MiguelBai/yanni-video-editor

>- One time install for the Yanni video editor. Use when the user says "set up the video editor", "install the editor", "setup", runs /yanni-editor:setup, or when the edit skill finds a missing tool (ffmpeg, node, whisper, hyperframes) or no ~/.yanni-editor/brand.json. Installs everything with Homebrew, downloads the speech model, and saves the user's brand (name stamp, accent color, vocabulary).

Was ist yanni-video-editor?

yanni-video-editor is a Claude Code agent skill that >- One time install for the Yanni video editor. Use when the user says "set up the video editor", "install the editor", "setup", runs /yanni-editor:setup, or when the edit skill finds a missing tool (ffmpeg, node, whisper, hyperframes) or no ~/.yanni-editor/brand.json. Installs everything with Homebrew, downloads the speech model, and saves the user's brand (name stamp, accent color, vocabulary).

Funktioniert mit~Claude Code~Codex CLI~Cursor

npx skills add https://github.com/MiguelBai/yanni-video-editor/tree/HEAD/skills/setup

Dokumentation

Setup: Yanni video editor

The user is not technical. Do the work yourself, explain in one short line what is happening, never paste walls of terminal output at them.

1. Install the tools

Run the bundled script from this skill's folder:

bash "<this skill's base directory>/setup.sh"

It is idempotent. Read its output:

MISSING Homebrew: Homebrew needs the user's Mac password, which you cannot type. Tell them, in plain words: "Paste this into the box below with a ! in front, press Enter, and type your Mac password when it asks (nothing shows while you type, that is normal):" ! /bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)" When it finishes, run setup.sh again.
Any other MISSING line: fix it (usually brew install <thing>) and rerun.
SETUP DONE: continue.

Speech model: small.en (470 MB) is the default and is good. If the user wants better spelling of hard words, run bash setup.sh medium.en and put "whisperModel": "medium.en" in brand.json.

2. Ask the brand questions (AskUserQuestion, one pop up, not a questionnaire)

Ask at most these, offering sensible defaults:

Name stamp. The big two line title that slams onto the screen in the first face shot (in the reference it was "ANABOLIC / ARCHIVES"). Default 100X / YANNI.
Accent color. Used for the second stamp line and for red stat numbers. Default the reference red #D6192B. Other good options: electric blue #1E5BFF, lime #7CD322.
Where finished videos go. Default ~/Movies/Yanni Edits.

3. Write `~/.yanni-editor/brand.json`

{
  "line1": "100X",
  "line2": "YANNI",
  "accent": "#D6192B",
  "projectsDir": "~/Movies/Yanni Edits",
  "whisperModel": "small.en",
  "vocab": ["BPC-157", "TB-500", "GHK-Cu", "semaglutide", "tirzepatide", "retatrutide",
            "ipamorelin", "CJC-1295", "MOTS-c", "creatine", "hypertrophy", "macros",
            "protein", "insulin", "testosterone", "cortisol", "IGF-1", "GLP-1"]
}

vocab is fed to the speech model so peptide and compound names come out spelled right. Add any brand, supplement or coach names the user mentions.

4. Smoke test

Run a 3 second test so the first real edit does not hit a surprise:

E="<plugin root>/skills/edit/engine"
T="$(mktemp -d)"
ffmpeg -v error -f lavfi -i "testsrc2=s=1080x1920:r=30:d=3" -f lavfi -i "sine=f=300:d=3" -shortest -c:v libx264 -c:a aac "$T/clip.mp4"
node "$E/prep.mjs" "$T/clip.mp4" "$T/proj"

It should print "ok":true. Then tell the user, in one or two lines: setup is done, and to edit a video they just say "edit my video" and drag the file in.