A scheduled Penny skill keeps a recurring method and its due date together. Use it when you want to repeat a review without writing the instructions again. Penny tracks when the skill is due; a connected AI session has to carry out the work.
Worked example: a Monday reading review
Suppose you record pages read in a tracker and want to review your progress each Monday. The review should inspect the previous seven days, explain missing entries, and suggest one adjustment. It should leave your reading goal unchanged unless you approve a change.
Start in a private conversation with Penny connected and enabled. Replace the tracker name and timezone with your own. Copy this prompt:
Prepare a Penny skill called Weekly reading review.
Use it to review my Reading pages tracker for the previous seven days.
Read the entries, report the total pages and days with recorded entries,
and identify any dates with no observation. Don't treat missing entries as zero.
Suggest one adjustment and ask before changing my goal or tasks.
Make it due every Monday at 09:00 in Europe/London.
Keep its permission scope to reading my Penny data and saving a private review note.
Don't send messages or take actions outside Penny.
Show me the instructions, cadence, timezone and permission scope before saving.
Read the proposed definition and approve it if it matches your intent. If the AI cannot create a schedule with its available Penny tools, keep the instructions as an on-demand skill and run it manually. A written promise to schedule it is not a saved schedule.
Check the saved skill
Open Skills and select the skill. Check its name, instructions, Active or Paused state, and the How this skill is offered section. Confirm the next due date, timezone, and permission scope. A Monday date in the wrong timezone can shift when the review is offered.
To try the method in a foreground conversation, copy:
Read my saved Weekly reading review skill and show its current status.
With my approval, run it now using the recorded tracker entries.
Save the resulting review privately in Penny and read it back.
Record the run's actual outcome, and report any step that failed.
Open Notes and find the review. Compare its time window and total with the underlying tracker entries. Ask the AI to read the skill's run record, where its available tools support that view. Check the next due date after a completed scheduled run. A due date, draft report, or started run alone does not show that the review finished.
Choose how it will run
Opening an eligible connected conversation and approving the due review is the manual route. Unattended execution needs a separate recurring session in a host that permits scheduled Penny access. Check that host's installation instructions; several coding integrations restrict private memory to foreground conversations. A Penny schedule does not override those restrictions or create a notification service.
See Save reusable instructions as a skill for the underlying feature and Run a weekly review with Penny for a broader review.