A workflow is just three things stitched together: a document, a cron, and a prompt. Below are six realistic shapes that show how those pieces fit. Steal them, adapt them, or use them as a starting point when you sit down to wire up your own.

1. Weekday morning standup digest

Scenario. You keep a "Team Standup" document with sections for Yesterday, Today, and Blockers. Every weekday morning you want it pre-populated from the previous day's PR activity, Linear updates, and Slack threads, so the actual standup is about discussion, not data entry.

Look at yesterday's merged PRs, closed Linear issues, and any Slack threads in #eng tagged with :standup:. Rewrite the Yesterday section as a tight bulleted summary grouped by person. Carry over any unfinished items from yesterday's Today into today's Today. Leave Blockers empty unless something is clearly on fire. Do not add commentary — just the bullets.

2. Hourly status-page refresh

Scenario. A customer-facing "System Status" document pulls live data from a handful of request pages (uptime, error rate, queue depth). You want the prose summary at the top to actually match what the queries say, refreshed every hour during business hours.

You are this Status document. Re-run every request page on the page, then rewrite the Current Status paragraph at the top to reflect what the data now shows. If every metric is green, say so in one sentence. If anything is degraded or red, name the specific service and the metric that tripped, and link to the relevant page. Do not touch the historical incidents table.

3. Friday afternoon weekly retro

Scenario. You want a draft retro waiting for you at end-of-week so the team starts Monday's retro meeting with something to react to instead of a blank page.

This is a weekly retrospective document. Create a new child page titled Week of {{Monday's date}}. Populate it with three sections: What went well, What didn't, and What to try next week. Draw from this week's merged PRs, closed incidents, and any pages in this document edited in the last 7 days. Be specific — name PRs, people, and incidents. Three to five bullets per section is plenty.

4. Every-10-minutes deploy watcher

Scenario. You're in the middle of a staged rollout and want a "Deploy Watch" document that polls the deploy dashboard, summarizes what changed, and pages you only if something looks wrong.

This is a deploy watcher document. Re-run the deployStatus and errorRate request pages. Append one row to the Run log table with the current timestamp, deploy stage, and error rate. If the error rate has risen by more than 25% over the last three samples, also write a one-line ALERT at the top of the page with the timestamp and the offending service. Otherwise leave the top of the page alone.

5. Nightly stale-draft sweeper

Scenario. Your workspace accumulates half-written draft pages. You'd like an agent to quietly tidy up overnight — archive anything that hasn't been touched in 30 days, and leave a short note about what it did.

This is a housekeeping document. List every page in the workspace whose updated timestamp is older than 30 days and whose title starts with Draft:. For each one, move it under the Archive/ parent page and prepend [archived YYYY-MM-DD] to its title. Then append a dated entry to the Sweep log at the bottom of this page summarizing how many pages were archived and their titles. If nothing qualifies, append a one-line "no-op" entry.

6. Monthly billing reconciliation

Scenario. On the 1st of each month, you want a reconciliation page that pulls last month's Stripe charges, compares them to your internal usage table, and flags any mismatches before finance asks about them.

This is a Billing Reconciliation document. Create a new child page titled {{previous month name}} reconciliation. Run the stripeCharges and internalUsage request pages for the previous calendar month. Build a table joining the two on customer ID, with columns: customer, Stripe total, internal total, delta. Sort by absolute delta, descending. At the top, write a one-paragraph summary calling out any customer with a delta greater than $50, and link to the relevant Stripe and internal records.

Patterns to notice

A few things worth highlighting across the six examples above: