视频下载

SkillFiles & storage

Lets your agent download source video or audio from YouTube, Douyin, Bilibili, X and similar sites to a local file.

Then ask your AI: use the 视频下载 skill

Details

Instructions available. Your AI can read the instructions. Execution depends on the setup they require.

Add it as a step in a pipeline you build on Ahel, or connect Ahel to Claude, ChatGPT, Cursor, Claude Code or Codex and ask your AI to use it.

About this skill

Download source video or audio from Douyin, YouTube, Bilibili, Twitter/X, Xiaohongshu, and other yt-dlp-supported URLs into the content-creation workspace. Use when the user says 下载视频, 下载音频, 保存这个链接, 把源视频拉下来, or when another video workflow needs a local source file before transcription, frame analysi

What this skill tells your AI

The instructions your AI receives, as published by pluviobyte/rnskill in skills/ra-video-download/SKILL.md and read by Ahel’s review.

Download source media only. Do not transcribe, rewrite, edit, or publish it.

Storage rules

  • Keep source URLs, source titles, source metadata, and downloaded reference media inside 01-内容生产/视频工作台/.internal/ by default.
  • Use 01-内容生产/视频工作台/制作中/<日期-主题>/原始素材/ only when the caller explicitly names a production project and the media is an approved production input.
  • Never place source information in a handoff, finished script, final archive, or public document.
  • Never overwrite an existing file. Let the downloader add a stable suffix or stop.

Workflow

  1. Confirm that the URL is source media and that downloading it is within the user's request.
  2. Run the bundled downloader. It uses this order:
    • Douyin: TikHub media URL, then yt-dlp with browser cookies, then plain yt-dlp.
    • Other supported sites: plain yt-dlp, optionally with browser cookies when required.
  3. Inspect the returned local path and media probe. A successful HTTP response alone is not delivery.
  4. Require a playable audio or video stream and a non-zero duration. Report the local path, duration, dimensions, and selected download method.
  5. If every route fails, preserve the concise failure log and state the missing prerequisite. Do not claim that the video was downloaded.

Command

Use the Python interpreter that can import tikhub for the best Douyin path. In this workspace, prefer /opt/homebrew/bin/python3 when available.

/opt/homebrew/bin/python3 scripts/download_media.py '<url>'

Useful options:

# Explicit production input directory
/opt/homebrew/bin/python3 scripts/download_media.py '<url>' --output-dir '01-内容生产/视频工作台/制作中/<日期-主题>/原始素材/<项目>'

# Extract audio only
/opt/homebrew/bin/python3 scripts/download_media.py '<url>' --audio-only

# Use an authenticated browser profile for a protected public page
/opt/homebrew/bin/python3 scripts/download_media.py '<url>' --cookies-from-browser chrome

Resolve scripts/download_media.py relative to this SKILL.md; do not assume the current directory is the skill directory.

Safety and failure handling

  • Use browser cookies only for the requested public media. Do not print, copy, or persist raw cookies.
  • Read TIKHUB_API_KEY from the environment or the workspace-root .env; never print it.
  • Do not bypass private-account access controls, paywalls, DRM, or deleted/private media.
  • Treat TikHub and browser cookies as acquisition fallbacks, not as proof that reuse rights exist.
  • Keep the original source file untouched after download.
  • If ffprobe fails, reject the file even if it has an .mp4 extension.

Handoff

Return a compact result containing:

  • local media path
  • source metadata path
  • method used (tikhub, yt-dlp-cookie, or yt-dlp)
  • duration, resolution, and codecs
  • any limitation such as watermark, missing audio, or authentication failure

Then route onward only when requested: ra-逐字稿提取skill for source speech extraction, video-use for editing, or ra-video-production-director for production.

Signals

GitHub stars
2k
Forks
192
Last commit
Sep 2026
Advanced
Item type
skill
Key
ra-video-download
Source
github.com/pluviobyte/rnskill