Make it work once. Then put it on a schedule.

Choose a result that stays useful

Good candidates have a predictable input and a small, reviewable output: summarize changes in a project note, prepare a weekly planning brief, or collect open questions from documents you provide. Avoid making your first routine depend on a long chain of logins and uncertain source material.

Run the task manually first. Check the result, refine the prompt and repeat the test with a slightly different input. A schedule should save you from remembering to start useful work, not from checking what it produces.

Describe the time and the stopping point

Every weekday at 9 in my local time, review the shared project notes and list three priorities, any blocked work and the source for each item. Put the brief in this conversation. Do not send messages or change project records.

This example distinguishes preparing a brief from taking action on it. If a source is unavailable, ask for a clear report of what was missing rather than a guessed update.

Remember that this team runs on your Mac

YankyBot’s local schedules require the app to be running and the Mac awake. Closing the laptop is not the same as leaving a cloud worker running. Check the schedule’s next run and your computer’s availability before relying on it for a time-sensitive job.

A machine waking late can change when work happens. Do not use a local reminder as the only control for something that must execute at an exact time. The local and cloud comparison explains this trade-off.

Leave a reviewable trail

Ask each run to include the period it reviewed and the sources it actually read. A note saying “no changes found in these three files” is more useful than a generic all-clear that gives you nothing to inspect.

When the agent is already working, additional messages use its queue. Avoid creating several overlapping schedules for the same job. Give each routine a clear owner and pause it when the source or responsibility changes.

Retire routines as deliberately as you create them

A task that was useful during a launch can become noise afterward. Review schedules when a project ends. Pause the old routine, save any useful output and decide whether a new one is needed.

The best automation has a reason to exist today. A short list of dependable routines will usually be easier to supervise than a busy list of forgotten ones.