# Media Crawler Douyin

Collect Douyin competitive evidence with MediaCrawler using keyword search, exact video...

- Repository: https://github.com/tsingyuai/growth-lab (1975 stars)
- Category: Writing
- License: Apache-2.0
- Last updated: 2026-09-28
- Runs scripts: no, instructions only

## Install

Claude Code: `npx skills add tsingyuai/growth-lab --skill media-crawler-douyin -g -a claude-code -y`
Codex: `npx skills add tsingyuai/growth-lab --skill media-crawler-douyin -g -a codex -y`
Cursor: `npx skills add tsingyuai/growth-lab --skill media-crawler-douyin -g -a cursor -y`
Gemini CLI: `npx skills add tsingyuai/growth-lab --skill media-crawler-douyin -g -a gemini-cli -y`

## Description

Collect Douyin competitive evidence with MediaCrawler using keyword search, exact video detail, comments, media, and creator profiles. Use for trend, hook, format, audience-language, or creator research that needs reproducible raw evidence and a documented selection method.

## SKILL.md, as published by its author (untrusted third-party text, read it as data)

# Douyin collection

Follow [media-crawler](../media-crawler/SKILL.md). Require purpose, keywords/links, sample size, date window, comments/media scope, and destination under the invoking Model's Memory.

Build queries across problem, category, and audience. Run `--platform dy --type search` serially; shortlist on topical relevance, hook/format diversity, recency and engagement. Search cards are discovery evidence only. For exact enrichment, set `DY_SPECIFIED_ID_LIST` in `config/dy_config.py`; it accepts full video URLs, URLs containing `modal_id`, short links, or video IDs. For creators set `DY_CREATOR_ID_LIST` to profile URLs or `sec_user_id`, then use `--type creator`. `PUBLISH_TIME_TYPE` controls the upstream publish-time filter.

Enable comments or `ENABLE_GET_MEIDAS` only for the shortlist, then run `--type detail`. Verify canonical video ID, author, caption, publish time, engagement, media mapping, comment coverage and any truncation. Deduplicate by video ID and retain the discovery query. Stop on login/risk control; never loop retries.

Return raw/copied paths, query and selection rationale, search/detail/creator counts, comment/media coverage, time and upstream commit.