Short-Form AI Video is a Claude AI skill — platform-native scripts with current hook formats, pacing, and CTA patterns for TikTok, Reels, and Shorts.
The script came back in under ten seconds. Hook. Three points. Call to action. Neatly formatted, grammatically clean, utterly unusable. Not because the structure was wrong — it wasn't — but because it had no idea what platform you were posting on, what format was actually performing this week, or whether your audience responds to pattern-interrupt openers or story-first builds. It was a video script the same way a fish is a bicycle: technically something, functionally nothing.
Short-form video is the most format-sensitive content medium that exists right now. The same topic, the same message, the same creator — formatted for TikTok versus Reels versus YouTube Shorts — will perform completely differently because each platform's algorithm reads engagement signals differently, each audience has trained differently on what constitutes a watch-worthy first two seconds, and the native content conventions on each are distinct enough that crossposted scripts feel wrong to a platform's audience even when they can't articulate why.
Generic AI doesn't know any of this. It knows what a video script looks like. That's not the same thing.
Why Platform-Blind Scripts Underperform
The first two seconds of a short-form video are the entire game. Not because of some motivational content principle — because that's when the algorithm measures the completion signal that determines distribution. A video that loses viewers at second three gets suppressed. A video that holds them gets pushed. The hook isn't just a writing device; it's an algorithmic input, and what works as a hook on TikTok in May 2026 is meaningfully different from what worked in January, and different again from what works on Reels aimed at a different demographic slice.
This is the problem with scripts written from training data. The formats that AI learned from were already dated when the model was trained — and short-form format trends move faster than any model's training cycle. The "controversial statement" hook that saturated TikTok eighteen months ago now reads as a red flag for low-quality content to the algorithm and the audience simultaneously. The POV format that drove massive engagement on Reels two years ago is now a signal of lazy creative. Generic AI produces scripts that were current when the data was current. Which is to say: not now.
A script written without knowing this week's platform format norms isn't just mediocre — it actively signals low-quality content to the algorithm before a human viewer ever decides whether to watch.
Beyond the hook, there's pacing. TikTok audiences have trained on rapid cuts and pattern interrupts. YouTube Shorts audiences, pulled from a different browse context, have slightly longer patience for premise setup before the payoff. LinkedIn video — even short-form — performs better with a slower, more deliberate rhythm because the feed context is professional and the audience is not in scroll-entertainment mode. A single script written for "short-form video" without knowing the platform is optimised for none of them.
That gap is exactly what the Short-Form AI Video Prompt skill for Claude was built to close.
What Live Research Changes About the Output
Before the Short-Form AI Video skill writes a single line, it researches what's currently performing on your target platform for your content type and niche. Not what performed historically — what's in distribution right now. Current hook patterns, pacing conventions, caption structures, CTA placements, and the specific verbal and visual formats that the algorithm is currently rewarding in your category.
That research changes the output in ways that are immediately visible. A script for a financial educator on TikTok targeting 25-to-34-year-olds gets different hook language than a script for the same educator on YouTube Shorts targeting the same audience — because the browse context, the scroll speed, and the established native content conventions are different on each platform. The skill knows the difference and writes for it, rather than averaging across both.
Platform-native means the script sounds like it belongs there — not like it was written somewhere else and formatted to fit.
The three-question interview at the start of the skill establishes your platform, your niche, and your content goal — whether you're driving profile follows, link-in-bio clicks, saves, or comments, because each of those objectives changes how the CTA is written and where it lands in the script's pacing. A save-optimised script plants value early and delivers a recap CTA. A follow-optimised script builds personal connection before the ask. Those are structurally different scripts, not the same script with a different last line.
How the Short-Form AI Video Skill Works
Three inputs, then a research pass, then your script — plus a second variant at a different pacing or hook angle so you're not locked into one interpretation of the brief.
Platform-Blind Script vs Platform-Native Script
Same topic, same creator, same 45-second length target. One written by vanilla AI, one by the skill. The gap isn't in polish — it's in whether the script actually belongs on the platform it's destined for.
Body: "Here are three things you should know about index funds. Number one: they're diversified. Number two: they have low fees. Number three: they're easy to start with."
CTA: "Follow me for more investing tips and drop a comment below if you found this helpful!"
Caption: "Investing tips 💰 #investing #finance #money #stocks #indexfunds"
Build (2–20s): "Everyone told me to pick stocks. Nobody told me that 90% of fund managers don't beat the index they're trying to outperform. So I stopped paying someone to underperform and put everything into a total market fund. Here's what changed."
Payoff (20–38s): [specific outcome, personal, stated as result not advice — calibrated to avoid platform content restrictions on financial claims]
CTA (38–45s): "If you're earlier in this than I was, follow — I'm posting the three moves I wish I'd made at 24."
Caption: First line mirrors hook for algorithm text-match. Two hashtags, not eight.
The left script uses a hook that was already a cliché in 2024 and gets lower initial completion rates because the "did you know" pattern has been conditioned out of the TikTok audience's attention — they've seen it thousands of times and their thumb moves before their brain does. The right script opens with a specific dollar figure and a personal outcome, which is the hook pattern currently driving the highest completion rates in the finance niche on TikTok. The CTA on the right doesn't ask for a follow — it gives a reason to follow, which is structurally different and performs differently. The caption uses two hashtags because the current platform data shows that over-hashtagging in finance content depresses reach rather than expanding it.
Who Gets the Most from This Skill
Creators posting consistently to one or two platforms who need to produce scripts faster without losing the platform-native quality that drives growth. Founders and marketers building organic video presence for their brand on TikTok, Reels, or Shorts. Anyone who's tried generic AI scripts and found themselves rewriting them from scratch anyway.
The skill is most valuable for creators who already understand their niche and audience — because that context, fed into the three-question interview, is what lets the skill produce scripts that sound like you rather than like a content template. If you're still figuring out your positioning, the scripts will be well-structured but may not reflect a clear point of view. The skill amplifies what you bring to it.
It's less suited to creators who want full creative control over every word, or who have a highly distinctive verbal style that needs careful imitation. The variants are jumping-off points, not finished performances — they still need your delivery, your energy, and your specific examples layered in. What they do is hand you that structure already solved, so your creative energy goes into the parts that actually require you.
The Output You Walk Away With
Each run of the skill produces two complete script variants: word-for-word copy from hook through CTA, timestamped to a target duration, with notes on pacing and delivery for each section. Alongside each script: caption copy written for the platform's current algorithm behaviour, on-screen text suggestions for the moments where text overlay significantly improves completion, and a one-line note on the hook pattern used and why it was selected for your platform and goal combination.
From download to first draft: open the skill in Claude, answer three questions, receive two usable scripts. The one you choose goes straight into filming — no reformatting, no rewriting the structure, no guessing whether the hook is current. The only editing left is making it sound like you, because that part was never something a skill was going to do for you anyway.
The creators who are growing consistently on short-form aren't the ones with the best ideas — there are plenty of good ideas on every platform. They're the ones whose scripts are formatted correctly for the platform's current distribution logic on the day they post. That's a research problem as much as a writing problem. The skill handles the research part. The rest is still yours.
The next piece most people tackle from here is product ad film prompts that brief a director properly. If you're working across the full Video & Pod workflow, the Video & Pod bundle covers everything in one place.
If you're doing this manually, the ai video prompt skill is the faster path.
Put this to work: the Short-Form AI Video Prompt skill for Claude turns everything above into one guided workflow you run in a normal Claude chat. Not ready to buy? Start with a free Claude skill and see how it works first.