ugc-ad-scripts
Skill0 starsChanged 21 days agoMITView on GitHub
Write UGC (creator-style) video ad scripts that sound like a person, not an ad. Picks a format beyond the overused testimonial, sizes the script to the video length by words per second, structures it beat by beat (hook, problem, demo, proof, ask), and writes it so it works with the sound off and when an AI video model or voice reads it.
1 other project uses a version of this.
Adopted by
Most recently taken up by
- mbfinotti/advertising-skillstheir own version, first seen Oct 11, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- mbfinotti/advertising-skills★ 2same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository6 other files it gives its agents.
- ad-copy-frameworksSkill
- ad-hooksSkill
- ad-library-researchSkill
- ad-teardownSkill
- campaign-analystSkill
- social-media-specsSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 4 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
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 59
Write each beat as what the camera sees plus the line, so whoever films or generates it knows both:
- Distinctive: line 74
Write numbers as they are spoken: when a voice or an AI video model will read the line: "ninety-nine dollars", "thirty percent off", "two …
- Distinctive: line 85
Write in the creator's register: contractions, short sentences, one aside.
- Distinctive: line 87
No claims the product cannot back up, and nothing a regulated category forbids (health, finance, weight loss).
≈ “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
UGC ad scripts
A UGC ad is a short video that looks like a person filmed it on their phone: one creator, one product, a real setting. It works because it does not look like an ad for the first two seconds. Most bad UGC scripts fail in the same three ways: they open slowly ("Hi guys, so today..."), they are the same talking-head testimonial everyone else runs, and they hold more words than the video has time for.
1. Pick the format first
The talking-head testimonial is the most overexposed format in paid social. Viewers recognise it as an ad almost instantly. Use it, but rotate in the others:
| Format | What happens | Good for |
| Testimonial | Creator tells the camera what changed for them | Products with a clear before and after |
| Scenario drop-in | Open in the middle of the problem moment, no setup | Everyday pains (sleep, cleaning, commuting) |
| Hot take | "Unpopular opinion: ..." then the product as the proof | Crowded categories with a default choice |
| Diagnostic | "If you have this exact problem, watch this" | Narrow audiences who self-identify |
| Demo | Hands and product, the creator narrating what they do | Anything with a satisfying physical action or a fast app flow |
| Third-person proof | Creator reacts to someone else's result (a partner, a friend, reviews) | Gifts, family products, B2B |
| List | "3 things I stopped buying after..." with the product as one | Bundles, routines, comparisons |
| Us vs them | The usual way side by side with this way | A product that replaces a habit |
Make 3 to 5 scripts in different formats for the same product before you make 5 variations of one.
2. Size it to the length
A creator speaking naturally to camera says about 2 words per second; a fast, punchy read reaches about 2.4. Budget for the slower rate and leave half a second at the end:
| Video length | Spoken words |
| 9 to 10 s | 15 to 18 |
| 15 s | 25 to 30 |
| 20 s | 35 to 40 |
| 30 s | 55 to 60 |
| 45 s | 85 to 90 |
| 60 s | 110 to 120 |
A script over budget gets read too fast, cut off, or (by an AI video model) silently drifts. Cut words, never speed.
3. Structure, beat by beat
- Hook (0 to 2 s). The visual and the first line land together. No greeting, no name, no logo intro. See the
ad-hooksskill for 17 mechanics by name. - Problem (2 to 5 s). One specific, relatable moment. "My back hurt after every flight" beats "back pain is common".
- The product, on screen by second 3 to 5. Shown in use, not held up next to the face.
- Proof (the middle). One concrete result with a number or a detail only a user would know. "I went from 6 alarms to 1", "it fits under the seat in front". One proof point lands; four are a list nobody remembers.
- The ask (last 2 to 3 s). One action, said plainly: "link's below", "it's 30% off till Sunday". If there is an offer, say it here, once.
Write each beat as what the camera sees plus the line, so whoever films or generates it knows both:
HOOK [Close-up: she pours coffee, it overflows because she's staring at her phone]
"I used to lose the first hour of every day to my inbox."
PROOF [Over the shoulder: the app sorts 40 emails into 3 piles in one tap]
"Now it's sorted before my coffee's done. Forty emails, three piles."
ASK [Back to her, sipping, relaxed]
"Free for two weeks. Link's below."4. Sound off, and voice-safe
- Most feeds start muted. The first frame and the captions have to carry the hook on their own. Put the key line on screen as captions (burned in during editing, short lines of 5 words or fewer).
- Write numbers as they are spoken when a voice or an AI video model will read the line: "ninety-nine dollars", "thirty percent off", "two times faster". A model reading "$99" or "2x" may say it wrong or skip it. Keep the digits for the on-screen caption.
- Keep on-screen text out of AI-generated footage. Video models garble text on signs, labels and screens. Add text, prices and logos in the edit.
- One action per shot. "She opens the box, smiles, and pours a glass" is three shots. Each extra action in one shot is one more thing that can go wrong on camera or in a generated clip.
5. Casting and voice
- Cast the person the target viewer would trust, not a default. Match age, situation and register to the buyer.
- Write in the creator's register: contractions, short sentences, one aside. Read it out loud; if it sounds like a brochure, rewrite it.
- No claims the product cannot back up, and nothing a regulated category forbids (health, finance, weight loss).
Checklist before handing it off
- The hook works on mute, in the first frame
- Word count fits the length table above
- Product on screen by second 3 to 5
- One proof point, with a number or a specific detail
- One ask
- Numbers written out for the voice, digits kept for captions
- At least three formats tried, not three wordings of one
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.

