YouTube Retention Curve to Script Keep/Cut Brief
Turn retention curves, Most Replayed heatmaps, transcripts, and comments into keep/cut script rules tied to exact timestamps.
Before & after
Same prompt, same agent — the only difference is the skill.
That's a great question to help improve your content strategy. Studying competitor videos is definitely smart. I'd suggest watching a few of their top-performing videos and taking notes on what keeps you engaged—things like pacing, editing cuts, hooks, and how they structure information. You might also look at the comments to see what resonates with viewers. Different niches have different retention patterns, so it helps to focus on channels in your specific space. Do you have specific competitors in mind?
RETENTION ANALYSIS REPORT Top 3 Competitor Videos: 00:00-00:15 HOOK ZONE (Critical) → Peak: 92% retention at 0:03 (cold open joke) → Matched comment: 'Laughed so hard I subscribed' → Pattern: Unexpected contrast to expected intro 00:15-01:00 CREDIBILITY BUILD → Valley: 71% retention at 0:45 (slow explainer section) → Matched comment: 'Speed this up' (3x likes) → Fix: 30s max on background, cut jargon 01:00-END ENGAGEMENT PEAKS → Peaks: 89% at 3:12 (demonstration), 85% at 5:40 (real-world result) → Comment theme: Reaction to proof, not theory Do/Don't Rules for Next Script: • DO: Open with counterintuitive claim (first 3s) • DO: Show proof immediately, skip theory • DON'T: Explain context before hook • DON'T: Slow pacing in first 90 seconds
About this skill
Your retention graph shows that people left at 4:10, but it doesn't tell your writer what to change. Most competitor breakdowns make it worse: they treat the public Most Replayed bar as if it showed where viewers dropped off, and they turn one viral spike into a rule. This skill matches every spike, cliff, and plateau to the transcript lines that caused it. It filters out platform artifacts and hands your writer rules they can test.
What it does
- Labels each data source for what it actually measures: Studio retention for your own videos, and the Most Replayed heatmap, chapters, transcripts, and comments for competitors
- Classifies each event as a cliff, slope, spike, plateau, or outro drop, and reads the 10-20 seconds before it to find the cause
- Catches false signals before you copy them: skip landings after recaps or sponsors, mid-roll ad dips, end-screen exits, and packaging-driven outliers
- Counts timestamped comments within each event window and separates confusion questions from payoff reactions
- Promotes a pattern to a rule only when it repeats across videos; anything weaker is marked as a hypothesis with a test for your next upload
- Lets you paste a draft script to check it line by line against the brief
What you get
An event table with timestamps, % of runtime, source, transcript excerpt, comment evidence, and cause. Then KEEP and CUT rules with a placement, a limit, and the evidence behind each, plus what to check in Studio after your next upload.
Best for
Faceless and documentary-style YouTube script leads, channel managers running monthly competitor teardowns, and retention coaches at agencies who need to give writers specific instructions, not general feedback.
SKILL.md preview
---
name: youtube-retention-script-brief
… (sign up to view the full skill)More creator skills
View all Creator skills →YouTube Script Writer (Shorts to Long-Form)
Shootable YouTube scripts from the title promise: hook in seconds, pacing by format, runtime-matched word count, one CTA, editor cues.
Faceless YouTube Doodle Image Set from a Timestamped Transcript
Turn a timestamped script into consistent MS Paint-style image prompts, sortable timestamp filenames, and an ffmpeg draft-video file.
YouTube Ideas by Search and Discovery Intent
Get YouTube video ideas with viewer intent, search or recommendation paths, distinct angles, and matching title concepts.
YouTube Hook Rewrites and Retention-Drop Fixes
Diagnoses each retention drop against your script, rewrites the first 30 seconds, and places re-hooks right before viewers leave.