Cs Free Image Generator Nv

• Outcome: Pass
• Latest audit: May 29, 2026 at 8:32 PM UTC
• Version:
  1.3.0

Overview

This skill is a straightforward NVIDIA-backed image generator, with some documentation gaps users should understand before use.

Install only if you are comfortable providing an NVIDIA_API_KEY, sending image prompts to NVIDIA's API, and having generated response JSON saved under /tmp. Avoid putting secrets, private data, or confidential business content in prompts, and be aware the documented model name should be clarified by the publisher.

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
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• MCP Least Privilege: Underdeclared Capability, Wildcard Permission, Missing Permission Declaration
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection

Findings (5)

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 88% confidence
• Finding: The skill documentation indicates use of environment variables, outbound network access to the NVIDIA API, and file writes to /tmp, but no permissions are declared. This creates a transparency and policy-enforcement gap: users and hosts cannot accurately assess or constrain the skill's capabilities, and a compromised or modified script could exfiltrate secrets or persist sensitive outputs without prior declaration.

Description-Behavior Mismatch

Medium

• Confidence: 80% confidence
• Finding: The manifest says the skill uses NVIDIA MoCL via NVIDIA API Playground, but the code calls a different model endpoint, black-forest-labs/flux.2-klein-4b. This mismatch undermines transparency and trust, and could cause users to send data under incorrect assumptions about the provider, model, or handling characteristics.

Context-Inappropriate Capability

Low

• Confidence: 83% confidence
• Finding: The script silently reads a user-specific ~/.openclaw/.env file to load credentials, which is broader capability than a simple prompt-to-image wrapper implies. Hidden credential loading increases surprise and weakens reviewability, especially in agent skill contexts where users may not expect filesystem access to personal configuration files.

Vague Triggers

Medium

• Confidence: 81% confidence
• Finding: The trigger phrases include very broad requests such as '生成图片', '画一张图', and 'text to image', which can overlap with many ordinary user intents and cause the skill to activate unexpectedly. Overbroad activation increases the chance of unintended network calls, file writes, and use of API-backed functionality without clear user consent or the most appropriate tool being selected.

Missing User Warnings

Medium

• Confidence: 87% confidence
• Finding: The script transmits the user-provided prompt to an external NVIDIA API without any explicit warning or consent mechanism. In a skill setting, prompts may contain sensitive personal or proprietary information, so undisclosed network transmission creates a privacy and data-governance risk.

VirusTotal

56/56 vendors flagged this skill as clean.