-system and command execution, increasing risk from remote content handling, path collisions, and unnecessary command execution surfaces.

Context-Inappropriate Capability

Medium

• Confidence: 96% confidence
• Finding: The package defines a postinstall hook that automatically executes JavaScript during installation, even though the skill's stated purpose is only to recommend prompts. Install-time code execution creates an unnecessary trust boundary expansion because users must run unreviewed code before they can inspect or use the prompt library, and such hooks can perform filesystem or network actions silently.

Description-Behavior Mismatch

Medium

• Confidence: 93% confidence
• Finding: The manifest states that the package auto-downloads content on install, which exceeds the expected behavior of a prompt recommendation skill and introduces hidden network and setup actions. For a data-only or prompt-library package, automatic retrieval at install time increases supply-chain risk and can expose users to unexpected code paths, remote content changes, or environment interaction.

Missing User Warnings

Medium

• Confidence: 91% confidence
• Finding: The README encourages users to paste articles, scripts, and notes for remixing, but provides no privacy or data-handling warning. In an agent-skill context, users may reasonably assume pasted content stays local, while setup and possible downstream processing could involve external resources or unclear handling, increasing the risk of inadvertent disclosure of proprietary or sensitive content.

Missing User Warnings

Medium

• Confidence: 81% confidence
• Finding: The README explicitly encourages users to paste full articles, video scripts, and notes into the skill, but it does not disclose whether that content may be processed by the hosting AI assistant, sent to remote services, or correlated with externally hosted preview/image data. This creates a real privacy and data-handling risk because users may provide proprietary or sensitive content under the assumption that the skill operates locally and safely.

Missing User Warnings

Medium

• Confidence: 95% confidence
• Finding: The skill mandates automatic freshness checks and silent remote updates on every run, potentially modifying local references without clear user awareness. This creates a supply-chain and transparency risk because remote content is pulled into the local environment automatically, and behavior can change over time based on external data.

Missing User Warnings

Medium

• Confidence: 93% confidence
• Finding: The skill instructs downloading remote images and deleting local files via shell commands without warning the user that network access and file operations will occur. Even if the source domain is intended to be trusted, this behavior increases the attack surface and obscures side effects unrelated to simple prompt recommendation.

Missing User Warnings

Medium

• Confidence: 91% confidence
• Finding: The package advertises automatic download behavior during installation without clear upfront warning about network access and system-side effects. In the context of a prompt-library skill, this lack of disclosure is risky because users may reasonably expect inert content, not install-time retrieval or setup actions that can change over time and behave differently across environments.

VirusTotal

VirusTotal findings are pending for this skill version.

Static analysis

No suspicious patterns detected.