Collect Douyin Competitive Evidence with MediaCrawler

Where this fits

When you are auditing a competitor's short-video strategy, you don't want your agent to improvise from screenshots or memory. You want reproducible raw evidence: search cards for discovery, exact video details for shortlisted items, comments, downloaded media, and creator profiles. media-crawler-douyin turns Douyin research into a defined collection pipeline that an agent can plan, execute, and document.

Why agents benefit

• The agent can construct search queries across problem, category, and audience, then run Douyin search serially without mixing discovery and enrichment data.
• It can keep discovery cheap: search cards are only for shortlisting, while comment and media scraping are reserved for the final shortlist.
• It supports exact re-collection by passing DY_SPECIFIED_ID_LIST or DY_CREATOR_ID_LIST, so runs are reproducible and auditable.
• It combines search, detail, creator, comments, and media steps into one workflow, giving the agent structured inputs for trend, hook, format, and audience-language analysis.
• PUBLISH_TIME_TYPE lets the agent scope collection to a date window upstream, instead of post-filtering noisy results.

Practical scenarios

Trend and hook research for a new campaign

The agent defines keywords around a problem, category, and audience, runs search serially, and shortlists by topical relevance, hook/format diversity, recency, and engagement. It then collects exact video details and comments for the shortlist to find recurring opening patterns.

Competitor creator audit

Set DY_CREATOR_ID_LIST to profile URLs or sec_user_id values and run --type creator. The agent collects creator profiles and video metadata, then compares posting cadence, format mix, and comment language across competitors.

Reproducible evidence for a content brief

Before writing a brief, the agent documents purpose, keywords, sample size, date window, and destination. It saves raw search cards and the shortlist method, then runs detail and comment collection. The result is a citable evidence folder, not a summary.

Add it to your agent workflow

Install the skill:

npx skills add tsingyuai/growth-lab --skill media-crawler-douyin

Before collection, define the task constraints in the invoking model's memory. For discovery, run:

python main.py --platform dy --type search

For exact enrichment, put canonical video URLs or IDs in DY_SPECIFIED_ID_LIST, or profile URLs in DY_CREATOR_ID_LIST, then run --type detail or --type creator. Verify canonical video ID, author, and caption in the output.

Compare before adopting

• Quality signals: Does it preserve canonical IDs, author, caption, and exact URLs? Can you re-run with the same config to verify output?
• Maintenance freshness: Douyin's client can change; check repo commit activity and issue handling before relying on it in production.
• Alternatives: Compare with paid API providers, one-off scraping scripts, or a general research skill; consider rate limits, structured output, and total cost.
• Workflow fit: Does it fit your agent's memory and storage conventions? Are purpose, sample size, date window, comments/media scope, and destination part of your regular task template?

Why it is worth tracking

This skill has solid community momentum on GitHub (1,985 stars) and enforces a research discipline that is useful for agent workflows: it requires a purpose, a selection method, and a destination before collecting evidence. The underlying MediaCrawler is widely used, which lowers the risk of a dead-end skill. Evaluate it as soon as your agent needs documented, reproducible Douyin evidence rather than ad-hoc searching. For ongoing comparison, track its listing at https://www.openagentskill.com/skills/tsingyuai-media-crawler-douyin.