Winners share a skeleton before they share a topic. The skill that pulls the skeleton from any three winning videos and writes your video into it, in your own words.
I used to watch a winning video and think, I could say this about my thing, in my own voice. The idea was never the problem. Staring at a blank page trying to reverse the structure by hand, that was the problem.
So we built the blueprint flow. Give it three to five winners. It pulls out the beats they share, not any one video's exact outline. Then it interviews you, and if your answer is thin, it pushes back for the real story instead of settling for a summary.
First version I ran gave me a flat script, because I gave it flat answers. The fix wasn't the tool. It was learning to answer with stories, not conclusions. Give it stories, get back a script you can read straight off a teleprompter.
Watch the winner three times, take notes, try to write something similar. Two hours later you've got a worse copy of their video and none of your own material in it.
Paste three references, answer the interview honestly, and get a structured script in your words with the thin spots flagged. Thirty minutes.
Blueprints saved and reused, each one tracking how the videos made from it performed against your own median, so the structures that keep producing outliers rise to the top.
It distills a beat-by-beat blueprint from your references, runs a short interview for your real material, probes where answers are thin, assembles the script in teleprompter format, then does a craft pass for spoken storytelling and an integrity sweep so nothing gets invented.
---
name: script-blueprint
description: Distills the common structure from 2–5 winning videos into a reusable beat-by-beat blueprint, then interviews you for your real material and writes your script into that structure — teleprompter-formatted, your phrasing kept, honest placeholders where you didn't answer. Trigger on "turn these videos into a template", "write my script in this structure", "blueprint these winners", "fill-in-the-blanks script".
---
# Script Blueprint
Winners in a niche share a structure long before they share a topic. This skill extracts
that structure, then writes YOUR video into it using only YOUR material. It never borrows
their words, their claims, or their facts. It borrows the skeleton.
## Stage 1: Distill the blueprint
**Input:** 2–5 reference videos (URLs) or your own best past scripts. Get transcripts
through a connected research tool (YouTube transcript connector, ScrapeCreators via a
connector hub, or the free repo at github.com/drleadflow/thumbnail-system-starter which
does this whole loop with a UI). Manual fallback: paste the transcripts.
With the transcripts in hand:
- Find the COMMON structure — not any one video's outline, but the pattern the strongest
performers share. Where they diverge, follow the stronger performer.
- Produce 6–10 beats, in order. Each beat gets: a short name, its retention purpose
(hook, proof, open loop, payoff, re-hook, pivot, close), a one-line writing instruction,
and a [bracketed placeholder] describing exactly what the creator must supply.
- Name the blueprint by its SHAPE ("Receipts-First Case Study", "Honest Experiment
Chronicle"), never by any source's topic.
- Structure only. Zero words, claims, or facts carried over from the sources.
Save the blueprint. It's reusable across every future video with that shape.
## Stage 2: The interview
Ask the shortest set of questions (5–8) that extracts everything the beats need and only
the creator can supply: their real numbers, their story, their specific examples, their
opinion, their call to action. One question at a time. Plain spoken language. Never ask
for anything the AI could invent — those aren't questions, they're padding.
## Stage 3: The probe (where most scripts go flat)
Read the answers. Any answer that is a conclusion instead of a story ("you have to
optimize for cost") will produce a flat beat. Ask up to four follow-ups that dig:
- What happened? What broke? What did it cost?
- The specific number, timeframe, before/after.
If every answer already has story + specifics, skip this stage.
## Stage 4: Assemble
Write the script INTO the blueprint using the interview as the ONLY source of facts,
numbers, stories, and claims. Rules:
- Where a beat needs material that never arrived, write `[YOUR …: description]`. Never
fill from general knowledge, never borrow subject matter from the references.
- Never change what a number refers to. 15 agents stays 15 agents.
- Keep the creator's phrasing where it's quotable. Rough spoken lines beat polished
copy.
- Teleprompter format: one idea per line, short lines at breath points, **bold** the
punch words, `## Beat · ~M:SS` headers at ~150 words per minute, blank line after
each beat.
## Stage 5: The craft pass
Rewrite for spoken storytelling: claims become moments (what happened, then the lesson),
rhythm varies (long line, short punch), one open loop early that pays off late — all from
interview material. Where a beat has no real material, do not pad it. Mark it:
`[THIN: the story the creator should add — e.g. the week this first broke]`.
Then an integrity sweep, claim by claim: any tool name, statistic, timeframe, or workflow
detail not in the interview gets deleted or placeholdered.
## Stage 6: Track what works
Each time a script from a blueprint gets published, log how the video performed against
the creator's own median. Over time, blueprints earn a track record — and the ones that
keep producing outliers become the ones you reach for first.
Inside the AI CEO Lab I teach the full pipeline from research to teleprompter, run by agents. Or my team builds it around your voice and your winners.
One email a week. What changed, why it matters, what to do about it. Skip it and find out the hard way.