download-online-video
Skill1 starsChanged 45 days agoMITView on GitHub
Download user-authorized videos from Bilibili, Douyin, TikTok, X, and YouTube at the best available native quality or a requested resolution, then verify the local media file. Use when the user supplies a supported video URL and asks to save the video locally; do not use for DRM bypass, paywall circumvention, or unauthorized access.
In this repository18 other files it gives its agents.
- AGENTS.mdAGENTS.md
- codebase-designSkill
- eli5Skill
- find-skillsSkill
- improve-codebase-architectureSkill
- planning-with-filesSkill
- refine-promptsSkill
- show-meSkill
- app-logo-generatorSkill
- decompose-work-atomicallySkill
- download-video-transcriptsSkill
- create-better-spritesheetSkill
- video-to-spritesheetSkill
- handoff-next-goalSkill
- handoff-next-taskSkill
- human-reasoningSkill
- refine-promptsSkill
- solve-from-first-principlesSkill
The file
We split this skill into the rules it gives an agent
7 rules. About 0 are common, 7 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈7
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 12
Treat "original", "source", or "native" quality as the highest-quality source stream the platform exposes.
- Distinctive: line 17
Do not download subtitles, thumbnails, comments, metadata sidecars, or alternate audio tracks unless requested.
- Distinctive: line 35
Add authorized browser cookies only if the public probe fails for an access reason.
- Distinctive: line 40
Do not infer quality solely from file size or a format label.
- Distinctive: line 51
Treat this as a starting shape, not a substitute for the format inspection.
- Distinctive: line 53
Remove or move aside only temporary files created by this run.
- Distinctive: line 65
Do not claim success until the final file passes inspection.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Download Online Video
Download the requested media as a real local file and report an evidence-backed result. Prefer yt-dlp for extraction, ffmpeg for stream merging or remuxing, and ffprobe for verification.
Preserve the request
- Treat "original", "source", or "native" quality as the highest-quality source stream the platform exposes. Never upscale and call it original.
- Honor an explicit resolution cap or codec/container requirement. If the exact request is unavailable, select the closest non-upscaled result and disclose the mismatch.
- Download only the referenced item by default. Use
--no-playlistunless the user explicitly requests a playlist, channel, collection, series, or multi-item post. - A URL playback-time parameter is not a trim request. Preserve the complete video unless the user supplies an explicit time range.
- Save in the user-specified directory, or the current task directory when none is specified. Avoid overwriting existing files.
- Do not download subtitles, thumbnails, comments, metadata sidecars, or alternate audio tracks unless requested.
Access and privacy boundary
Proceed only with public media or content the user is authorized to access. Do not defeat DRM, a paywall, geographic access controls, account permissions, or other technical access restrictions.
Try public extraction without cookies first. When extraction requires a login and the user is already authorized in a local browser, --cookies-from-browser <browser> may be used without printing, exporting, or copying cookie contents. Never persist a cookie file unless the user explicitly requests it and understands the exposure risk.
Workflow
- Normalize the URL without removing selectors such as an X
/video/2suffix or a Bilibili page index. Quote URLs in shell commands. - Inspect the local tools. Use an existing current
yt-dlp; if absent, installyt-dlp[default]in a task-local virtual environment rather than modifying system Python. Locateffmpegandffprobebefore choosing separate video and audio streams. Probe the source before downloading:
yt-dlp --no-playlist --list-formats "URL"Add authorized browser cookies only if the public probe fails for an access reason. Read references/platforms-and-recovery.md when the site requires platform-specific handling or extraction fails.
- Choose formats from the observed list:
- For highest native quality, prefer the best source video plus the best compatible source audio. A same-quality progressive file is preferable when it avoids an unnecessary merge.
- For a resolution cap, select the best video at or below the requested height plus compatible audio. Do not silently choose a higher resolution.
- Preserve source codecs and remux when possible. Transcode only when the user requests compatibility or the target container cannot carry the selected streams.
- Do not infer quality solely from file size or a format label; compare resolution, bitrate, frame rate, codec, dynamic range, and whether audio is present.
Download with an unambiguous output template and
--no-overwrites. A typical adaptive-stream command is:yt-dlp --no-playlist --no-overwrites \ -f "bestvideo*+bestaudio/best" \ --merge-output-format mp4 \ -o "%(extractor)s_%(id)s_%(title).80B.%(ext)s" \ "URL"Treat this as a starting shape, not a substitute for the format inspection. Use explicit format IDs when they better preserve the requested native quality or codec.
- Verify the final artifact with
ffprobe. Confirm the file is non-empty and inspect at least container, duration, video codec, width, height, frame rate, audio codec, sample rate, channels, and file size. If the source has no audio, say so rather than treating it as a successful audio download. - Remove or move aside only temporary files created by this run. Do not remove pre-existing media or unrelated work.
Result contract
Return a clickable absolute file path and report:
- actual resolution and frame rate;
- video and audio codecs;
- duration and file size;
- whether this is the highest native source tier or a user-requested cap;
- any material mismatch, missing stream, authentication dependency, or source limitation.
Do not claim success until the final file passes inspection. Distinguish platform-advertised estimates from properties verified on the downloaded file.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

