ClawHub
Back to skill

Security audit

Seedream 5.0 Lite 商品精修

Security checks across malware telemetry and agentic risk

Overview

This skill is a disclosed AI Hive product-photo retouching helper that uploads user-chosen images, stores an API key only if initialized, and downloads generated results.

Install only if you are comfortable sending selected product images and prompts to AI Hive, storing an API key locally if you use init, and reviewing outputs carefully so retouching does not misrepresent real product condition, labels, quantity, or other commercial facts.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • 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
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Lp3

Medium
Category
MCP Least Privilege
Confidence
90% confidence
Finding
The skill invokes a local Python script with shell, network, environment-variable, and file read/write capabilities, but does not declare those permissions. That creates a trust and review gap: users may authorize or run the skill without understanding that it can store credentials locally, access files, and send data to a remote API. In this context the behavior appears related to the stated image-editing workflow, but undeclared capabilities still increase the risk of unintended data exposure or misuse.

Tp4

High
Category
MCP Tool Poisoning
Confidence
87% confidence
Finding
The skill is presented as a constrained product-retouching tool, but the described commands support broader free-form image editing, task polling, image upload, and credential setup/storage. This mismatch can mislead users and security reviewers about the true operational scope, making it easier to use the tool for unauthorized or deceptive image manipulation or to expose credentials and images beyond the expected retouch-only use case. Because the skill already handles commercial product images and local API keys, the mismatch materially raises abuse potential.

Description-Behavior Mismatch

Medium
Confidence
95% confidence
Finding
The skill is described as a narrowly scoped product-retouch tool, but the CLI exposes generic task querying and raw media upload subcommands that are broader than that stated purpose. This expands the effective capability surface and can let callers use the skill as a general API client for media handling and task inspection, undermining least-privilege expectations and policy scoping.

Description-Behavior Mismatch

Medium
Confidence
96% confidence
Finding
`--param key=value` is passed through directly into the model request without an allowlist tied to the skill's promise to preserve product structure, labels, and commercial facts. That means a caller can supply arbitrary backend-supported parameters to alter behavior beyond constrained retouching, potentially turning a 'safe retouch' skill into a more general image manipulation interface.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.