ClawHub
Back to skill

Security audit

沙箱智能体对话 Chat

Security checks across malware telemetry and agentic risk

Overview

This skill is mostly transparent about using a hosted dLazy chat service, but its very broad trigger phrases could cause ordinary chat requests to be sent to a third-party service unexpectedly.

Install only if you intend ordinary dLazy chat requests to be handled by this third-party hosted service. Avoid sending sensitive prompts or attachments unless they are appropriate for dLazy, and prefer explicit product-specific invocations such as dlazy chat or dLazy sandbox agent.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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 (3)

Intent-Code Divergence

Medium
Confidence
95% confidence
Finding
The documentation states the skill itself '不会越权访问网络或文件系统', yet nearby text explicitly says prompts are sent to api.dlazy.com and local files passed via --files are uploaded to files.dlazy.com. This mismatch can mislead users about data exposure and trust boundaries, causing them to provide sensitive prompts or files under a false assumption of local-only behavior.

Vague Triggers

Medium
Confidence
86% confidence
Finding
The trigger phrases are broad conversational terms such as '对话' and '和 agent 聊', which are likely to appear in normal user requests unrelated to this specific external SaaS skill. That increases the chance of unintended invocation, potentially causing accidental network calls, project creation, or upload of user context to a third-party service without clear user intent.

Vague Triggers

Medium
Confidence
92% confidence
Finding
The trigger keywords include very generic phrases like 'chat' and 'talk to the agent', which can cause unintended invocation in ordinary conversation. In this skill's context, accidental activation is more concerning because it is a thin client to a hosted SaaS agent that can continue project-scoped sessions and transmit prompts, project context, and optionally attached files to remote endpoints.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.