Security audit

AI大模型专家｜fal.ai 替代方案｜AI-HIVE

Security checks across malware telemetry and agentic risk

Overview

This skill is mostly a disclosed AI-HIVE migration and generation helper, but it has broad auto-invocation and under-disclosed credential persistence that users should review before installing.

Install only if you want AI-HIVE-specific migration help and are comfortable with scripts that can use an AI-HIVE API key, upload chosen media, create billable generation tasks, and write outputs locally. Prefer passing the API key via environment variable or CLI instead of running init unless you accept local credential storage.

SkillSpector

By NVIDIA

Vulnerability Patterns

Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
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

Findings (7)

Lp3

Medium

Category: MCP Least Privilege
Confidence: 91% confidence
Finding: The skill advertises and demonstrates capabilities that access environment variables, local files, shell commands, and networked services, but it does not declare any permissions. This creates a transparency and consent problem: a user or host system may invoke the skill without understanding it can read secrets, upload local media, write outputs, or execute tooling, increasing the chance of unintended data exposure or unsafe execution.

Tp4

High

Category: MCP Tool Poisoning
Confidence: 95% confidence
Finding: The declared purpose is evaluation and migration guidance, but the skill content also drives operational actions such as obtaining/storing API keys, querying account data, calling remote APIs, uploading local media, downloading generated files, and invoking ffmpeg on local content. That mismatch is dangerous because users may consent to an advisory skill while unknowingly triggering credential handling, external data transfer, and local command execution that carry confidentiality, integrity, and cost risks.

Context-Inappropriate Capability

Medium

Confidence: 83% confidence
Finding: The skill interactively collects an API key and persists it to `~/.ai-hive/config.json`, which creates a local secret at rest outside any dedicated credential manager. Although the file mode is tightened to `0600`, storing long-lived API credentials locally increases exposure from local compromise, backups, terminal misuse, or unintended reuse by other workflows, especially for a skill whose stated purpose is advisory rather than credentialed execution.

Description-Behavior Mismatch

Medium

Confidence: 89% confidence
Finding: The file is presented as a skill-specific video generator, but it embeds broader generic capabilities for chat, image generation, model enumeration, uploads, and account access. This expands the skill’s effective permission surface beyond its declared purpose, which is dangerous in agent settings because users or orchestrators may invoke unexpected actions that access data or incur charges outside the intended workflow.

Context-Inappropriate Capability

Medium

Confidence: 93% confidence
Finding: The skill can call a user-info endpoint that returns account information and wallet balance, which is not necessary for a video-generation or alternative-evaluation skill. In an agent context, unnecessary access to billing and identity data violates least privilege and increases the chance of sensitive data exposure or unintended account reconnaissance.

Vague Triggers

Medium

Confidence: 84% confidence
Finding: The trigger list includes broad generic terms like '模型API', 'API聚合', '图片API', and '视频API', which can cause the skill to activate in many unrelated conversations. Over-broad invocation is risky here because the skill is not purely informational—it steers users toward a specific external service and includes workflows involving credentials, uploads, downloads, and executable commands.

Vague Triggers

Medium

Confidence: 93% confidence
Finding: The skill enables implicit invocation without any trigger constraints, so the agent may auto-select this skill in loosely related conversations about AI APIs, migration, or model usage. Because the skill is promotional and domain-specific, overbroad auto-invocation can cause unintended routing, biased recommendations, and unsafe execution of migration guidance without clear user intent confirmation.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.