agentleFS
TomSzenessyTomSzenessy/AI_TEMPLATE/.agents/skills/brand-sweep/SKILL.md

brand-sweep

Skill0 starsChanged 11 days agoMITView on GitHub

Names and rebrands the build so it is the user's own: name candidates with the trademark, domain, store and handle checks to run, a new palette checked for contrast, a logo brief, a voice guide, and a sweep tool that finds anything of the reference product left in the codebase. Works for a product you are matching a reference and for an original product with no reference. Use when the user says "name my app", "rebrand", "anything of X left", "trademark checks", or "before launch".

The file

We split this skill into the rules it gives an agent

2 rules. About 0 are common, 2 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Not a logo, a brief for whoever makes it (the user, a designer, an image model):

    line 92
  2. Distinctive:

    must work at 16px (favicon) and as a 1024px app icon

    line 97

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “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-sweep

Nothing launches under the reference product's identity. This skill is the line between a copy and your own product.

Two modes. Reference mode reads the angle from reference/fixes.md (review-mining) and the token roles in reference/design/tokens.json (design-tokens). Original mode (your own product, no reference) reads the angle from the confirmed goal and users in VISION.md and docs/design.md, and skips only the reference comparisons, marking them not applicable. Original mode never invents a competitor: if the brief is missing, ask the owner the few blocking questions rather than fabricate one. Tool in this folder (commands run from the project root):

python3 .agents/skills/brand-sweep/sweep.py . --avoid "Original Name,Its Company" --domains original.com --colors "#006bff"
python3 .agents/skills/brand-sweep/sweep.py . --config reference/brand.json        # same, from the brand file

The sweep only reads. You (the agent) write brand.md and brand.json (avoid, domains, colors, optional exclude) under reference/ in reference mode or docs/design/ in original mode, which sweep.py --config then reads. Colours are hex (#006bff, 006bff, #06f, #006bffaa; alpha is ignored); rgb(...) or a typo exits 2, never a silent "Clean".

Step 1: the name

Reference mode reads the angle from reference/fixes.md; original mode reads it from VISION.md and docs/design.md. Generate 20 candidates across styles: descriptive (Booklink), compound (Slotwise), invented (Calvo), metaphor (Harbor), verb (Book). Then cut to 5 with these filters:

  • Not confusingly similar to the reference product in sound, look or meaning, and not to any other app in the same category. Trademark offices consider confusion when assessing similarity; use that as a screening test, not as legal clearance (see Step 2). No puns on their name, no "-ly" twin.
  • Short, spellable after hearing it once, no awkward meaning in big languages.
  • Says something about the angle, or at least does not fight it.

Step 2: the checks, run, not assumed

For each of the 5, a row per check. Mark each to run, or the result with the date it was run. Never write "available" for a check nobody ran.

checkwhere
US trademarktmsearch.uspto.gov, the app's class (usually 9 and 42)
EU trademarkeuipo.europa.eu eSearch, or TMview for many offices
Canadaised-isde.canada.ca trademarks database
globalWIPO Global Brand Database
domainwhois name.com, or the registrar's search
App Store and Playsearch the exact name
handlesX, Instagram, TikTok, GitHub
the websearch "name + category"

These are screening checks, not legal clearance. Before spending money on the name, a trademark lawyer should do a proper search.

Step 3: palette

A new palette, written into the same token roles design-tokens set up (reference/design/ or docs/design/). Reference mode: pick a primary brand hue from a different family than the reference product's (if theirs is blue, yours is not a nearby blue). Original mode: pick a hue that fits the goal angle and any colours the owner already chose; there is no reference to rotate away from. Then:

python3 .agents/skills/design-tokens/contrast.py reference/design/tokens.json   # or docs/design/

Zero failures. Reference mode: add the reference product's brand colours to reference/brand.json so the sweep catches any that survive. Original mode: if no reference colours exist, record that in brand.json and rely on the manual artifact checks in Step 6.

Step 4: logo brief

Not a logo, a brief for whoever makes it (the user, a designer, an image model):

  • the idea in one line, tied to the name and angle
  • mark type: wordmark, symbol plus wordmark, or monogram
  • must work at 16px (favicon) and as a 1024px app icon
  • deliverables: SVG, app icon 1024x1024 with no transparency for iOS, favicon set, social image 1200x630
  • must not resemble the reference product's mark (reference mode): no shared shape, colour pair or letterform trick. Put their mark next to the drafts and check. Original mode: the mark must be your own — no copying of any known brand's shape, colour pair or letterform; check against the design references in docs/product/research.md if any exist.

Step 5: voice

Three words for how it sounds, with what each does not mean ("direct, not blunt"). Five do and don't pairs. Then rewrite the 10 most-seen strings in the build (sign up, empty states, the main button, the confirmation, the error) in that voice. All fresh, none echoing the reference product's phrasing.

Step 6: the sweep

Replace every placeholder name, colour and string. Then:

python3 .agents/skills/brand-sweep/sweep.py . --config reference/brand.json

It searches file contents and file names for the reference product's name (also inside identifiers like CalendlyEmbed), domains and colours, skipping node_modules, build output and the top-level reference/, .agent/ and .review/ folders (where their name belongs). Pass --include-reference to check reference/ too, for example if it sits inside public/.

product-kickoff requires naming the reference product in docs/product/research.md and docs/design.md; those are scanned, so exclude them (repeatable --exclude <glob>, or "exclude": [...] in brand.json):

python3 .agents/skills/brand-sweep/sweep.py . --config reference/brand.json \
  --exclude docs/product/research.md --exclude docs/design.md

Exclude only research and planning docs, never shipped code or copy. Exit 1 means something is left; exit 2 means bad input. Fix until it says clean. Also check by eye: the favicon, the page titles, the email templates, the OG image, the app icon.

Original mode with no reference product: there is no reference name, domain or colour to sweep. Do not fabricate avoid-terms to force a clean CLI exit, and do not claim the sweep passed. Record in brand.md that no reference identity exists (not applicable), skip the sweep.py invocation, and run the manual artifact checks above by eye instead. The exit 1/2 gate is unchanged: it still blocks on leftover reference material when a reference does exist.

Output

brand.md (name with checks, palette, logo brief, voice) and brand.json under reference/ (reference mode) or docs/design/ (original mode), updated tokens, rewritten strings, and a clean sweep (reference mode) or the manual artifact-check record (original mode). That clean sweep is the pre-launch gate: nothing ships while it exits 1.

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.