Weekly customer feedback digest from tl;dv calls

Every Friday at 4pm in my local timezone, run an agent that synthesizes the past week of customer calls from tl;dv into a single product feedback digest. Use a cron trigger on a weekly Friday 4pm schedule.

Step 1: Call tl;dv List Meetings to pull every meeting from the last seven days. Filter out internal meetings using the title and template, so anything that looks like a standup, retro, planning, all-hands, 1:1, or interview is skipped. Keep only customer-facing calls (discovery, demo, onboarding, success check-in, renewal, churn, win-back, support escalation, etc.).

Step 2: For each remaining meeting, call tl;dv Get Meeting Notes to retrieve the structured AI notes (topics plus markdown summary). Capture the meeting title, date, attendees or customer name, the meeting link, and the notes.

Step 3: Cluster the notes across all meetings into recurring themes. Default buckets are: feature requests, complaints and bugs, churn risks, competitor mentions, and pricing pushback. Allow new themes to emerge if multiple customers raised the same topic. For each theme, pick a representative verbatim quote from the notes, list the source meetings (with links), and count the number of distinct customers (not raw mentions) who raised it. Rank themes by distinct-customer count, breaking ties by severity (churn risk and pricing pushback ranked higher).

Step 4: Send the ranked digest to Slack. Use Slack Send a Message to post into the #product-insights channel. The message should include the week range at the top, a ranked list of themes with the count of distinct customers, the representative quote, and links to the source meetings. Use Slack formatting (bold headings, bullet lists). Keep it scannable; if the message is long, post the headline summary in the channel and put the per-theme detail in a thread reply.

Step 5: Append one row per theme to a running Google Sheets tracker so trends accumulate. Use Google Sheets Append Values against a sheet I will configure. Columns: week (ISO week or Friday's date), theme, distinct customer count, severity (low / medium / high), and source meeting links (comma-separated). One row per theme, in the same ranked order as the Slack post.

Other notes: if there are no customer calls in the window, post a short Slack message saying so and skip the spreadsheet write. Make the customer-meeting filter, the Slack channel, the spreadsheet ID and tab name, and the theme buckets easy to edit at the top of the agent instructions so I can tune them later without rewriting the workflow.