ClawHub
Back to skill

Security audit

GLM-V-Caption

Security checks across malware telemetry and agentic risk

Overview

This skill is a coherent Zhipu GLM-V captioning tool, but users should expect submitted media, prompts, and selected local images to be sent to Zhipu for processing.

Install this only if you are comfortable sending chosen media, media URLs, prompts, and local images to Zhipu under your ZHIPU_API_KEY. Use a dedicated revocable API key if possible, monitor quota or billing, and avoid submitting confidential or regulated material unless Zhipu's privacy and retention terms are acceptable.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • 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 (3)

Missing User Warnings

Medium
Confidence
92% confidence
Finding
The skill does not prominently warn that user-provided media URLs, files, or content will be sent to the external ZhiPu API for processing. This can lead to inadvertent third-party disclosure of sensitive documents, images, or metadata, especially because the skill supports documents and multiple input types.

Missing User Warnings

Medium
Confidence
84% confidence
Finding
The tool reads local images, converts them to base64 data URLs, and transmits their contents to a third-party API without an explicit user-facing warning at submission time. In an agent context, this creates a meaningful privacy and data-handling risk because users may not realize local sensitive images are leaving the host boundary.

Ssd 3

Medium
Confidence
93% confidence
Finding
Mandating display of the full raw model output can expose sensitive information extracted from user-supplied documents, images, or videos, including personal, proprietary, or policy-restricted content. The danger is higher here because the skill explicitly handles documents and summarization tasks, where the model output may reproduce confidential source material verbatim.

VirusTotal

67/67 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.