Douyin Data Brief

Problem

When you have a Douyin link, a keyword, or pasted copy with counts and comments, the hard part is turning scattered evidence into a traceable read: why the post performed, whether the opening hook works, what the audience actually said, and whether it is worth following. This skill targets that kind of post-level analysis instead of publishing workflows, full account audits, or batch content generation.

How it works

The required input is simply where the Douyin content comes from: a public video link, a keyword, or user-pasted material. By default it treats one video as the unit, reads one page of first-level comments, and outputs four sections: metrics (views, likes, comments when available), hook, one audience quote, and follow or skip. User-provided numbers are labeled as provided; queried numbers are labeled as fetched with a read date. Optional paid queries can cover a video, comments, keyword search, creator page, or hot-list lookups, and each query must be confirmed before execution, including scope, price, and count. Execution goes through the bundled client, with tool parameters passed as JSON over stdin and a stable client_request_id per paid request to avoid duplicate submission and duplicate billing.

Boundaries

It does not estimate missing data or interpolate from similar videos. Inferences must be marked as inferences and supported by evidence. If the user already has enough material, a free brief can be produced without querying. New queries, extra pages, changed parameters, or reruns are treated as new paid work and require fresh approval. The output is a one-page sourced brief, not a full account diagnosis, a publish-ready hot-topic post, or generated media.