Enforce safety constraints on system commands before execution. Use this skill whenever the agent needs to run shell commands, terminal operations, or system-level actions. It classifies commands into BLOCKED, CONFIRM, or ALLOWED and prevents dangerous operations from executing.