The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill is designed to generate and execute arbitrary shell scripts (Bash and PowerShell) locally by writing them to the filesystem and then registering them for execution via platform-native schedulers.
[COMMAND_EXECUTION]: The implementation utilizes high-privilege system utilities for persistence, including Register-ScheduledTask on Windows, systemd-run --user on Linux, and launchctl bootstrap on macOS.
[PROMPT_INJECTION]: The instructions explicitly guide the agent to bypass security constraints and permission models when scheduling other AI agent sessions. This includes using flags such as --permission-mode bypassPermissions and --dangerously-skip-permissions for Claude Code, --allow-all for GitHub Copilot, and --sandbox danger-full-access for OpenAI Codex.
[COMMAND_EXECUTION]: The Windows implementation specifically employs -ExecutionPolicy Bypass when executing scheduled PowerShell scripts, which overrides local security settings regarding script signing.
[COMMAND_EXECUTION]: The Linux reference suggests using loginctl enable-linger, which enables persistence across user sessions and may require elevated privileges via sudo or polkit to execute.

schedule