The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/check.sh script utilizes various system-level tools including ps, df, stat, ss, netstat, and powershell.exe to gather system metrics and configuration details. These actions are consistent with the skill's stated goal of providing a comprehensive health scan.\n- [PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection.\n
Ingestion points: The agent context receives raw output from commands that list user-influenced strings, such as process names via ps or directory listings via ls.\n
Boundary markers: The script generates a formatted text report but lacks clear instruction-ignoring delimiters around gathered data.\n
Capability inventory: The skill possesses the ability to execute shell commands and read system-wide configuration metadata.\n
Sanitization: No sanitization or filtering of command output is performed to prevent embedded instructions in system strings from being interpreted as commands by the AI.\n- [SAFE]: The script accesses sensitive file paths like ~/.ssh/id_rsa and ~/.aws/credentials solely to check file permissions using the stat utility. No reading of file contents or network-based data exfiltration was detected.

system-health-check