What problem does it solve? Running an AI coding agent periodically without manual launching is tricky: most agents are one-shot and amnesiac, cron cannot go below one minute, and unattended runs silently hang on permission prompts. This Skill provides a decision framework and concrete commands for scheduling agents reliably. ## Core Features & Use Cases - Two-camp decision framework: Choose between wrapping one-shot agents (Claude Code, Codex, Pi) in an external clock versus using Hermes with its built-in scheduler daemon. - External clock patterns: Cron entries, systemd timers, and while-sleep loops for sub-minute intervals, with JSON output and tool allowlists for unattended execution. - Hermes recurring jobs: Create cron-expression, interval, or one-shot jobs with zero-token script mode, job chaining, self-terminating loops, and loop-safety guards. - Use Case: You want an agent to check a task list every 5 minutes and only act on due items. Use the heartbeat pattern with a recurring Hermes job or a while-sleep loop that stays silent when nothing is due. ## Quick Start Ask the agent to set up a recurring job that runs a specific prompt every 10 minutes and verify it fires by checking the log or job list.