ClawHub
Back to skill

Security audit

Ai Assistant

Security checks across malware telemetry and agentic risk

Overview

The skill is mainly a document-analysis helper, but it asks for broad command execution and includes network troubleshooting commands outside its stated document-review purpose.

Review this skill before installing if you do not want a document-analysis helper to have local command execution. If installed, use it only on documents you intend the agent to read, and require explicit approval for any shell commands, especially network, firewall, proxy, or permission troubleshooting.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Context-Inappropriate Capability

Medium
Confidence
95% confidence
Finding
The skill is for long-document reasoning and contract/risk analysis, yet it includes shell-based network diagnostics such as running `ping` and checking firewall/proxy settings. That expands behavior beyond the stated purpose and normalizes local command execution, which could be abused by prompting the agent to run unnecessary system commands in the user's environment.

Description-Behavior Mismatch

Medium
Confidence
92% confidence
Finding
The manifest describes `exec` as needed only for document-file handling, but the body later extends it to environment and network troubleshooting. This mismatch undermines informed consent and makes the skill's effective privilege scope broader than advertised, increasing the risk that users or agents permit command execution they did not expect.

Missing User Warnings

Medium
Confidence
90% confidence
Finding
The skill declares `exec` capability but does not clearly warn that it may run system commands on the local machine. In an agent setting, lack of explicit disclosure can cause users to invoke a document-analysis skill without realizing it may execute shell commands, creating avoidable local-environment risk.

Missing User Warnings

Low
Confidence
93% confidence
Finding
The troubleshooting section directly instructs use of `ping` without a clear shell-execution safety notice. Even though `ping` is usually low risk, embedding command instructions in a non-admin skill trains unsafe execution patterns and can be extended to more dangerous commands in practice.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.