ClawHub
Back to skill

Security audit

AI大模型专家｜Nano Banana Pro 图片生成与编辑

Security checks across malware telemetry and agentic risk

Overview

The skill mostly matches an AI-HIVE image-generation workflow, but it enables implicit use of a third-party, potentially billable service that can upload local media and use a stored API key without tight trigger scoping.

Install only if you are comfortable sending prompts and selected media to AI-HIVE and storing an AI-HIVE API key locally. Keep implicit invocation disabled or require confirmation before uploads and generation tasks, especially for private product assets or billable API usage.

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 (3)

Lp3

Medium
Category
MCP Least Privilege
Confidence
91% confidence
Finding
The skill documentation describes shell commands, local config writes, environment variable use, file access, and network calls, but no explicit permission declaration is present. This creates a trust and review gap: users may invoke a skill that can access sensitive local resources and external services without clear up-front disclosure, increasing the chance of unintended secret exposure or unsafe execution.

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding
The skill is presented as a Nano Banana Pro image generation/editing workflow, but the documented behavior appears to extend into broader AI-HIVE functions such as generic chat, video generation, wallet/account queries, model enumeration, non-image media upload, and interactive API-key collection/storage. This mismatch weakens user consent and review boundaries: a user may approve an image tool while the underlying implementation can access broader account data, handle additional content types, and trigger operations outside the advertised scope.

Vague Triggers

Medium
Confidence
95% confidence
Finding
The manifest enables implicit invocation with no visible trigger scoping, approval gate, or contextual constraints. That increases the chance the agent will invoke an external image-generation/editing workflow and persist task IDs/results without clear user intent, which is especially relevant here because the skill interfaces with a third-party service and can upload reference materials automatically.

VirusTotal

65/65 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:74