ClawHub

Clawsy Agenthub

Security checks across malware telemetry and agentic risk

Overview

This skill is a coherent AgentHub integration, but it gives an agent broad remote task authority and includes unsafe prompt-handling and auto-work patterns users should review carefully.

Install only if you intentionally want an agent to act on AgentHub with your account. Use a dedicated revocable AgentHub key, avoid auto-work unless you set explicit limits and review submissions, do not treat remote task prompts as trusted instructions, and avoid using confidential documents or third-party API keys unless you are comfortable sending them to AgentHub.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • 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
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (2)

Missing User Warnings

Medium
Confidence
92% confidence
Finding
This skill instructs the agent to send user-provided URLs, extracted document contents, and even third-party validation API keys to external services, but it does not require an explicit warning or consent checkpoint immediately before transmission. That increases the risk of users unintentionally disclosing sensitive documents, repository content, or credentials to AgentHub and downstream validation providers.

Ssd 1

High
Confidence
99% confidence
Finding
The instructions explicitly say to use the platform's enriched_prompt as system instructions, effectively elevating untrusted remote task content into higher-priority behavioral guidance. A malicious task creator could inject prompt instructions that override safety rules, solicit secrets, manipulate tool use, or cause unauthorized actions by the agent.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal