ClawHub
Back to skill

Security audit

Comfyui Handler

Security checks across malware telemetry and agentic risk

Overview

The skill is a coherent ComfyUI integration, but it needs Review because a remote ComfyUI server can control downloaded filenames and potentially write files outside the intended output folder.

Install only if you trust the configured ComfyUI host and are comfortable sending prompts and input images to it. Until filename sanitization is added, avoid using this skill with untrusted or shared ComfyUI servers, and require explicit confirmation before sending outputs through messaging plugins or opening local files.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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)

Context-Inappropriate Capability

Medium
Confidence
87% confidence
Finding
The skill documentation explicitly encourages use of external messaging plugins and local file openers, which extends behavior beyond the stated ComfyUI image-generation/control purpose. This broadens the operational scope from image generation to data exfiltration or arbitrary local application invocation, increasing the chance an agent will send generated or user-supplied files to third-party services or execute local actions without clear user consent.

Missing User Warnings

Medium
Confidence
91% confidence
Finding
The README explicitly states that the skill will automatically upload local input images to a remote ComfyUI host and download generated outputs, but it does not warn users about the network transfer, trust boundary, or sensitivity of the data being sent. In an agent skill context, this is dangerous because users or downstream agents may invoke the tool assuming local-only processing, leading to unintended disclosure of private images or generated content to a remote system.

Missing User Warnings

Medium
Confidence
79% confidence
Finding
Remote content from the ComfyUI server is written directly to disk using a server-controlled filename without validation or sanitization. If the configured server is malicious or compromised, it could return filenames containing path traversal sequences or overwrite unexpected files under the workspace, making this more dangerous than a mere lack of disclosure.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.