agentleFS
charlie947charlie947/reel-editing-skills/skills/explainer-video/SKILL.md

explainer-video

Skill9 starsChanged 13 days agoMITView on GitHub

Edit a talking-head video in a clean Apple-style motion look, white canvas, one accent colour, one evolving object per beat that changes on the spoken word, switching between full face, split screen and full graphic every 3-5 seconds, with blur-in captions. The style behind my best-performing motion reel. For how your product works, a process, a framework, onboarding. Use when someone says "Apple style video", "premium motion edit", "clean motion graphics video", "edit my video like Apple", "white motion edit" or "motion graphics over my talking head".

We split this skill into the rules it gives an agent

7 rules. 0 are common, 6 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 16 edited versions we read keep these.

  • 6

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Before you start: run video-intake (cut take, words.js, face frames, starter).

    line 12
  2. Distinctive:

    No orange or bright colours except a brand's real logo.

    line 28
  3. Distinctive:

    Hold any on-screen line 1.5s: after its words are sharp.

    line 41
  4. Distinctive:

    Words and objects sit above your head,: never behind it.

    line 43
  5. Distinctive:

    Small objects in a big panel look cheap: Every beat called "not premium" measured at about a third of its panel.

    line 67
  6. Distinctive:

    Ticks and badges sit inside their card,: never on a clipped corner.

    line 69
  7. Distinctive:

    A "checks its own work" beat shows different designs,: never one design repeated.

    line 70

“Common” is measured: kept by at least 6 in 10 of the 16 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Explainer video

Use it for: how your product works, a process, a framework, onboarding.

This is the edit style of my best motion reel: 25,998 plays, about 11x my normal. A white canvas, one accent colour, and one object per beat that changes exactly when you say the word.

Before you start: run video-intake (cut take, words.js, face frames, starter). Run opening-hook for the first 2 seconds.

Every size. The same build makes 9:16, 4:5, 1:1 and 16:9. Size objects from PANEL, then run every-format.

Step 1: Ask for three things, then wait

  1. The cut transcript with word timings (from video-intake).
  2. The beats. Split the script into ideas of 3-5 seconds. One idea, one object.
  3. The real assets. Logos, app icons and screenshots the script names. Attached files only, 512px or bigger.

The look

  • White canvas. Text #1d1d1f. One accent, #0071e3 by default. One accent only.
  • SF Pro or Inter. Big words at weight 700-800 with tight tracking.
  • White cards: 44px corners, a hairline edge, a soft long shadow.
  • Captions: one line in a white pill at y 962, weight 650, about 52px. Each word fades in from a 7px blur as it is spoken.
  • No orange or bright colours except a brand's real logo.

The grammar

LayoutWhenWhat fills it
Full faceThe open and the call to actionYou, with the text hook from frame 0
SplitMost beatsOne object fills the top panel (x 74-1006, y 170-880), your face card below
Full graphicAbout 40% of the videoOne designed object fills the frame, no face
  • Switch layout every 3-5 seconds. One scene for the whole video reads as static.
  • One object per beat, and it changes ON its word. "Three steps" = the third step lands on "three". Stretch the beat's own clock so the last change hits the word.
  • The object carries across the cut. It swings away as the next one rises. Never cut to an empty frame.
  • Hold any on-screen line 1.5s after its words are sharp. 0.8s is too fast to read.
  • No zoom or push-in while a shot is on screen.
  • Words and objects sit above your head, never behind it.

Step 2: Write the beat sheet, then wait

One row per beat: start, end, layout, the spoken words, the object, and the word it changes on. Show it and wait for a yes. Then build the first 10 seconds and show stills before the rest.

Step 3: Build brief (paste into Claude Code inside the video folder)

Edit video.js only. Follow my beat sheet exactly, with its times from words.js.
Look: white #ffffff, text #1d1d1f, one accent #0071e3, SF Pro or Inter. White cards with 44px corners and a soft long shadow.
Each beat draws ONE object that fills its panel and changes on the word I named. Use spring() and easeOut() from the starter. No linear moves, no bounce on text.
Split beats: object inside x 74-1006, y 170-880. Full beats: object fills the frame but keeps y 930-1040 clear for the caption.
The outgoing object swings away while the next one rises. No empty frames. No zoom while a shot is on screen.
Use only my attached logos and screenshots. Never invent a number, a logo or a result.
After the build, open index.html?render&t=<mid-point of every beat> and check: object fills its panel, nothing over the caption, nothing below y 1500, nothing right of x 950 between y 1000 and 1700.

Step 4: Render and check

node render.mjs --stills <times> for a quick look, then node render.mjs for the video. Run video-check before you post.

What I learned the hard way

  • Small objects in a big panel look cheap. Every beat called "not premium" measured at about a third of its panel. Make it fill.
  • Blue key words and outlined captions look cheap too. Plain near-black words in a white pill read as premium.
  • Ticks and badges sit inside their card, never on a clipped corner.
  • A "checks its own work" beat shows different designs, never one design repeated.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.