ClawHub
Back to skill

Security audit

ERNIE-Image文生图SKILL

Security checks for vulnerabilities and agentic risk

Overview

This skill is a straightforward ERNIE-Image text-to-image wrapper, but its documentation overstates image-editing support and users should treat prompts as shared with Baidu.

Install only if you are comfortable using a Baidu AI Studio API key and sending image prompts to Baidu for processing. Treat this version as text-to-image only; do not rely on the advertised --input-image or editing capability. Prefer an environment variable over passing the API key in chat or on the command line, and choose output filenames carefully because existing files can be overwritten.

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 (3)

Tp4

High
Category
MCP Tool Poisoning
Confidence
94% confidence
Finding
The skill claims support for image editing and an `--input-image` option, but the described behavior does not actually implement that functionality. This mismatch can mislead users and downstream agents into handling files or prompts under false assumptions, increasing the chance of unsafe invocation patterns, privacy mistakes, or operational errors.

Vague Triggers

Medium
Confidence
88% confidence
Finding
The activation cue is overly broad and may cause the skill to trigger for many generic image-related requests without sufficient user intent verification. Over-broad triggering is risky here because the skill sends prompts to a third-party API and may process user content in contexts where the user did not explicitly consent to that external transmission.

Missing User Warnings

Medium
Confidence
97% confidence
Finding
The skill omits a clear warning that prompts and any potential input images are sent to Baidu's ERNIE-Image API. This creates a privacy and data-handling risk because users may unknowingly provide sensitive text or images to a third-party service.

Static analysis

No suspicious patterns detected.