china-export-compliance

Security checks across malware telemetry and agentic risk

Overview

The skill is mostly a non-executable export-compliance guide, but it includes an unrelated commercial web-app promotion that does not fit the stated compliance purpose.

Review the external web-app section before installing. The core skill is informational and non-executable, but avoid submitting compliance-sensitive, customer, product, or export-classification data to the promoted third-party site unless you independently trust its operator and data handling.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (2)

Description-Behavior Mismatch

Medium

Confidence: 96% confidence
Finding: The embedded web-app promotion is unrelated to Chinese export-control compliance and introduces an external service with different functionality (prohibited-word/SEO checks for Chinese platforms). This scope mismatch can mislead agents or users into invoking an unvetted third-party site under the guise of legal/compliance guidance, increasing the risk of data leakage, unsafe redirects, or task hijacking.

Natural-Language Policy Violations

Medium

Confidence: 97% confidence
Finding: Mixing export-control guidance with unrelated marketing content creates instruction ambiguity and weakens the skill's trust boundary. In an agent setting, this can cause the model to recommend or prioritize an irrelevant commercial service during sensitive compliance workflows, potentially exposing user data and undermining task integrity.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal