ClawHub
Back to skill

Security audit

Model Center

Security checks for vulnerabilities and agentic risk

Overview

The NVIDIA API functionality is mostly coherent, but the skill documentation contains unrelated merged learning-history records with user-like identifiers and needs clearer consent around sending prompts to NVIDIA.

Review before installing. The code itself appears to be a straightforward NVIDIA API client, but the published documentation contains unrelated user-like records that should be removed, and users should avoid sending secrets, proprietary text, personal data, or regulated content through the API unless they are comfortable with NVIDIA receiving that data.

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Description-Behavior Mismatch

Low
Confidence
88% confidence
Finding
The embedded 'B站学习' and merged-learning sections are unrelated to the stated NVIDIA model skill and indicate the file contains extraneous, likely user-derived content. Unrelated historical content in a skill manifest is dangerous because it can leak private data, pollute agent behavior, and signal that documentation may have been automatically or carelessly merged from untrusted sources.

Vague Triggers

Medium
Confidence
80% confidence
Finding
The trigger phrases are broad terms such as model-related and API-related concepts that can match many ordinary conversations, increasing the chance the skill activates outside intended contexts. Overbroad activation is risky here because the skill can access an API key and send prompts or data to external services, so accidental invocation can cause unintended disclosure or external calls.

Missing User Warnings

Medium
Confidence
94% confidence
Finding
The documentation describes chat, embedding, and image-generation API operations using an NVIDIA API key but does not clearly warn that user prompts, images, or text inputs may be transmitted to a third-party service. This undermines transparency and informed consent, especially for sensitive prompts, proprietary data, or personal information entered by users.

Ssd 3

Medium
Confidence
96% confidence
Finding
The learning-history section exposes likely user-derived identifiers, usernames, UUID-like values, and media references unrelated to the skill’s function. This is a direct data-leakage issue: even if the identifiers are partial or contextual, publishing unrelated user-linked records in a skill artifact can expose personal data, internal provenance, or cross-context information to anyone reviewing or using the skill.

Static analysis

No suspicious patterns detected.