AGENT SKILL

video-maker-brand

Create or reuse the Video Maker brand of the project open in this session — name, spoken name for the voice, tagline, a 4-color palette with measured contrast, logo with transparency, fonts, voice and music — derived from the repository's own files (README, landing copy, CSS, Tailwind or theme config, web manifest, logo assets). Use it before producing any video with Video Maker, or when the user asks to set up, fix or change the brand, colors, logo or voice of their videos.

License 许可
MIT
最近更新
2026-11-15
Stars
★ 0
Forks
0

01 前言 Front Matter

name
video-maker-brand
description
Create or reuse the Video Maker brand of the project open in this session — name, spoken name for the voice, tagline, a 4-color palette with measured contrast, logo with transparency, fonts, voice and music — derived from the repository's own files (README, landing copy, CSS, Tailwind or theme config, web manifest, logo assets). Use it before producing any video with Video Maker, or when the user asks to set up, fix or change the brand, colors, logo or voice of their videos.

02 资源文件 2

03 SKILL.md 正文

Video Maker brand

Every video uses a brand of the studio: it gives the colors, fonts, logo, intro/outro look, voice and music bank. Your job is to get a correct brand for this project with the least effort for the user.

Talk to the user in their language and show the brand as a short visual summary (name, colors as hex swatches, logo file, font, voice described in words), never as JSON.

Field reference of the manifest: ../video-maker/references/brand.md (schema: ../video-maker/references/schemas/brand.schema.json). Derivation details and edge cases: references/deriving-a-brand.md.

1. Reuse first

  1. list_brands. If one matches the product (same name or an id derived from it), get_brand and show it.
  2. Ask only "use it as is, or adjust something?". Built-in brands cannot be changed: to adjust one, create a user brand with duplicateFrom set to it.
  3. If none matches, derive a new one (§2).

2. Derive from the repo

Collect each item with the file it came from:

Manifest field Where it comes from
name product name as the README, manifest or landing page writes it
spokenName the same name written the way a voice must say it ("Acme Pay" for "AcmePay", "Jey Ess Kit" for "JSKit" only if that is how people say it); ask the user when unsure. Exact tokens (package names, commands, URLs) are never spoken: they go on screen, see the motion-brief skill
tagline the landing hero line or the package description, ≤ 8 words
url the public site of the product (homepage in the package manifest, landing page); it is the default outro caption
palette.background, palette.accent, palette.accent2, palette.text CSS variables, Tailwind theme, design tokens, web manifest theme_color / background_color
logo an SVG or PNG wordmark in the repo; must have transparency (§3)
typography.display, typography.body and the fonts files the families the product uses, as woff2/ttf/otf files
voice.id a voice from list_voices in the audience's language

Palette (4 colors, contrast measured)

  • background: the color most videos sit on (dark or light, from the product's own surfaces).

  • text: the main text color on that background.

  • accent and accent2: the two brand colors used for highlights, gradients and buttons.

  • Measure contrast with scripts/contrast.mjs (WCAG ratio):

    node <this skill>/scripts/contrast.mjs '#0b1020' '#f5f7ff' '#6d5dfc' '#22d3ee'
    

    text on background must be ≥ 4.5; each accent on background ≥ 3. If a pair fails, propose the closest shade that passes and say which one you changed.

Logo with transparency (pix_fmt rule)

A logo without an alpha channel is drawn as a box over the video. For raster files check the pixel format with ffprobe -v error -select_streams v:0 -show_entries stream=pix_fmt -of csv=p=0 <file>: rgba, bgra, ya8, pal8 or any format with an a has transparency; rgb24, yuv420p, gray do not. Prefer, in this order: an SVG wordmark, a PNG with alpha ≥ 800 px wide, a square mark from the web manifest. Never pick one without alpha silently: offer to go without a logo or ask for a transparent file. Horizontal wordmarks work better than square icons; keep the square one as the optional isotipo.

Voice and music

  • The brand manifest has no tone field. How the product talks ("close and direct, no jargon") belongs to the brief and the script of each video, not to the brand.
  • Voice: list_voices; describe two candidates to the user in words (gender, accent, energy). voice.lengthScale 1.0 is normal speed, 1.08 slightly calmer.
  • Music: a new brand starts with an empty bank, and the brief music field only takes ids of the brand's own bank. Ways to fill it:
    • musicFrom (id of another brand) only for a sub-brand of the same owner that shares its parent's music. Never use it to borrow another company's bank.
    • Tracks the user owns: upload_brand_file kind music, then update_brand adding the entry.
    • Free tracks, with no download on your side: in a project of this brand run search_free_music (each result shows title, author, license and page URL), pick one whose license is not share-alike, then call import_brand_music with the project slug, the resultId, the brand id, an estimated bpm, a short mood and optionally videoTypes. The studio downloads the file (mp3, wav or m4a), stores it in the brand and appends the track with its credit (title, author, license). Use the answered track.id in the brief.
    • Any entry needs id, file, credit, bpm and mood; drop and videoTypes are optional.

3. Propose, then create

Show the summary and wait for the OK. Then:

  1. create_brand with name (and id, kebab-case starting with a letter). The studio creates a starter manifest with default fonts and a generated wordmark.
  2. upload_brand_file for each file: kind logo (svg, png, webp, jpg), font (woff2, ttf, otf), music (mp3, wav, m4a; a free track goes through import_brand_music instead), image (photos the scenes can use), device (product hardware). Each upload answers the stored relative path; use exactly that path in the manifest.
  3. get_brand, change only what you derived, and send the whole manifest with update_brand. The studio validates it and names the offending field when invalid: fix that field and send again.
  4. get_brand again and confirm to the user, in words, what the brand now has.

Never edit a built-in brand, never delete a brand (the user does that in the studio) and never upload files the repo does not own the rights to (a client's logo, a font without a license file).

4. Optional parts that make videos better

  • images: photos the motion scenes may use, each with a one-line caption saying what it shows (the storyboard AI reads the caption, it does not see the photo).
  • devices: when the product runs on hardware (a screen, a kiosk, a terminal, a tablet stand), upload a clean product photo as kind device and describe its screen area as a 4-point polygon so scenes can put the product's UI inside it. See references/deriving-a-brand.md §4.
  • bumpers: limit the intro/outro looks to the ones that suit the brand (bumpers.backgrounds, bumpers.reveals), and set bumpers.outroCaption if the outro should type something other than the url.

04 同仓库更多 Skills

e-burgos
e-burgos/video-maker

video-producer

Productor de video conversacional del Video Maker Studio. Parte de un prompt del usuario y del material que haya en `media/<slug>/raw/` (videos, fotos, pantallas grabadas, locución, música, guion, o nada), acuerda con el usuario qué video hacer y lo produce con las herramientas ya implementadas (guion, voz Piper, montaje, captions, mezcla, render), buscando medios libres solo cuando el usuario lo pide o no hay material. Invocar con `/video-producer media/<slug>` o cuando el usuario pida "armá/generá un video" sobre un proyecto.

★ 0
e-burgos
e-burgos/video-maker

tucu-ui-pro

Usar siempre que un frontend React/Next use (o vaya a usar) @e-burgos/tucu-ui — montarlo en Vite, Next 16 o una app con Tailwind propio, elegir ThemeProvider/layout/preset, tematizar con customPaletteColor y tokens --tucu-*, elegir el componente correcto con sus props válidos y esquivar las trampas que rompen en runtime, con o sin el MCP tucu-ui conectado.

★ 0
e-burgos
e-burgos/video-maker

setup-graphify

Instala y configura graphify (grafo de conocimiento del repo) para un dev que lo quiera usar. Elige un backend gratuito, valida el modelo con una medición real y construye el primer grafo. Invocar solo si el dev pide habilitar graphify.

★ 0
e-burgos
e-burgos/video-maker

sdd-steward

Conserje del kit SDD - puerta de entrada para status del harness, actualización de la librería, arranque de ideas, costos y salud de los arneses. Clasifica cualquier pedido con la tabla de ruteo, ejecuta solo lo que no tiene otro dueño y delega el resto sin bypassear gates. Invocar ante cualquier pregunta u operación sobre el kit en sí.

★ 0
e-burgos
e-burgos/video-maker

sdd-reviewer

Skill del Agente Reviewer SDD. Valida la calidad de todo el output del ciclo y ejecuta los gates de cierre. Invocar cuando todas las tasks de implementación del ciclo estén completas.

★ 0
e-burgos
e-burgos/video-maker

sdd-planner

Skill del Agente Planner SDD. Convierte historias de usuario en tasks técnicas ordenadas y estimadas. Invocar después del Funcional, en paralelo con el Arquitecto.

★ 0