What this agent does
Individual pieces of customer feedback are easy to read and easy to forget — the pattern only becomes obvious once someone tags fifty of them and looks at the aggregate, and that tagging rarely happens consistently by hand. This agent tags every incoming piece of feedback the moment it arrives and rolls it into a weekly "what's actually recurring" report.
stage-1-auto-tagging— reads each new support ticket, review, or feedback form submission and tags it with a theme (bug, feature request, pricing complaint, praise, churn-risk) and a sentiment score.stage-2-weekly-rollup— aggregates the week's tagged feedback, ranks themes by frequency and severity, and writes a short report naming the top 3-5 issues worth acting on, each with representative quotes.
Every run appends to a running feedback log, so themes can be tracked as trends over multiple weeks, not just a single snapshot.
Requirements
- A feedback source: a support-ticket system export, a review-platform feed, or a feedback-form spreadsheet.
- Works on Zo Computer natively and on Claude AI (claude.ai or Claude Code) — no dependency on a specific helpdesk vendor; the package includes adapters for CSV/Sheet input so it works with whatever export your tool provides.
Setup
- Unzip the package into your workspace, keeping the folder structure intact.
- Fill in your source location (Sheet ID, CSV export path, or feed URL) and the theme categories relevant to your product.
- Run the starter prompt against a sample batch of feedback to confirm the tagging matches your expectations.
- Schedule it weekly to generate the recurring-themes report.
What you get
- Every incoming ticket, review, or form response automatically tagged by theme and sentiment as it arrives
- A weekly report naming the top 3-5 recurring issues worth acting on, each backed by representative quotes
- A running feedback log so themes can be tracked as trends across weeks, not just a single snapshot
- Setup that works out of the box on both Zo Computer and Claude AI, with adapters for whatever export format your feedback tool provides
