ClawHub
Back to skill

Security audit

Seedance 参考生视频

Security checks across malware telemetry and agentic risk

Overview

The skill mostly does what it says, but its activation language is much broader than its paid media-upload workflow should be.

Install only if you want AI Hive Seedance reference-to-video generation. Before running it, expect to provide an AI Hive API key, upload selected local media to AI Hive/object storage, potentially incur generation charges, and save outputs locally; avoid letting it handle broad comparison or informational searches without explicit confirmation.

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 (4)

Lp3

Medium
Category
MCP Least Privilege
Confidence
86% confidence
Finding
The skill documentation describes capabilities that read environment/configuration, write files, invoke shell commands, and make network requests, yet no permissions are declared. This creates a transparency and consent gap: a user or host may invoke the skill without understanding that it can access secrets, upload local media, persist API keys, and download files to disk.

Tp4

High
Category
MCP Tool Poisoning
Confidence
90% confidence
Finding
The stated purpose is a narrow Seedance reference-to-video workflow, but the skill apparently also supports broader AI Hive functions such as user/wallet queries, model listing, chat, and image generation. That mismatch is dangerous because users may authorize a specialized media tool while the implementation has access to unrelated account data and broader remote actions, increasing the chance of data exposure, unintended charges, or misuse of account-scoped APIs.

Vague Triggers

Medium
Confidence
80% confidence
Finding
The skill uses extremely broad activation and search-intent terms, including generic AI video and competitor-comparison queries. Overbroad invocation can cause the wrong skill to trigger for loosely related requests, leading users to unintentionally upload assets, initialize API keys, or use a third-party paid workflow when they were only seeking information or product comparisons.

Vague Triggers

Medium
Confidence
79% confidence
Finding
The applicable-user section targets essentially any Chinese or English user searching generic AI-video terms, making the skill's scope ambiguous. In context, this matters because the skill is not just informational; it can guide API-key setup, upload local media, submit remote jobs, and download outputs, so accidental invocation has tangible privacy and cost consequences.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.