brand-kit · aykutucesai — Claude Code skill: video brand settings
The brand-kit Claude Code skill builds and checks brand settings for video projects. It activates when a video needs consistent colors, fonts, logos, or voice.
MIT ★ 128 updated 2026-11-03
/plugin marketplace add aykutucesai/motionforge
/plugin install motionforge@motionforgeWhat brand-kit does
The skill creates, loads, and validates a brand.json containing color roles, font licenses, logo files, watermark and outro settings, plus voice and style rules. It comes into play for branded videos and when a site, repository, logo, or screenshots provide the visual reference.
A Python 3 script extracts a draft from a site, repository, or CSS and HTML files; repository contents are read without execution. The checker flags missing required fields, invalid colors, and broken logo paths. It cannot see CSS added by JavaScript at runtime or resolve Tailwind palette classes. Screenshot colors are sampled manually.
What it helps with
- draft
brand.jsonfrom a site or CSS - catch missing fields and invalid logo paths
- derive brand colors from screenshots
- define video colors, fonts, and voice rules
Example prompts for the agent
- Create a
brand.jsondraft from this website. - Check
brand/<id>/brand.jsonand list any errors. - Choose brand colors for this video from these screenshots.
How to install brand-kit
1Plugin marketplace
/plugin marketplace add aykutucesai/motionforge/plugin install motionforge@motionforgeThis installs the whole motionforge plugin with all its skills
2Skills CLI (Vercel)
npx skills add aykutucesai/motionforge --skill brand-kit -a claude-code3Manually: put the files in place
~/.claude/skills/brand-kit/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/brand-kit/
1Skills CLI (Vercel)
npx skills add aykutucesai/motionforge --skill brand-kit -a codex2Manually: put the files in place
.agents/skills/brand-kit/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add aykutucesai/motionforge --skill brand-kit -a cursor2Manually: put the files in place
.cursor/skills/brand-kit/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install aykutucesai/motionforge skills/brand-kit2Manually: put the files in place
.github/skills/brand-kit/SKILL.md1Skills CLI (Vercel)
npx skills add aykutucesai/motionforge --skill brand-kit -a gemini-cli2Platform CLI
gemini skills install https://github.com/aykutucesai/motionforge --path skills/brand-kitManually: put the files in place
.devin/skills/brand-kit/SKILL.mdThe legacy .windsurf/skills/ path still works
Claude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about brand-kit
How do I install brand-kit in Claude Code?
Add the marketplace with /plugin marketplace add aykutucesai/motionforge, then install the skill with /plugin install motionforge@motionforge.
What does brand-kit do?
It creates, loads, and validates video brand settings in brand.json, covering color roles, fonts, logos, watermark, outro, and voice rules.
What does brand-kit need to run?
Python 3 is needed for extraction and validation. Brand details can come from a website, repository, CSS and HTML files, or manual review of screenshots.
Source text
SKILL.md · MIT
---
name: brand-kit
description: Create, load or check a brand.json (color roles, fonts with licenses, logo files, watermark, outro, voice and tone, do/don't) for videos made for a company, product or creator. Use when a video is for a brand, when the user shares a website, repository, logo or screenshots to match, or when brand colors/fonts/logos need to be applied consistently across videos.
license: MIT
---
# Brand kit
A brand kit is decided once and reused in every video. It lives in a `brand/` folder next to the video projects (one subfolder per brand when there are several) and is copied into each `videos/<slug>/brand.json`.
```
brand/<brand-id>/
brand.json
BRAND.md # short human summary: tone, sample lines, what to avoid
logo/ # logo files exactly as supplied (svg preferred)
fonts/ # only fonts whose license allows shipping them
```
When a brand has its own rules, they **override** the default style rules for color, ground, fonts and logo. Rules about honesty, hooks, banned phrases, real logos/screens and consent for paid steps still apply.
## 1. Find out what exists
Ask, in this order, and stop at the first that gives enough:
1. An existing `brand/<id>/brand.json` from an earlier video → reuse it; re-check only if the user says the brand changed.
2. A brand guide (PDF, Figma, notes) → read it and fill brand.json by hand.
3. The brand's own website or front-end code → run the extractor (step 2).
4. Only screenshots or a logo → sample colors from them (step 2, manual path).
5. Nothing → ask for colors, fonts and logo. **Never invent a brand identity for a real company.**
## 2. Extract a draft
```bash
python3 skills/brand-kit/scripts/brand.py extract --url https://brand.example -o brand/<id>/brand.json
python3 skills/brand-kit/scripts/brand.py extract --repo ../their-app/web -o brand/<id>/brand.json # read only, nothing executed
python3 skills/brand-kit/scripts/brand.py extract --css theme.css --html index.html -o brand/<id>/brand.json
```
How the draft is built (standard-library Python, no installs):
- Reads CSS custom properties, `:root`/`body` colors, `@font-face`, `font-family` declarations, Google Fonts links, `@fontsource/*` dependencies, `theme-color` and `og:*` meta, and logo-like files or `<img>` tags in the header.
- Assigns color roles from variable names (`--background`, `--foreground`, `--primary`, `--border`, …), resolving `var()` chains; falls back to `body`/`:root` declarations; the accent fallback is the most used saturated color and is marked as a guess.
- Records the **source** of every value. Anything not found is `null` and listed in `missing`.
- Limits: CSS injected by JavaScript at runtime is not seen; Tailwind default palette classes are not resolved; a site that blocks scripts returns nothing. Then use screenshots.
**Screenshots only (manual path):** open the image, pick background, text and the one strong brand color by eye, confirm with a color picker or `ffmpeg -i shot.png -vf "scale=1:1" -f rawvideo -pix_fmt rgb24 - | xxd` for an average, and mark each value `"source": "sampled from <file>"`.
## 3. Complete it with the user
Show the draft as a short table (role, hex, where it came from) and ask the user to:
- confirm or correct each color role and say what it **means** (`use`);
- pick the logo files (one for dark grounds, one for light grounds, a square icon);
- name the font license (OFL/Apache/commercial). Commercial fonts are used only from the user's own licensed files and never committed to a public repo;
- set the voice: register (informal/formal address), tone words, typical call to action, words to prefer and avoid;
- decide the watermark and the outro;
- list do/don't rules.
## 4. brand.json schema
| Field | Required | Content |
|---|---|---|
| `name` | yes | brand name as written |
| `colors.<role>` | `bg`, `ink`, `accent` | `{hex, name, use, source?}`; other roles: `surface`, `accent_2`, `muted`, `line`, `bg_light`, `ink_on_light` |
| `gradients[]` | no | `{name, stops[], angle, use}`; say where it may appear |
| `ground` | no | how backgrounds look (`dark`, `light`, `never[]`) |
| `fonts.display/body/mono` | display + body | `{family, weights[], license, subsets[], files[], fallback, usage}`; include `latin-ext` when any Latin language beyond English is used |
| `logo` | yes for brands | `primary_on_dark`, `primary_on_light`, `icon`, `min_width_px`, `clearspace` |
| `watermark` | no | `{file, corner: top-left|top-right|bottom-left, size_px, margin_px, opacity, from_s, hide_during[]}` |
| `outro` | no | `{duration_s, layout, handle, site, cta{lang: text}, icons[]}` |
| `captions` | no | `{font, weight, size_px, active_word, zone_y[], stroke}` |
| `voice` | yes | `{register, tone[], audience, goal, cta, words_prefer[], words_avoid[]}` |
| `do[]`, `dont[]` | yes | short, testable rules |
See `examples/northwind.brand.json` (a fictional agency) for a complete kit with a corner watermark and an outro with handle and site.
## 5. Check it
```bash
python3 skills/brand-kit/scripts/brand.py check brand/<id>/brand.json
```
FAIL: missing name/bg/ink/accent, invalid hex, a font without license, a logo path that does not exist, an invalid watermark corner or opacity. WARN: contrast below 4.5:1 (ink/bg) or 3:1 (accent/bg), no `use` for a color, no latin-ext subset, a bottom-right watermark (sits under the platform buttons), an outro under 1.5 s, empty `dont`.
Low contrast is reported, never "fixed" by changing the brand color: use the accent on larger elements or on another surface instead.
## 6. Apply it in scenes
- HyperFrames: generate CSS variables from `brand.json` (`--bg`, `--ink`, `--accent`, …) and `@font-face` rules for each file and subset; scenes use only the variables, never raw hex.
- Remotion: import `brand.json` and read colors and fonts from it in one theme module.
- Watermark: one fixed layer for the whole video at the given corner, size and opacity; hidden in the windows listed in `hide_during`.
- Outro: logo for the ground in use, handle and site in the body font, the CTA line in the video language, held at least 1.5 s; the last spoken sentence (the CTA) lands on it.
- Captions: active word in the brand's `captions.active_word` color.
- Words: run `voice.words_avoid` through the banned-phrase scan in `script-and-hooks` together with the language list.
© motionforge contributors · MIT · aykutucesai/motionforge
Similar skills
All skills for Design →brand-kit by other authors comes first.