agentleFS
matteotittamatteotitta/claude-marketing-context/skills/brand-kit/SKILL.md

brand-kit

Skill3 starsChanged 31 days agoMITView on GitHub

Extracts a company's visual identity from screenshots (best) or its website (fallback) and writes a DESIGN.md brand file: machine-readable tokens for colours, typography, spacing, corner radius and logo usage, plus short prose explaining how and when to use each, and a list of do's and don'ts. Any later page, deck, carousel or ad can read the tokens instead of guessing. Triggers: "brand kit", "brand guidelines", "extract our brand", "design tokens", "colours and fonts from", "visual identity", "DESIGN.md". Not for tone of voice (use /tov-guidelines) or messaging (use /messaging).

17 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

8 rules. 0 are common, 5 are distinctive, and 3 are very specific to this project.

  • 0

    Common

    Most of the 14 edited versions we read keep these.

  • 5

    Distinctive

    This author's own choices. Worth a look.

  • 3

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    When the team keeps producing off-brand pages, decks or ads.

    line 21
  2. Distinctive:

    When a brand guide exists as a PDF nobody opens, and you want it in a form AI tools can read.

    line 22
  3. Very specific:

    Don't use it for voice (/tov-guidelines), for messaging (/messaging), or to design a new identity from scratch unless that's explicitly …

    line 24
  4. Very specific:

    Paths: Write to context/brand/MMYY-brand-kit.md.

    line 37
  5. Distinctive:

    do's and don'ts

    line 62
  6. Very specific:

    every {token} reference resolves

    line 65
  7. Distinctive:

    every colour token is used somewhere

    line 68
  8. Distinctive:

    Every hex value was seen in a screenshot or the CSS.

    line 73

“Common” is measured: kept by at least 6 in 10 of the 14 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

Brand kit

Turn what a brand looks like into a file other work can rely on. The tokens at the top are the source of truth: when a later page needs the primary colour, it reads the hex value from here instead of approximating it from a description. The prose underneath explains why each choice exists and how to use it.

When to use it

  • Building the context folder, before any visual output.
  • When the team keeps producing off-brand pages, decks or ads.
  • When a brand guide exists as a PDF nobody opens, and you want it in a form AI tools can read.

Don't use it for voice (/tov-guidelines), for messaging (/messaging), or to design a new identity from scratch unless that's explicitly what the user asked for.

Inputs

Required: one visual source and the company name. The visual source is either:

  • 3 to 5 screenshots: homepage, about, pricing, a feature page. This gives the highest fidelity.
  • a website URL, where you capture screenshots yourself if you can.
  • a brand PDF.

Optional: the website URL alongside screenshots (to read exact hex codes and font names from the CSS), a Figma file, logo files, an existing brand guide.

Decide the job first: apply the brand as it is (most common), fill in only the missing pieces, or create a new identity (only when asked).

Paths. Write to context/brand/MMYY-brand-kit.md. Agency or consultant? Swap context/ for clients/<client>/context/ throughout.

Tools. Your own vision on screenshots does most of the work. Page fetching or Firecrawl can read CSS for exact values. A browser automation tool can take screenshots from a URL if you have one.

Steps

  1. Capture. Get the screenshots, including pages that show icons, illustrations and photos in context, not just the logo.
  2. Analyse. Colours (primary, secondary, accent, background, text), fonts and sizes, spacing rhythm, corner radius, buttons and cards, imagery style.
  3. Cross-check against the CSS if you have the URL. When the CSS and the screenshot disagree, trust the screenshot.
  4. Score confidence 0 to 5 on every token, and note where it came from.
  5. Write the tokens. A YAML block at the top of the file:
    • colors: at least a primary, plus secondary, accent, background, surface, text, muted. Hex values.
    • typography: heading and body font families, weights, a size scale.
    • spacing: a base unit and scale.
    • rounded: corner radius values.
    • logo: which file to use on light and dark backgrounds, minimum size, clear space.
    • Refer to other tokens as {colors.primary} rather than repeating the hex.
  6. Write the prose sections, in this order:
    • overview and mood
    • logo usage
    • colours, with what each is for
    • typography
    • layout and spacing
    • shapes
    • imagery
    • do's and don'ts
  7. Make the do's and don'ts quantitative. Accent colour on no more than about 10% of a screen. No more than two font weights per screen. One primary action per screen. Name the brand's one signature element and limit it to one use per surface.
  8. Check the file.
    • every {token} reference resolves
    • a primary colour exists
    • text and background pairs meet WCAG AA contrast
    • every colour token is used somewhere
  9. Mark gaps as [NEEDS VERIFICATION] with your best estimate.

What good looks like

  • Every hex value was seen in a screenshot or the CSS. Nothing is guessed.
  • Font names are confirmed, or described by shape with likely candidates flagged.
  • Logo rules only describe files that exist.
  • A designer, or an AI building a landing page, could produce an on-brand page from this file alone.
  • The do's and don'ts are specific to this brand, not generic design advice.

Mistakes to avoid

  • Inventing an icon set or illustration style the brand doesn't have. "No icon set" is a valid finding.
  • Approximating colours from a description ("a warm red") instead of extracting the value.
  • Describing the brand in adjectives with no tokens.
  • Letting the signature element creep into every section.

Chain

Step 7 of the suggested chain: /company-context → /icp-research → /competitor-research → /win-loss → /expert-pov → /tov-guidelines → brand-kit → /positioning → /messaging → /funnel-strategy → /goals.

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.