Security audit

AI大模型专家｜AI API中转站

Security checks across malware telemetry and agentic risk

Overview

This appears to be a disclosed AI-HIVE API helper, though users should know it stores an API key locally and can upload chosen media to AI-HIVE.

Install only if you intend to use AI-HIVE with your own authorized API key and media. Treat generated API calls as potentially billable, avoid uploading sensitive or unlicensed files, and prefer environment variables or a protected local config for credentials.

SkillSpector

By NVIDIA

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: 88% confidence
Finding: The skill advertises shell, network, filesystem, and environment-dependent workflows but does not declare permissions up front. That creates a transparency and consent problem: a user or host may invoke scripts that read local files, use local secrets, write outputs, and make remote API calls without clear permission gating. In an API-relay skill that handles keys and task artifacts, undeclared capabilities are more dangerous because they naturally intersect with sensitive credentials and external data transfer.

Tp4

High

Category: MCP Tool Poisoning
Confidence: 81% confidence
Finding: The documented purpose centers on API aggregation and relay management, but the skill also appears to include additional behaviors such as blueprint generation, fixed-model media generation tools, browser-assisted key acquisition, wallet/user queries, and chat calls. This mismatch reduces informed consent and can mislead operators about what the skill will do, especially when it touches authentication flows, account information, and billing-related resources. In a credential-handling relay context, hidden or under-described behaviors materially increase security risk.

Description-Behavior Mismatch

High

Confidence: 95% confidence
Finding: The script’s generated blueprint materially diverges from the skill’s stated purpose of AI API relay management and instead emits short-drama/content-production artifacts such as storyboards and prompt lists. In an agent ecosystem, this kind of capability mismatch can mislead downstream automation, cause operators to trust incorrect outputs, and potentially route sensitive business workflows into unrelated content-generation processes.

Intent-Code Divergence

Medium

Confidence: 88% confidence
Finding: The module docstring states it creates a production brief for an AI API relay product, but the body produces creative-planning content unrelated to relay administration. Mislabeling functionality in security-sensitive agent environments is dangerous because reviewers, orchestrators, or users may grant the script inappropriate trust, permissions, or operational use based on false expectations.

Vague Triggers

Medium

Confidence: 74% confidence
Finding: The trigger language is broad and overlaps with many generic API, gateway, relay, and AI integration requests, which can cause the skill to activate outside its intended scope. Over-broad activation is risky because this skill can steer users toward running scripts, handling API keys, and interacting with external services even when a narrower or safer response would suffice. The context makes this more concerning because the skill is positioned around token management and relay operations.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

Detected: suspicious.exposed_secret_literal

File appears to expose a hardcoded API secret or token.

Critical

Code: suspicious.exposed_secret_literal
Location: SKILL.md:126