ClawHub
Back to skill

Security audit

Doubao Assistant

Security checks across malware telemetry and agentic risk

Overview

This skill appears to be a broad Doubao integration guide, but it requests powerful agent tools and gives under-scoped examples for tool execution and data access.

Review before installing. Use this only in a workspace where read and exec access are acceptable, keep Doubao and vector database credentials in environment variables or a secrets manager, and require explicit user approval plus allowlisted tools before any create, modify, delete, ticketing, file-writing, or command execution action.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

Intent-Code Divergence

Medium
Confidence
94% confidence
Finding
The skill advertises only generic inputs like content/mode/style, but the body describes much broader capabilities including command execution, file operations, tool calling, RAG, and ticket creation. This mismatch can mislead users and agent platforms about the true privilege and data-handling scope, increasing the chance of unintended execution or approval of risky actions.

Intent-Code Divergence

High
Confidence
99% confidence
Finding
The document states sensitive operations require user confirmation, but the example tool-calling flow executes model-requested tools directly via executeTool(...) with no approval gate. In an agent context with exec/read/glob/grep tools available, this creates a direct path for prompt-influenced or malicious model outputs to trigger external actions without human consent.

VirusTotal

61/61 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

Detected: suspicious.prompt_injection_instructions

Prompt-injection style instruction pattern detected.

Warn
Code
suspicious.prompt_injection_instructions
Location
SKILL.md:189