ClawHub
Back to skill

Security audit

Ssh Exec

Security checks across malware telemetry and agentic risk

Overview

This is a small instruction-only SSH helper whose powerful remote-command behavior is disclosed and matches its purpose.

Install this only if you want your agent to run commands on a remote machine over SSH. Use least-privilege SSH keys and accounts, verify the target host and port, and review any command or local script before allowing execution.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

Description-Behavior Mismatch

Medium
Confidence
90% confidence
Finding
The skill is advertised as executing a single remote command, but the documentation also shows piping and executing an entire local script on the remote host. This materially expands the capability from constrained command execution to arbitrary remote code execution, increasing the chance of misuse, policy bypass, or operators running more powerful actions than intended.

Missing User Warnings

Medium
Confidence
83% confidence
Finding
The documentation normalizes remote command execution and remote script delivery without warning users about trust boundaries, target verification, privilege level, or the risk of destructive commands. In a skill specifically designed for SSH execution, omission of safety guidance makes accidental harmful use more likely, especially when examples can be copied directly into privileged environments.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.