ClawHub
Back to skill

Security audit

research Decision

Security checks across malware telemetry and agentic risk

Overview

This appears to be a research/troubleshooting skill with read-only local checks and web research behavior that is mostly purpose-aligned, but users should be aware it may reveal software inventory and search topics.

Install if you are comfortable with the agent running read-only environment/dependency checks and using web searches for your research topic. Avoid using it with confidential project names or sensitive incident details unless you first instruct the agent to ask before local inspection and to generalize or restrict external search queries.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (3)

Missing User Warnings

Medium
Confidence
92% confidence
Finding
The skill explicitly tells the agent to run local version, dependency, and environment inspection commands automatically before asking the user to do them. Even though these commands are commonly used for troubleshooting, they can reveal installed software, package sets, project structure, and other host-specific metadata without informed user consent. In this skill context, that creates an avoidable privacy and system-inspection risk.

Missing User Warnings

Medium
Confidence
96% confidence
Finding
This workflow section operationalizes automatic host inspection by directing the agent to run commands such as --version, npm list, pip list, and composer show and to include the results in the report. Those commands can disclose local software inventory and dependency information that may be sensitive in enterprise or regulated environments, especially when done by default and without notice. The surrounding skill purpose makes the behavior understandable, but not safe by default.

Natural-Language Policy Violations

Medium
Confidence
92% confidence
Finding
The file explicitly states that Chinese and English queries must be used and that high-quality sources must be covered, which can cause the agent to disclose user-provided topics to additional external search providers or regions without explicit user consent. In a research skill, this expands data exposure and may violate user expectations, organizational policy, or data residency/privacy requirements, especially when prompts contain confidential project names, incident details, or internal technologies.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.