I asked Claude for a prompt others could use to implement this: I want to schedule [my sync script] to run automatically instead of triggering it manually. It's a Python script (or similar) that pulls data from an API using locally-saved credentials/tokens — not something that can easily run in GitHub Actions or another CI runner, since those start with a clean environment and can't complete auth (password/MFA/token refresh) unattended. Set this up as a local scheduled task (Windows Task Scheduler / cron, depending on OS) instead: 1. Write a small wrapper script that runs the actual sync script and logs its output (stdout+stderr) with a timestamp, to a log file I can check later. Watch for encoding mismatches if mixing native command output with the shell's own log-writing (e.g. PowerShell's Add-Content vs pipe redirection can default to different encodings and produce garbled log text — force a consistent encoding for both). 2. Register the scheduled task to run daily at [time]. If the machine can sleep, explicitly check whether wake timers are enabled and whether the task is set to wake the computer — don't assume it will just work. 3. Actually verify: trigger the task manually first to confirm the wrapper itself works, then leave the real schedule running for several real days and check the log / task history afterward — don't just trust that registering the task means it will fire reliably. A machine that sleeps can silently miss scheduled runs even with wake timers enabled, if it falls back asleep in the gap between waking and the task actually launching. If you see missed days, check the OS's power/sleep event log around the scheduled time to see whether it woke at all, and consider just disabling sleep entirely if this machine is meant to run this kind of automation (most desktops draw negligible extra power idling vs. sleeping). 4. Leave the resulting data-file changes for me to review/commit myself, don't auto-commit unless I ask for that separately.