ClawHub
Back to skill

Security audit

autodl-train

Security checks across malware telemetry and agentic risk

Overview

This is a disclosed SSH helper for launching and monitoring AutoDL training jobs, with sensitive but purpose-aligned server access.

Install only if you are comfortable giving this skill SSH access to the target AutoDL server. Review config.json and any .env file before running it, especially host, username, project_path, ssh_key_path, train_command, log_path, and resume arguments. Prefer SSH keys and a least-privileged remote account where possible, keep passwords and private configs out of published skill directories, and monitor launched jobs because they can keep running and consuming GPU time, disk, and cloud credits.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (1)

Lp3

Medium
Category
MCP Least Privilege
Confidence
89% confidence
Finding
The skill directs the agent to read local files, consume environment variables, and execute shell commands over SSH, but it does not declare permissions for those capabilities. This creates a trust and policy gap: an orchestrator or reviewer may treat the skill as lower-privilege than it actually is, while the skill can access credentials such as SSH passwords and perform remote operations with significant side effects.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.