ClawHub
Back to skill

Security audit

Agent Handshake

Security checks across malware telemetry and agentic risk

Overview

This skill is documented as an agent connection helper but also instructs users to deploy a persistent HTTP remote-command service with broad file and command access.

Review carefully before installing. Treat the VPS deployment mode as granting remote administrative access to the target machine, not just agent discovery. Do not expose it publicly or through a tunnel unless you understand the risk, use strong unique credentials, restrict network access, and are comfortable with a persistent service that can run arbitrary commands.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (3)

Description-Behavior Mismatch

High
Confidence
97% confidence
Finding
The documented behavior goes far beyond a simple 'handshake' or service discovery skill: it explicitly describes deploying `agent_shell.py`, exposing `/run`, and using it as a substitute for SSH to execute arbitrary remote tasks. This is dangerous because the skill framing can mislead users into enabling a remote administration backdoor they may not recognize as full remote code execution.

Context-Inappropriate Capability

High
Confidence
99% confidence
Finding
The skill explicitly enables arbitrary remote system administration over an HTTP endpoint protected only by HTTP Basic Auth, including reading files, writing files, and running any command. That creates a high-risk remote code execution surface that is weaker and less auditable than standard hardened SSH deployments, especially if exposed to the internet or tunneled publicly.

Missing User Warnings

High
Confidence
95% confidence
Finding
The documentation normalizes broad remote command execution and file access ('跟 SSH 一样', '可以读文件、写文件、跑任何命令') without a prominent safety warning, threat model, or operational constraints. In context, this makes dangerous capability more likely to be adopted casually, especially because the skill also promotes public exposure and tunneling workflows.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.