retention-by-scene
Skill2 starsChanged 29 days agoMITView on GitHub
Find where viewers leave a published YouTube video and which scene, chapter or sentence was playing - pulled straight from the YouTube Analytics API with the creator's own login, compared with YouTube's own benchmark for similar videos. Use for "why do people stop watching", "analyze my retention", "which part lost viewers", "what should I clip", or a video post-mortem.
In this repository13 other files it gives its agents.
- channel-pulseSkill
- check-renderSkill
- check-thumbnailSkill
- check-voiceoverSkill
- find-topicSkill
- localizeSkill
- make-shortsSkill
- morph-clipsSkill
- produce-videoSkill
- research-factsSkill
- setupSkill
- upload-privateSkill
- write-scriptSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 2 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
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 44
If the opening holds and the slide is gentle but views are low, the problem is the packaging (title and thumbnail), not the video.
- Distinctive: line 50
Not a list of ten.
≈ “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
Retention by scene
The retention curve is the most honest feedback YouTube gives. This skill pulls it directly (no CSV export), lays it over the video's chapters or script, and names the moments that cost viewers.
$SKILL below is this skill's folder.
Setup
Same OAuth client as upload-private (see its references/oauth-setup.md), with a read-only analytics login:
python $SKILL/scripts/retention.py --login --profile mychannelRun
A video under 72 hours old or under 100 engaged views gets "too early to judge" and a date to come back instead
of a report: a curve from a handful of viewers is noise. Write no lesson from it. --anyway shows the numbers,
marked NOT JUDGED.
python $SKILL/scripts/retention.py <video_id> --profile mychannel --srt voice/captions.srtSections come from the chapters in the description, or --timeline sections.json, or the captions grouped into
30-second windows.
Read it in this order
- Opening - how much is lost by 30 seconds, and YouTube's relative retention at 10 and 30 seconds (0.5 is the median of similar-length videos). The first seconds can be inflated by feed autoplay previews, so trust the relative number here. An opening problem is fixed in the script's first 15 seconds, never in the edit.
- Cliffs - the steepest drops after 30 seconds, with what was being said. A cliff is a moment: a topic turn with no signpost, a long setup, an ad read.
- Slide - steady loss per minute by section. A flat slide is pacing; the fix is cutting, not rewriting.
- Rewatch bumps - where the curve rises, people rewound. Those spans are your best clip and Short candidates.
If the opening holds and the slide is gentle but views are low, the problem is the packaging (title and thumbnail), not the video.
Report back
The single biggest leak, the one change it suggests for the next video, and one line for the channel's lessons. Not a list of ten.
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.

