ClawHub
Back to skill

Security audit

Box-KVCache

Security checks across malware telemetry and agentic risk

Overview

This local LLM cache tool mostly matches its stated purpose, but it needs review because one loader uses pickle-enabled NumPy loading that can execute code from a malicious cache file.

Install only if you are comfortable reviewing or patching the pickle-enabled load path first. Do not load compressed cache files from other people unless allow_pickle is removed or the file is trusted. Run the scripts as an unprivileged user, expect local Ollama/system inspection commands, and stop Ollama manually if you do not want it left running.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (1)

subprocess module call

Medium
Category
Dangerous Code Execution
Content
def check_ollama_running() -> bool:
    """检查 Ollama 是否在运行"""
    result = subprocess.run(
        "tasklist | findstr ollama",
        shell=True, capture_output=True, text=True
    )
Confidence
95% confidence
Finding
result = subprocess.run( "tasklist | findstr ollama", shell=True, capture_output=True, text=True )

VirusTotal

67/67 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.