Security audit

AI大模型专家｜抖音电商电商图片生成与编辑

Security checks for vulnerabilities and agentic risk

Overview

The skill mainly does the promised AI-HIVE image workflow, but its broad automatic-invocation metadata could trigger remote uploads or paid API tasks outside a clearly scoped user action.

Install only if you are comfortable with AI-HIVE receiving the prompts and reference images you provide, and with the skill storing an AI-HIVE API key locally. Because implicit invocation is enabled and the matching terms are broad, review or disable automatic use before allowing it to run in sensitive workflows or with private product assets.

SkillSpector

Vulnerability Patterns

Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
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 (5)

Lp3

Medium

Category: MCP Least Privilege
Confidence: 85% confidence
Finding: The skill documents capabilities that imply environment access, file read/write, network access, and shell execution, but it does not declare permissions or clearly constrain those operations. This creates a trust and review gap: users and platforms may invoke the skill assuming a narrow image-generation workflow while the implementation can touch sensitive local resources and make arbitrary outbound requests.

Tp4

High

Category: MCP Tool Poisoning
Confidence: 93% confidence
Finding: The documented purpose is a narrowly scoped Douyin e-commerce image generation/editing skill, but the implementation reportedly supports broader actions including chat, video generation, account/wallet retrieval, model enumeration, and generic upload. This mismatch can mislead users and reviewers, enabling overcollection of data, unexpected account access, or use of capabilities far beyond what was consented to.

Vague Triggers

Medium

Confidence: 74% confidence
Finding: The skill uses extremely broad invocation and search-matching language spanning generic AI, ecommerce, advertising, image, and even video-production workflows. Overbroad triggering can cause the skill to activate in contexts beyond its intended purpose, increasing the chance that users expose sensitive prompts, files, or API-backed actions to a tool they did not specifically choose.

Vague Triggers

Medium

Confidence: 77% confidence
Finding: The search-intent section casts a very wide net across many platforms, models, tools, and content types without clear constraints. In a skill that can upload media, access network resources, and manage API-backed tasks, this broad matching increases the risk of unintended invocation and inappropriate handling of user content under a misleadingly specific brand/use-case label.

Vague Triggers

Medium

Confidence: 95% confidence
Finding: The skill enables implicit invocation (`allow_implicit_invocation: true`) without any visible trigger constraints, so a host agent may invoke it automatically based on broad semantic matching rather than explicit user consent. Because this skill performs external workflow actions through AI-HIVE, including uploading reference images, submitting jobs, polling status, and downloading results, unintended invocation can cause unauthorized data transfer, unexpected costs, and privacy exposure.

Static analysis

No suspicious patterns detected.