ClawHub
Back to skill

Security audit

comfyui-runner

Security checks for vulnerabilities and agentic risk

Overview

The skill is a small ComfyUI health-check CLI with documentation mismatches, but no evidence of hidden, destructive, persistent, or exfiltrating behavior.

Install only if you need a status check for a ComfyUI server at the configured host and port. Do not rely on this version to start or stop ComfyUI, and avoid placing credentials in its .env file unless a future version actually uses them and documents how they are protected.

A.I.G

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs

SkillSpector

Vulnerability Patterns
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (3)

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding
The skill claims start/stop/status control of a ComfyUI instance, but the observed behavior only probes a health endpoint over HTTP and does not implement the promised run/stop actions. This mismatch is security-relevant because operators may authorize the skill under false assumptions, and hidden or incomplete behavior reduces auditability and can mask future capability creep or unauthorized network interactions. In this context, the discrepancy is more dangerous because the skill targets a specific internal IP and may be used in automation where accurate capability declaration is important.

Lp3

Medium
Category
MCP Least Privilege
Confidence
86% confidence
Finding
The skill advertises network access and environment-variable-based configuration but does not declare any explicit tool scope such as permissions or allowed-tools. This weakens security boundaries because a runtime may grant broader capabilities than reviewers or users expect, especially given the use of curl and credential-bearing environment variables. In the context of a skill that talks to a local service, undeclared network and env access increases the chance of unintended data exposure or misuse.

Description-Behavior Mismatch

Medium
Confidence
97% confidence
Finding
The manifest description says the skill supports start/stop/status for a ComfyUI instance, implying lifecycle control operations. In this file, any action other than "status" is explicitly rejected, so the implemented behavior is narrower than the stated capability.

Static analysis

Detected: suspicious.env_credential_access

Environment variable access combined with network send.

Critical
Code
suspicious.env_credential_access
Location
bin/cli.js:4