A skill that makes Claude brief and judge a logo the way a senior identity designer does, one stage at a time, with you deciding at each fork. It turns your notes into a one-page logo brief: the name exactly as written, the one thing the mark should make people feel, and every place it will live, with sizes. It maps your competitors' logos and counts the clichés in your category. Then it writes three concept directions that differ in idea and structure, each with a plan for the 16 pixel browser tab, and a brief and prompts for your designer or image tool.
Inside are five reference files: the brief and competitor landscape, six structures a logo can take with type registers and symbol approaches, ten critique tests with a fix library and a usage spec, dated sizes and rules, and a worked example for a fictional animal clinic. Claude never copies or imitates an existing mark, and never names a brand or a living artist in a prompt. It never invents competitors, test results, Pantone colours or thread numbers. The favicon, app icon, avatar, embroidery and print figures, the image-tool models and the US trademark and copyright pointers were checked on 17 October 2026, each with its source.
Claude cannot draw the finished logo or generate images. It does the thinking, the briefs and the prompts, and with code execution switched on it builds a test sheet: your draft at 16, 24, 32, 48 and 64 pixels, in pure black, reversed, blurred and beside your competitors, with the thinnest stroke measured. It is not a trademark clearance search; it tells you which searches to run and when to see an attorney. It works best with four to six competitor logos, your current mark if you have one, and drafts exported as large PNG files on white.
Claude skill · version 1.0
Logo brief
Checking your access.
In the download · 9 files
- SKILL.mdThe method Claude follows: brief, landscape, three directions, handoff, critique and spec, one stage at a time.
- README.mdHow to install it in claude.ai or Claude Code, how to start, and the licence.
- references/brief-and-landscape.mdThe logo brief field by field, the application inventory, five personality scales, mapping competitors, category clichés and trust cues, and red flags in a brief.
- references/directions.mdWhere logo ideas come from, six logo structures, type registers with open-licence typefaces, symbol approaches, the direction card, a designer brief and image-tool prompts.
- references/critique-and-spec.mdTen tests with pass marks, the test sheet Claude builds with code, a fix library, feedback a designer can act on, the usage spec and the file package.
- references/sizes-and-rules.mdFavicon, app icon, avatar, print, embroidery and billboard sizes, contrast rules, image-tool models, and US trademark, copyright and font pointers, checked 17 October 2026.
Install it and begin
01Install it in Claude CodeTerminal mkdir -p ~/.claude/skillsunzip ~/Downloads/logo-brief.zip -d ~/.claude/skills/# Start it with /logo-brief. If this is your first skill, type /reload-skills once.On claude.ai there is nothing to unzip: upload the zip as it is. Every step is on Install and use a Claude skill.
02Your first message to ClaudeText Use the logo brief skill. The business is [NAME, EXACTLY AS WRITTEN], a [TYPE OF BUSINESS] in [CITY AND STATE]. Customers meet the logo on [SMALLEST AND HARDEST USES]. The competitors are [FOUR TO SIX NAMES, LOGOS ATTACHED]. Here are my notes: [PASTE YOUR NOTES]. Start with the brief and tell me what is missing.Attach four to six competitor logos and your current logo if you have one. For a critique, export each draft as a PNG at 1000 pixels or larger on a white background.
Common mistakes
Logo brief mistakes, and what to do instead
Instead ofJudging the logo at full size on a large monitor.
Do thisJudge it first at 16 and 32 pixels, in a circle avatar and in black on an office printer. Most marks fail at the small sizes first.
Instead ofA brief that lists six values for the mark to carry at once.
Do thisWrite the one thing the mark should make people feel, and a short list of what it must reject. "Not cute" removes half a category's clichés.
Instead ofThree directions that are one idea set in three typefaces.
Do thisThree ideas from different sources with different structures: one close to the category, one stretch, one bold. Cover the names and you should still tell them apart at 48 pixels.
Instead ofChoosing the colour first.
Do thisApprove the mark in solid black on white. If the idea needs colour to be understood, it is not a logo yet. Add the colour plan last and check its contrast.
Instead ofPrompting an image tool with "like the [famous brand] logo".
Do thisDescribe the shapes, how many, the line weight and the negative space. Never name a brand or a living artist, and run a reverse image search on anything you keep.
Instead ofShrinking the main logo to make the favicon.
Do thisDraw the 16 and 32 pixel versions separately on the pixel grid. At 16 pixels one pixel is 6.25 per cent of the frame, so thinner detail disappears.
Instead ofTreating "nobody else has this" as trademark clearance.
Do thisA resemblance check by eye is a first screen. Run the USPTO design-code search, then have a trademark attorney clear the mark before it goes on a sign.
What to do it with
- ClaudeFree tier, paid above it
A language model that writes, edits, reads long documents and runs tools, including agents that work inside a codebase.
Bad at It is agreeable by default, so it will improve a bad idea instead of telling you it is bad. Ask it to argue against you and the quality of the answer changes.
- Claude CodeSubscription
An agent that reads a whole repository, edits files, runs the tests and explains what it changed.
Bad at It is fast enough to make a large mistake quickly. Work on a branch, read the diff, and never let it push. Watch the starting mode: on Pro, Max and Team plans it opens in auto mode, which Anthropic's own list says may read a .env file and send the key in it to the API that key belongs to. Switch it off.
- RecraftFree tier, paid above it
Generates vector and raster artwork, and can hold a defined style across a set.
Bad at Vector output still wants cleaning by hand before it goes into an identity. Style consistency is good, not locked.
- IdeogramFree tier, paid above it
An image model built around getting words right inside the picture.
Bad at Strong at type, weaker at photographic realism. A poster comes out well, a product on a plinth less so.
- MidjourneySubscription
Generates images from a written description, with the strongest default aesthetic of anything in the category.
Bad at That default aesthetic is the problem as often as it is the draw. It has a house look, it pulls everything towards it, and getting a plain honest photograph of a real product out of it is harder than it should be. Text inside an image is unreliable.
- GPT ImagePay per use
Generates and edits images with unusually good instruction-following, and can hold text inside a picture better than most.
Bad at Faces drift between generations unless you feed a reference every time. Skin can come out plasticky at default settings, and a heavy fidelity setting costs many times more and takes many times longer.
- IllustratorSubscription
Vector drawing, where logos and dielines are actually built.
Bad at Its generative vector features produce shapes that look right and have terrible point structure underneath.
Tells you whether two colours are legible together, against the published standard.
Bad at The ratio is a floor and not a design. Passing text can still be unpleasant to read.
Why it matters
Ask a chat for a logo and you get a description of something that looks good at full size. Most logos fail in a 16 pixel browser tab, stitched on a shirt, printed in black on an office printer, or beside a competitor that looks the same. This skill makes Claude plan for the hardest uses first, keep three directions properly apart, and judge drafts by measured tests instead of taste.
- Type
- technique
- Period
- ongoing