mac-network-troubleshooting

Pass

Audited by Gen Agent Trust Hub on Oct 27, 2026

Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPRIVILEGE_ESCALATIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill functions by executing the kilobat CLI utility and its MCP server equivalent to perform diagnostic tasks on macOS systems.
  • [DATA_EXFILTRATION]: The instructions designate several commands as "read-only, run without asking," which permits the agent to collect and potentially exfiltrate sensitive local data without user confirmation. This includes mapping all devices on the local network (kilobat devices), identifying the network destinations for every running process (kilobat destinations), and accessing detailed historical application usage statistics (kilobat report).
  • [PRIVILEGE_ESCALATION]: The skill invokes commands that trigger macOS administrator prompts for elevated access, such as modifying system DNS settings (kilobat dns) and capturing packet headers (kilobat capture). While the instructions mandate asking the user first, the skill is designed to leverage these high-privilege operations.
  • [INDIRECT_PROMPT_INJECTION]: The skill provides an attack surface for indirect prompt injection by ingesting untrusted data from the environment (e.g., process names, SSIDs, and diagnostic strings) without sanitization.
  • Ingestion points: Data from kilobat diagnose, kilobat status, and kilobat scan.
  • Boundary markers: Absent; the agent is instructed to quote and interpret tool output directly.
  • Capability inventory: The agent has the capability to modify system settings, block apps, and monitor network traffic.
  • Sanitization: No sanitization or validation is applied to strings ingested from tool outputs before they are processed by the agent.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 27, 2026, 10:42 PM