A good Reels script rests on one structure: a hook in the first seconds, the pain-point context, a step-by-step part and a close with a call to action. Explaining that structure to Claude anew in every chat is a waste of time. The fix is to package it as a skill: a saved instruction Claude applies on command instead of on a fresh brief. Below is the full text of such a skill and how to call it.
What a Skill is and why it belongs here
A Skill in Claude is not a one-off prompt but an instruction that lives in the Customize section and is triggered by a slash command in any chat. The difference from a regular prompt is simple: a prompt has to be written (or pasted) every time, a skill is set up once and then just named. For content that means the script structure (hook / pain / steps / close) is locked in for good, and only the video topic changes.
Step 1 — open the Skills section
In the Claude interface go to your profile → Customize → the Skills tab. This is a separate store of skills, distinct from ordinary chats and Projects — whatever you add here is available from any conversation.
Step 2 — create a skill with the instruction
Hit Add skill and paste the whole instruction text. Here is a working version — no brands, no specific numbers, structure only:
You are a script producer for short vertical videos (Reels/Shorts).
Keep the four-part script structure regardless of the topic:
1. Hook (0–3 sec) — the opening line that has to stop the scroll.
No preamble: go straight to a conflict, a question or a surprising fact.
2. Pain-point context (3–8 sec) — why this matters to the viewer, which
problem they recognize in themselves.
3. Steps (8–35 sec) — the step-by-step part as a list. Every item is
a concrete action or technique, no generalities.
4. Close — a one-line takeaway and an explicit call to action: save,
share, comment.
Rules:
— never use lines like "this changes everything" without specifics;
— every step is a measurable action, not a wish;
— at the end of the script offer two alternative hooks and ask
which one to develop further.
When I name a topic — write the script by this structure right away,
without asking about the format.
Save it — and the structure is fixed for good, no matter how many new chats you open.
Step 3 — call it with a single command
From then on, instead of writing a brief from scratch you just name the topic and invoke the skill with a slash command:
write a reels about editing with neural nets /viral-content-structure
The command name after the slash matches the name you gave the skill when creating it — Claude layers its instruction on top of your request and immediately returns the text in the hook → pain → steps → close structure, with no clarifying questions about the format.
How to check the output
Before you publish the script, run through this checklist:
- The hook reads in 2–3 seconds and needs no outside context.
- The pain-point context names a concrete situation, not an abstract problem.
- The steps have no filler — every item can be repeated right away.
- The close contains one clear action for the viewer, not two or three at once.
If something is off — ask Claude for an alternative hook or to rewrite one specific step without touching the rest of the structure.
How we do it
The qvib.pro knowledge base already holds more than five hundred cards like this one: skill teardowns, ready prompts and content structures — no DMs, no code words, everything in the open. The same principle — a saved instruction instead of a repeated brief — is at the core of the qvib engine: it builds code and content along a structure fixed in advance, not from a blank page on every run.
Related links
- The "Reels script generator" skill: qvib.pro/arsenal/skills/generator-scenariev-reels
- A breakdown of viral content structure: qvib.pro/arsenal/skills/struktura-virusnogo-kontenta
- How the qvib engine works: qvib.pro/