ClawHub

Code Runner Local

Security checks across malware telemetry and agentic risk

Overview

This skill does what it claims by running code locally, but users should treat any snippet it runs as fully trusted software.

Install only if you want an agent to run code on your machine. Review snippets before execution, avoid running untrusted code, and use a disposable container or sandbox for code that might read files, access the network, run shell commands, or modify the system.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (2)

Vague Triggers

Medium
Confidence
94% confidence
Finding
The skill’s invocation guidance is extremely broad—covering generic requests to run, test, verify, or demonstrate code—without meaningful exclusions or stronger trust boundaries. In an agent setting, that increases the chance the skill is invoked for untrusted user-supplied code, which can lead to execution of filesystem, network, or system-command payloads even though the document includes only high-level cautions later.

Missing User Warnings

Medium
Confidence
98% confidence
Finding
This script is intentionally designed to execute user-supplied code across many languages, including shell-capable runtimes, by constructing and invoking system commands with exec(). In an agent-skill context, that creates a real remote code execution surface on the host environment with access to local files, network, installed tools, and interpreter behavior; the timeout only limits duration, not capability.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal