Viral Short Video Maker

Problem

Short-video content analysis often breaks down in three places: a single video lacks structured post-mortem beyond the title and comments; account pages scatter performance signals, making it hard to infer topic selection, opening hooks, tone, and positioning; teams that want to reuse an author’s style or a corpus need a durable constraint instead of one-off prompts.

How it works

Single-video report: Given a Douyin, WeChat Channels, or Xiaohongshu link, the skill auto-detects the platform, collects video metadata, downloads the video, transcribes audio, and generates an asset JSON plus *.workbuddy_prompt.md. The model then writes a shootable script based on the original entities, concepts, and themes, and the final output is one WORK_ID_final_report.html file.

Account analysis: Given a Douyin or Xiaohongshu profile, the built-in collector gathers public posts and engagement data, then summarizes the shared patterns of high-like, high-comment, and high-save items. The output covers account positioning, target audience, title hooks, script structure, expression habits, and reusable topic directions. Xiaohongshu collection offers 20, 50, and 100-post modes with a hard limit of 200 posts.

Style persistence: Create a reusable style profile from an external account, a .docx/.txt corpus, or both. The profile records content positioning, tone, opening patterns, structure, and closing style, so later report generation can refer to a stable style instead of re-explaining the target voice.

Limits

This skill focuses on content understanding, structured breakdown, and executable writing. It is not for ad optimization, account operations, or automatic publishing. WeChat Channels currently supports single-video analysis only, not batch account collection. Analysis works on user-provided public content and does not bypass login or platform access controls; cookies, tokens, and API keys should not appear in commands, reports, or replies.