Weekly social listening to Linear research issues

Build me an agent workflow that mines Brandwatch for emerging product feedback themes and files them as Linear research issues for our product team. It should run on a cron, every Monday at 9am in our local timezone.

Each run, the agent should:

1. Use Brandwatch List Queries to find our saved product-feedback query (a boolean for our product name plus complaint and feature-request language). Match by name; if multiple match, prefer the most recently updated. If none match, exit cleanly with a note in the run log.

2. Use Brandwatch Get Topics over the trailing 14 days against that query to surface trending and bursting topics, including their sentiment and volume. Pull both topic labels and the underlying volume/burst signal.

3. Filter the topic list down to real product signal. Keep only themes that read as recurring product pain points or concrete feature requests. Drop generic brand mentions, off-topic chatter, news cycle noise, competitor talk, and themes that are clearly already-resolved issues. Use judgement on the topic label, sentiment skew, and the example mentions. If you are unsure, lean toward dropping it.

4. For each surviving topic, use Brandwatch Retrieve Mentions filtered to that topic over the same 14-day window to pull 3 to 5 representative example posts. Prefer posts that are clearly on-topic, have author info, and link back to the source.

5. Before filing, use Linear List Issues against our Product team and check for any issue created in the last 30 days whose title contains the same theme (case-insensitive contains match on the topic phrase). If one exists, skip this topic; do not duplicate.

6. For each remaining topic, use Linear Create Issue in the Product team with:

- Title: [Social signal] {topic}

- Description (Markdown): a short summary of the theme; the mention volume and how it has trended over the 14-day window (e.g. bursting, steady rise); overall sentiment skew; then a quoted list of the example posts, each with the author handle and a link to the original post.

- Label: social-feedback (create the label or pick the closest existing one if it does not exist yet).

At the end of the run, log a short summary: how many topics were considered, how many passed the filter, how many were skipped as duplicates, and how many issues were created. If Brandwatch returns a rate-limit error, back off and retry; if Linear errors on a single issue, log it and move on so one bad topic doesn't kill the run.