Nex Brand Extractor

What it does

• Ships brand_extract.py, a self-contained async Python module with one public function extract_brand(url, logo_url, archetype_slug) returning a BrandResult dataclass: primary, secondary, accent, background, and text colors plus heading and body fonts.
• Runs four strategies in order: rendered page via Playwright (handles JS-built sites), CSS custom properties and theme-color meta, logo color extraction via colorthief, and archetype default palettes as the floor.
• Filters near-white, near-black, and low-saturation greys so a page never gets themed in accidental neutrals, and cleans font-loader hashes from font names.
• Blocks private and internal hosts with an SSRF guard before any fetch; every strategy returns None on failure, the module never throws.
• Includes references/strategies.md (why this strategy order) and references/archetype-defaults.md (default palettes and fonts per sector archetype).

Why use this

You get read-the-colors-off-this-URL as a drop-in module instead of writing scraping, filtering, and fallback logic yourself. Pulled from the brand-extraction step of the production demo factory at a Belgian AI agency, where it themes generated demo sites to match each prospect's existing branding. Core dependencies are just httpx and beautifulsoup4; Playwright and colorthief are optional and degrade gracefully.

See it in action

Prompt:

Pull the brand colors from janssens-vastgoed.be so my generated demo matches.

Result:

from brand_extract import extract_brand

result = await extract_brand(
    https://janssens-vastgoed.be,
    archetype_slug=vastgoed,
)
# BrandResult(primary=..., secondary=..., accent=...,
#             heading_font=..., body_font=..., source=css)

The source field tells you which strategy produced the palette, so you know whether you got real brand colors or the archetype floor.

Use cases

• Auto-theming generated demo sites to match a prospect's existing branding
• Building a theming step into a site generator or page builder
• Extracting a logo's dominant colors when the CSS has no usable variables
• Reading theme-color meta and CSS custom properties off a live site
• Guaranteeing a sensible palette even when nothing is detectable (archetype fallback)

Known limitations

• It reads an existing site; it does not generate a palette from scratch and does not scrape page content.
• The logo strategy needs a raster image; SVG logos are not supported. Contrast and WCAG checks are out of scope.
• The rendered-page strategy requires installing Playwright plus a browser; without it the module starts at the CSS strategy.

FAQ

What exactly is in the download?

SKILL.md, README.md, manifest.json, CHANGELOG.md, LICENSE.md, scripts/brand_extract.py, and two reference docs: references/strategies.md and references/archetype-defaults.md.

Which agents does it work with?

Claude Code, Claude Cowork, Cursor, Codex CLI, and any agent that reads SKILL.md skills. The Python module also runs standalone.

Do I need anything else?

Python 3.11 with httpx and beautifulsoup4. Optional: playwright for the rendered-page strategy and colorthief for logo extraction. No API keys.

What happens on a site with no detectable brand colors?

The extractor falls back to the archetype defaults you pick (default premium-services), so extract_brand always returns a complete, usable BrandResult.