home-network

Pass

Audited by Gen Agent Trust Hub on Nov 20, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from the local network environment that could be manipulated by an attacker on the same network.
  • Ingestion points: The skill reads device names, hostnames, and SSIDs via commands like arp -a, dns-sd, and system_profiler SPAirPortDataType (documented in diagnostics.md).
  • Boundary markers: The instructions do not specify the use of delimiters or 'ignore' instructions when the agent processes these ingested strings.
  • Capability inventory: The skill utilizes shell command execution for diagnostics and writes log data to the file system (e.g., ~/ITGuy/machine.md).
  • Sanitization: While the skill provides guidelines for redacting sensitive information like MAC addresses for privacy, it lacks explicit logic to sanitize or escape ingested strings to prevent them from being interpreted as commands by the AI.
  • [COMMAND_EXECUTION]: The skill includes numerous recipes for executing built-in macOS networking utilities.
  • Evidence: Commands such as networkQuality, scutil, ping, and ipconfig are utilized for diagnostics. These are legitimate tools used within the scoped purpose of network administration, and the skill includes warnings against privilege escalation (e.g., noting that wdutil requires sudo and should be handled by the user).
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 20, 2026, 12:15 AM