skills/ljagiello/agent-skills/utmapp/Gen Agent Trust Hub

utmapp

Pass

Audited by Gen Agent Trust Hub on May 15, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill enables full automation of UTM virtual machines on macOS through the bundled utmctl command-line utility and AppleScript/JXA interfaces. This allows for tasks such as starting/stopping VMs, executing commands within the guest environment, and managing file transfers. Evidence is found in SKILL.md, references/utmctl.md, and references/applescript.md.
  • [COMMAND_EXECUTION]: The documentation includes a setup instruction to create a symbolic link for the utmctl binary using sudo. This is a standard procedure to make the tool accessible from the shell's global path as shown in SKILL.md.
  • [EXTERNAL_DOWNLOADS]: The instructions guide the user to download OS installation media and VM gallery images from the official UTM website and recognized GitHub repositories. These sources include mac.getutm.app and github.com/TuringSoftware/CrystalFetch.
  • [COMMAND_EXECUTION]: The skill's ability to execute commands in a guest VM and capture the output represents a surface for indirect prompt injection. 1. Ingestion points: Output from utmctl exec and content from utmctl file pull (specified in SKILL.md and references/utmctl.md). 2. Boundary markers: Boundary markers are absent from provided command examples. 3. Capability inventory: Commands for VM management, guest script execution, and host/guest file transfers across utmctl and AppleScript. 4. Sanitization: No explicit sanitization of guest-provided output is performed in the examples.
Audit Metadata
Risk Level
SAFE
Analyzed
May 15, 2026, 06:11 AM
Security Audit — agent-trust-hub — utmapp