ClawHub

Agent Gateway

Security checks across malware telemetry and agentic risk

Overview

The skill has a coherent LLM-gateway purpose, but its privacy claims understate external provider data sharing and it relies on an unpinned external npm CLI for the real gateway behavior.

Review before installing. Use this only if you trust the external `@claws-shield/cli` package and understand that cloud model routing may send prompts, request metadata, and provider credentials to Anthropic, OpenAI, or Google. Prefer a pinned CLI version, expose only the API keys you intend to use, set provider-side limits, and use local Ollama only if you need requests to stay local.

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 (3)

Intent-Code Divergence

High
Confidence
98% confidence
Finding
The privacy statement is materially misleading because the gateway is explicitly designed to send requests to Anthropic, OpenAI, and Gemini. In that operating mode, prompts and authentication material are necessarily used in communications with third-party providers, so claiming API keys are never transmitted to third parties can cause users to make unsafe trust and data-sharing decisions.

Missing User Warnings

Medium
Confidence
94% confidence
Finding
The skill omits an important warning that user prompts and request contents will leave the local environment when cloud-backed models are selected. This is dangerous because users may route sensitive data through the gateway under the mistaken belief that all processing remains local, especially given the surrounding privacy-focused language.

Natural-Language Policy Violations

High
Confidence
99% confidence
Finding
This is a true vulnerability because the file makes an affirmative privacy assurance that conflicts with the documented third-party routing functionality. Misleading privacy representations can directly cause data exposure by encouraging users to input confidential prompts or deploy the gateway in environments with stricter data residency or vendor-sharing requirements.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal