ClawHub
Back to skill

Security audit

沟通偏好工具箱(专业版)

Security checks for vulnerabilities and agentic risk

Overview

This skill is a documented communication-preference manager with purpose-aligned sync, history, and scene-switching features, but users should treat remote sync and automatic behavior changes carefully.

Before installing, review whether you want preference files synced to Git or object storage, keep credentials in the recommended secure locations, and enable automatic scene switching only if you are comfortable with the agent changing communication style from inferred context.

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • 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
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (3)

Intent-Code Divergence

Medium
Confidence
93% confidence
Finding
The skill claims command execution is protected by a whitelist and avoids user-input concatenation, but the surrounding documentation explicitly encourages use of user-supplied remotes, file paths, webhook URLs, and environment-derived values in exec-driven workflows without defining any actual validation or allowlisting. In a skill with read/exec/write capabilities, this mismatch can lead operators or downstream implementations to trust unsafe command patterns, increasing the risk of command injection, data exfiltration, or unintended outbound connections.

Missing User Warnings

Medium
Confidence
83% confidence
Finding
The skill promotes automatic scene detection and switching, which can silently alter the agent's communication behavior based on inferred context. Without strong warnings, confirmation gates, or visibility into when switching occurs, users may experience unintended behavioral changes that affect decisions, outputs, or policy-sensitive interactions.

Missing User Warnings

Medium
Confidence
94% confidence
Finding
The skill documents remote push/pull and merge workflows for preference data but does not clearly warn that this transmits potentially sensitive user/team preference information to external Git or storage systems. In an enterprise/team context, such preference files may contain behavioral metadata, role information, or other internal details that users may not expect to leave the local environment.

Static analysis

No suspicious patterns detected.