The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes Node.js child_process to execute shell commands for gathering process information and performing DNS checks. Specifically, the system_network tool allows for a user-provided hostname to be used in a DNS connectivity check, which may lead to command injection if input is not sanitized.
[DATA_EXFILTRATION]: The system_network tool exposes MAC addresses, which are persistent, unique hardware identifiers that can facilitate host tracking or identification.
[DATA_EXFILTRATION]: The system_processes tool returns full command-line strings for running processes. This is a high-risk behavior as command lines often contain sensitive data like credentials, API keys, or tokens passed as environment variables or arguments.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection. Ingestion points: system process names, command lines, and network interface metadata (SKILL.md). Boundary markers: absent. Capability inventory: executes shell commands via child_process (SKILL.md). Sanitization: none specified for telemetry data. An attacker could influence agent behavior by manipulating process names or other system metadata read by the skill.

system-info