Douyin Content Research

Problem being addressed

When researching short-video topics, monitoring competitors, or observing public sentiment, the friction is usually not that content is invisible, but that public signals are scattered: keyword search results, high-engagement videos, a creator's historical posts, active comments, and live trend lists often require manual paging, copying, and cleanup. This makes it hard to reproduce a workflow or build a comparable dataset of likes, publish times, and comment text. 🚀 Douyin Content Research is aimed at that workflow, turning public Douyin data into structured logs.

How the skill works

The skill runs with Node.js and relies on an external API token (GUAIKEI_API_TOKEN) to fetch data. Its core capabilities include:
- Keyword search: search videos by keyword, then inspect results by likes or recency to identify topic directions and high-engagement patterns.
- Creator post retrieval: fetch public posts from a specific creator, useful for comparing competitor account strategies.
- Comment retrieval: read detailed comments on a video, helping identify discussion direction, user concerns, and sentiment cues.
- Trend tracking: retrieve Douyin's real-time trending list to spot active topics quickly.
- Structured export: results are saved as JSON logs, making later filtering, comparison, tabulation, and analysis easier.

In practice, you usually start from a concrete task, such as finding high-like videos for a keyword in the last 30 days, collecting recent posts from a benchmark account, reading comments on one video, or checking the current trending list. The skill requests the relevant data with the provided inputs and writes outputs to log files, reducing manual cleanup.

Boundaries and notes

This capability targets public Douyin data only; it is not suitable for private, hidden, or login-gated content. A valid token is required before calls, otherwise permission or runtime issues may occur. Outputs should be used for personal or internal team research, such as topic analysis, competitor observation, and comment synthesis; they should not be used for abusive scraping, bulk redistribution, or circumventing platform limits. If results are sparse, adjust keywords, time range, or sorting criteria.