ClawHub
Back to skill

Security audit

AI Act Check - 欧盟 AI Act 合规检查工具

Security checks across malware telemetry and agentic risk

Overview

This is a disclosed EU AI Act self-check skill that sends checklist answers to a fixed cloud scoring service and stores only an API key or anonymous trial ID locally.

Install only if you are comfortable sending your EU AI Act checklist answers to compliancehub.cn for scoring. Use the offline preview mode for local-only review, and prefer the environment variable or a private 0600 key file for the API key.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • 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 (5)

Lp3

Medium
Category
MCP Least Privilege
Confidence
70% confidence
Finding
Without declared permissions the skill's intent is opaque and cannot be validated.

Intent-Code Divergence

Low
Confidence
97% confidence
Finding
The docstring says the function fetches "pipl-check items from the cloud rule library," but the code actually requests this skill's EU AI Act rule URL via RULES_URL derived from the skill slug. This is an active documentation contradiction, likely from copy-paste, and can mislead reviewers about what dataset is being accessed.

Natural-Language Policy Violations

Low
Confidence
83% confidence
Finding
The manifest states 'Locale: zh-CN' and says interaction and prompts are primarily in Chinese, with English only available on demand. This sets a default language preference rather than offering an explicit user choice up front, which can conflict with language/locale policy expectations.

Natural-Language Policy Violations

Medium
Confidence
93% confidence
Finding
The generated HTML hard-codes `lang="zh-CN"` and presents the report UI text in Chinese only. This is a natural-language locale policy issue because the skill does not offer users a language choice or clearly justify why output must be Chinese.

Env Variable Harvesting

High
Category
Data Exfiltration
Content
1) env COMPLIANCEHUB_API_KEY (ephemeral, safest)
      2) private store ~/.config/compliancehub/<slug>.key (mode 0600)
    """
    env_key = os.environ.get("COMPLIANCEHUB_API_KEY")
    if env_key and env_key.strip():
        return env_key.strip()
    p = _key_path()
Confidence
70% confidence
Finding
os.environ.get("COMPLIANCEHUB_API_KEY

VirusTotal

50/50 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.