ClawHub
Back to skill

Security audit

Autotask Mcp

Security checks across malware telemetry and agentic risk

Overview

This skill is a disclosed local Docker wrapper for an Autotask MCP server, with credential and auto-update risks that are mostly documented and purpose-aligned.

Before installing, verify that you trust the upstream Docker image publisher, manually create and edit the .env file without exposing its contents to the agent, and consider pinning a known-good image digest before enabling weekly auto-updates. Review the scheduler scripts before running them, since they create user-level persistent tasks that can pull new images and restart the local service.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (3)

Intent-Code Divergence

Medium
Confidence
97% confidence
Finding
The skill declares a strict command allowlist for agents, but later instructs execution of `./scripts/cron_install.sh` and `./scripts/cron_uninstall.sh`, which are not on that list. This contradiction weakens operator and agent safety guarantees, making it easier for an agent to justify executing broader system-modifying commands than the policy originally permits.

Vague Triggers

Medium
Confidence
84% confidence
Finding
The invocation description is broad enough that an agent could select this skill for routine Autotask-related requests, even when local Docker setup and privileged credential handling are unnecessary. Over-broad triggering increases the chance that secrets, local services, or container-management actions are introduced into otherwise simple tasks.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The skill recommends automatic updates and scheduled restart behavior via LaunchAgent or systemd user timers without a prominent warning that this creates persistent scheduled tasks and may restart services automatically. That can surprise users, alter system state beyond the current session, and in the worst case keep pulling `latest`, increasing supply-chain and persistence risk.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal