agentleFS
whanoftheworldwhanoftheworld/graphic-skills/skills/logo-lab/SKILL.md

logo-lab

Skill0 starsChanged 34 days agoMITView on GitHub

Develop the logo from a reason and a concept, test finalists through a gauntlet, and produce the logo kit.

  • Installs packages

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

Files
SKILL.md

Logo lab

A logo earns its place by working: small, in one color, on a sign, in a crowd. This skill goes from the reason for a mark, through concepts and rough thumbnails, to a tested choice and a usable kit.

Load the brand-context and design-craft skills (the Skill tool, or read their SKILL.md), and read logo.md before starting. Reply in the user's language. The brand memory needs a decided Direction (a chosen route, or "existing identity" kept at intake); without one, tell the human to run /design-routes.

1. Settle the reason and the type

Answer, in the brand memory, whether the brand needs a symbol at all and which type of mark fits (wordmark, lettermark, symbol, combination, emblem), using the reason test and the table in logo.md. A wordmark must use the font decided in Typography; if none is decided, tell the human to run /type-system first.

2. Research the category

Collect the marks of the competitors named in Strategy. Write the clichés of the category in one line each, and the open visual space around them. Use web search or fetch for the marks you can reach; ask the human for the rest.

3. Concepts, then thumbnails

Write five to eight concepts in one sentence each, drawn from the name, the onlyness, or a true detail of the business. Cut to three with the human on strength of idea alone. For each survivor, draw at least a dozen small, rough SVG thumbnails in a single contact-sheet file, brand/explore/logo-thumbnails.html, then choose the strongest three to five marks in total to refine. When an image generator is available, you may also generate concept sheets with the logo-brand use case (see image-prompts.md in design-craft) to widen the field; they are raster inspiration only, so redraw every chosen concept as SVG.

4. Refine on a grid

Refine the finalists with simple geometry on a visible grid: consistent stroke, at most two colors, no gradient or shadow in the master, optical balance over mathematical balance. Turn every wordmark into outlines with scripts/text-to-svg.py (pip install fonttools uharfbuzz; it shapes Thai marks correctly). It reads a TTF or OTF file, not woff2: take the static or variable TTF from the font's Google Fonts download or the google/fonts repository, and pass --wght for a variable font. For Thai or mixed lockups, follow logo.md and thai-typography.md and run the stacking test on the Thai letters.

5. Run the gauntlet

Build brand/explore/logo-gauntlet.html showing every finalist through the seven tests in logo.md (favicon, one color, reversed, squint, small print, in place, name-swap). Look at it rendered, screenshot it when a browser tool exists, and mark each finalist pass or fail per test. A finalist that fails a test is fixed or dropped, not excused.

6. Choose and build the kit

Present the results and your recommendation with the reason, and the strongest argument against it. When the human chooses, produce the five SVG masters in brand/assets/logo/ named in logo.md (primary, alternate, icon, black, white). Add clear space, minimum size, and the misuse list to Logo in the brand memory, with the concept and the rejected finalists in the Decision log.

The skill is done when one logo passed the gauntlet (or each failure is documented and accepted by the human), the five SVG files exist (or an icon or alternate is recorded n/a with the reason, as logo.md allows for wordmarks), the clear-space and minimum-size rules are written, and Logo is set to decided.

State plainly what this is: a sound concept with clean first-pass vectors. A brand carrying heavy spend should have the master finished by a professional designer.

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.