Weekly Fillout feedback digest with Linear issues and a Slack summary

Build an agent workflow that triages our Fillout feedback form on a weekly schedule and turns it into a Slack digest plus Linear issues for the things worth acting on.

Trigger: cron, every Monday at 9am in the org's timezone.

Inputs the user should configure when launching the workflow:

- Fillout form ID (the feedback form to read from).

- Linear team ID where new issues should land, plus optional default labels and priority.

- Slack channel ID where the digest should be posted.

- Threshold for filing a Linear issue. Default: 'mentioned by more than one respondent, or explicitly flagged as blocking.'

Steps the agent should follow each run:

1. Call Fillout List Form Submissions on the configured form, filtered to the last 7 days using the afterDate and beforeDate query params. Paginate until all submissions in the window have been fetched.

2. If there are zero submissions, skip step 3 and 4 and just post a short Slack note like 'No new feedback submitted this week.' Then stop.

3. Read through every response and cluster them into themes: bugs, feature requests, praise, and confusion. For each cluster capture (a) a count, (b) two or three representative verbatim quotes with the submitter's email, and (c) a one-line summary of the theme.

4. For any bug or feature request that meets the configured threshold, call Linear Create Issue in the configured team. Title should be a short, action-oriented summary of the theme. Description should include the theme summary, the submitter emails, and the quoted feedback so engineering can follow up. Apply the configured labels and priority if provided. Collect the issue URLs to include in the digest.

5. Also compare this week's counts per theme against last week's counts so the digest can call out sentiment trend (up, down, flat). If there is no prior-week baseline yet, say so instead of guessing.

6. Call Slack Send a Message to post the digest to the configured channel. The message should include: total submissions this week, counts per theme, the top three quotes across the whole week, the sentiment trend versus last week, and a bulleted list of any Linear issues opened with their URLs.

Keep quotes short and unedited. Do not paraphrase what the customer said. If a response is off-topic or empty, drop it from the analysis instead of stuffing it into a theme.

This is agent work because the clustering, threshold judgement, and quote selection are open-ended.