Security audit

商品360环绕展示视频｜AI-HIVE

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed AI-HIVE product-video workflow that uses user-provided media and credentials to create or process videos, with no artifact-backed deception or automatic unsafe behavior.

Install only if you intend to use AI-HIVE for product video generation. Review prompts and routing before submitting paid tasks, avoid uploading sensitive or unlicensed media, and store the API key only on machines you trust.

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 (7)

Lp3

Medium

Category: MCP Least Privilege
Confidence: 92% confidence
Finding: The skill embeds executable workflows that use environment variables, local file I/O, network access, and shell-adjacent tooling (for example ffmpeg) without declaring permissions or clearly constraining those capabilities. This creates a hidden trust boundary: a user or host may believe the skill is only a planning/template tool, while it can actually access sensitive local resources and invoke external services that may incur cost or expose data.

Tp4

High

Category: MCP Tool Poisoning
Confidence: 96% confidence
Finding: The documented purpose is narrow, but the described/observed behavior includes broader AI-HIVE chat, image generation, model listing, wallet queries, uploads, and local media manipulation unrelated to the stated product-360 workflow. This mismatch is dangerous because reviewers and users may grant trust, inputs, or execution approval under a narrower mental model, while the skill can perform materially broader actions with privacy, billing, and data-exfiltration implications.

Description-Behavior Mismatch

High

Confidence: 98% confidence
Finding: The file presents itself as a specialized product-360 video skill, but actually embeds a broad general-purpose AI client with chat, image generation, model listing, uploads, and account access. This scope mismatch is dangerous because a caller or reviewer may grant the skill permissions and trust appropriate for a narrow media workflow while it can perform materially broader actions than advertised.

Intent-Code Divergence

Medium

Confidence: 94% confidence
Finding: The module docstring and CLI description explicitly describe a universal AI capability caller rather than a narrowly scoped product-orbit video skill. Misleading identity and documentation weaken security review, increase the chance of overbroad deployment, and make it easier to smuggle extra capabilities past policy or operator expectations.

Context-Inappropriate Capability

Medium

Confidence: 91% confidence
Finding: Exposing account information retrieval is unrelated to generating product-orbit videos and expands the data-access surface without clear necessity. In a narrow media-production skill, unnecessary account and wallet visibility can leak sensitive metadata and enables capability creep beyond the declared purpose.

Context-Inappropriate Capability

High

Confidence: 97% confidence
Finding: The generic text-chat path enables arbitrary prompt execution against external models, which is outside the declared product-orbit video role. In context, that broadens the skill into a general LLM proxy, increasing risk of policy bypass, unreviewed data egress, and misuse under the cover of a specialized media skill.

Context-Inappropriate Capability

Medium

Confidence: 92% confidence
Finding: Standalone image generation is broader than the stated task of producing product-orbit videos and is not justified by the skill description. This adds unneeded capability that can be abused for unrelated content generation and undermines least-privilege expectations for the skill.

Static analysis

No suspicious patterns detected.