ClawHub

洞明

Security checks across malware telemetry and agentic risk

Overview

This is a strategy-analysis skill that may use web research and create a limited report/PPT outline file, but those behaviors are disclosed and fit its purpose.

Install if you want a Chinese strategy-analysis workflow that can use live web research and produce structured reports. Be aware it may perform outbound source checks, create a PPT-outline file, and optionally call separate expert or PPT skills; review those downstream skills before using slide generation or expert orchestration.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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
Findings (5)

Description-Behavior Mismatch

Medium
Confidence
89% confidence
Finding
The skill is presented as a strategy-analysis system, but it explicitly expands into live web retrieval, URL reachability checks, file writing, and downstream orchestration. This scope expansion increases the attack surface and can cause the agent to perform actions or access resources beyond what a user would reasonably expect from a strategy skill.

Context-Inappropriate Capability

Medium
Confidence
92% confidence
Finding
Live web cross-checking and HTTP reachability validation introduce network interaction capabilities that are not essential to the core strategic reasoning function. If invoked on attacker-influenced URLs or sources, this can enable unintended outbound requests, SSRF-like behavior, or validation of arbitrary endpoints.

Vague Triggers

Medium
Confidence
90% confidence
Finding
The implicit trigger condition is broad enough to activate on normal strategic discussion, which can cause the agent to enter a more powerful workflow than the user intended. Overbroad auto-triggering increases the chance of unrequested data collection, external retrieval, or file-generation behavior.

Vague Triggers

Medium
Confidence
88% confidence
Finding
The usage description lists very broad natural-language triggers such as market research, competition analysis, and strategy transformation without strong boundary conditions. This makes accidental activation likely and broadens the circumstances under which the agent may use expanded capabilities.

Natural-Language Policy Violations

Medium
Confidence
84% confidence
Finding
The skill metadata and usage guidance strongly prescribe Chinese-language interaction without any indication of user choice. Forced language behavior can undermine user intent, conceal meaning from reviewers, and increase the chance of misunderstandings in security-sensitive or policy-constrained contexts.

VirusTotal

62/62 vendors flagged this skill as clean.

View on VirusTotal