ClawHub
Back to skill

Security audit

Local GMNCODE Vision Pro

Security checks for vulnerabilities and agentic risk

Overview

This vision skill appears purpose-built for GMNCODE image analysis, but it uploads local images to an external API without clear user-facing disclosure or consent controls.

Install only if you are comfortable with selected images or screenshots being uploaded to gmncode.cn using your GMNCODE_API_KEY. Avoid sensitive screenshots, private documents, credentials, customer data, or internal UI captures unless you have explicit approval and trust the provider’s data handling. A safer version should add a clear upload warning, an explicit opt-in flag, file/path limits, and guidance for redacting sensitive content.

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Lp3

Medium
Category
MCP Least Privilege
Confidence
93% confidence
Finding
The skill advertises executable scripts, an environment variable, and an external model route, which implies shell, env, and network capabilities, but it does not declare permissions or warn about those capabilities. This creates a transparency and governance gap: an agent or reviewer may treat the skill as lower risk than it is, increasing the chance of unintended credential access or external data transmission.

Vague Triggers

Medium
Confidence
84% confidence
Finding
The description contains broad trigger language such as 'when built-in image tools are unavailable or unreliable' and 'production-ready visual automation,' which can match many common requests. Over-broad activation increases the chance the skill is invoked unnecessarily, exposing more user data, triggering external API use, or causing script execution in cases where simpler and safer tooling would suffice.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The skill states that it uses a `GMNCODE_API_KEY` and a remote model route but does not warn that data may leave the local environment or that credentials will be consumed. In a vision skill, this is especially sensitive because screenshots and images may contain secrets, personal data, or internal UI content, so silent external transmission materially raises confidentiality risk.

Missing User Warnings

Medium
Confidence
88% confidence
Finding
The script base64-encodes the provided local image and transmits it to an external API, but provides no user-facing disclosure, consent step, or data-sensitivity warning. In an agent or automation context, this can cause unintentional disclosure of screenshots, documents, or other sensitive visual data to a third party.

Static analysis

No suspicious patterns detected.