Security audit

Temu 电商图片生成与编辑

Security checks for vulnerabilities and agentic risk

Overview

This skill is mainly a Temu-focused AI image workflow that uses an external API and local API-key config, with some broader unused client code present but no artifact-backed malicious behavior.

Install only if you are comfortable sending selected product/reference media and prompts to the AI Hive API and storing an API key locally. Keep API keys private, confirm cost on a single SKU before batch generation, and do not provide unrelated private files as image references or uploads.

SkillSpector

Vulnerability Patterns

Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
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

Findings (6)

Lp3

Medium

Category: MCP Least Privilege
Confidence: 91% confidence
Finding: The skill exposes shell, network, file read/write, and environment access but does not declare permissions, preventing users and policy systems from understanding its effective capabilities. This is dangerous because hidden capabilities can enable data exfiltration, local file access, or command execution beyond the user’s expected image-generation workflow.

Tp4

High

Category: MCP Tool Poisoning
Confidence: 96% confidence
Finding: The documented purpose is narrowly scoped to Temu ecommerce image creation, but the underlying behavior reportedly includes API key setup, account and wallet queries, generic chat, media upload, and video generation. This mismatch is dangerous because users may authorize or run the skill expecting limited image editing while it actually has broader access paths and functionality that can collect credentials, incur costs, or process unrelated sensitive data.

Description-Behavior Mismatch

Medium

Confidence: 95% confidence
Finding: The file is presented as a Temu image-generation/editing skill, but the implementation is a generic AI Hive client exposing chat, image, video, model listing, media upload, and account-query capabilities. This violates least-privilege for the declared skill purpose and expands what a caller can do if they invoke the script directly, increasing abuse potential beyond Temu product-image workflows.

Context-Inappropriate Capability

Medium

Confidence: 97% confidence
Finding: The chat_text pathway gives this skill unrestricted text-chat capability, which is unrelated to the advertised Temu image-editing scope. In a tool-execution environment, that hidden extra capability can be repurposed to send arbitrary prompts and uploaded media to external models, bypassing the intended task boundary.

Context-Inappropriate Capability

Medium

Confidence: 96% confidence
Finding: The script contains full video-generation support despite being described as an image-generation/editing skill. This broadens data egress and cost/abuse surface, enabling submission of arbitrary prompts and media for video tasks that are outside the declared business purpose.

Context-Inappropriate Capability

Low

Confidence: 88% confidence
Finding: User-info and model-listing features are not necessary for Temu image production and reveal account metadata, available model inventory, and potentially wallet/billing-related information. While not immediately code-execution dangerous, they expand reconnaissance and information exposure available through a supposedly narrow skill.

Static analysis

No suspicious patterns detected.