The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses launchctl (macOS), crontab (Linux), and schtasks (Windows) to create persistent background tasks. This is a form of system persistence that allows code to run outside the immediate agent session.
[COMMAND_EXECUTION]: It performs dynamic generation of executable configuration files, such as creating .plist files in ~/Library/LaunchAgents/ using shell redirection (cat > ...).
[COMMAND_EXECUTION]: The skill modifies the system's crontab by piping existing lists into new ones with appended commands, which can lead to misconfiguration or unintended persistence if the logic is manipulated.
[COMMAND_EXECUTION]: This skill presents an attack surface for indirect prompt injection.
Ingestion points: Untrusted user descriptions for tasks (e.g., "remind me to...") are accepted in SKILL.md.
Boundary markers: None identified in the command templates.
Capability inventory: Full shell execution via system schedulers (cron, launchd, schtasks).
Sanitization: No visible sanitization or validation of the task strings before they are injected into the scheduling commands.

scheduled-tasks