Security audit

Ai Artist Workstation

Security checks for vulnerabilities and agentic risk

Overview

This AI image-order workflow is mostly coherent, but it asks for broad agent authority and would handle face images and order data without enough scoping or disclosure.

Review before installing. Use it only for image-generation order workflows, avoid giving it unnecessary workspace access, and confirm any upload of customer selfies, prompts, or order details to third-party image providers. Do not rely on its security claims unless the surrounding agent or service actually enforces encryption, access control, logging, and command restrictions.

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (4)

Intent-Code Divergence

Medium

Confidence: 92% confidence
Finding: The skill makes concrete security assurances about encrypted data handling, authorized access, backups, and strict access control, but the document only describes a local/CLI workflow and provides no implementation details or enforceable controls. This creates a misleading trust boundary: operators may process prompts, images, and order data under the assumption that confidentiality and authorization protections already exist when they do not.

Intent-Code Divergence

High

Confidence: 97% confidence
Finding: The skill explicitly requires exec capability and accepts user-influenced inputs such as prompts, paths, order metadata, and delivery parameters, yet only claims that command execution is whitelist-restricted without defining any whitelist or safe invocation pattern. In an agent context, this gap can lead to command injection or unsafe file/network operations if later implementations interpolate user-controlled values into shell commands.

Vague Triggers

Medium

Confidence: 78% confidence
Finding: The invocation text is extremely broad, suggesting use whenever AI model calls, agent orchestration, or LLM applications are needed. In agent ecosystems, overbroad activation criteria can cause the skill to be selected in unrelated contexts, increasing the chance that it handles inappropriate data, invokes external services unexpectedly, or performs exec-enabled workflows without clear user intent.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The skill processes user prompts and reference images through external image-generation services, but it does not clearly warn users that these inputs may be transmitted to third-party APIs. Because the workflow includes face images and commercial order data, the privacy sensitivity is elevated and users may unknowingly disclose personal or proprietary content.

Static analysis

No suspicious patterns detected.