Daily Adobe Commerce sales pulse posted to Slack before standup

Every weekday at 8am in the workspace timezone, generate a plain-English sales pulse from Adobe Commerce and post it as a single Slack message before the team's standup.

Pull yesterday's orders from Adobe Commerce using the List Orders operation. Filter with searchCriteria where status is in ["processing", "complete"] and created_at is between the start and end of yesterday in the store's local timezone. Also pull the previous 7 days of orders using the same status filter and a created_at range from 8 days ago through yesterday. This second pull is the rolling baseline used for comparisons; do not include today. If order counts are large, page through results using searchCriteria[pageSize] and searchCriteria[currentPage] until total_count is reached.

From yesterday's orders, compute: gross revenue (sum of grand_total), order count, average order value, top 3 SKUs by units sold (sum item qty_ordered grouped by SKU), and top 3 SKUs by revenue (sum item row_total grouped by SKU). Read the SKU and quantities from each order's items array. Also count any orders that landed in canceled or closed status yesterday and surface that as a refund and cancel signal.

From the trailing 7-day window, compute the same daily averages (revenue per day, orders per day, average order value) so yesterday can be compared to normal. Compute percentage deltas versus the 7-day average for revenue, order count, and AOV.

Have the model write one tight Slack-flavored digest, five to eight short lines total, that leads with yesterday's headline numbers, shows the day-over-week deltas in parentheses, lists the top SKUs by units and by revenue, notes any refund or cancel activity, and includes exactly one call-out on anything unusual (a revenue spike, a slump, an outsized single order, or a SKU jumping into the top list for the first time this week). If nothing is genuinely unusual, skip the call-out entirely; do not manufacture drama. Use Slack mrkdwn (single asterisks for bold, backticks for numbers if useful) and simple line breaks, not Block Kit.

Post the finished digest as a single Slack message using the Slack Bot Send a Message operation. The user picks the destination channel or DM when they set the workflow up. Only one message per run; do not thread, do not split.