Brand init

Skill cofoundy/brand-skills/skills/brand-init

Open-source brand skills that turn an idea into a real brand — name, identity, voice & brand book — without leaving your AI agent. Works in Claude Code, Cursor & 70+ agents.

Install

npx -y skills add cofoundy/brand-skills --skill brand-init

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
10 stars10 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Initialize a structured, versioned brand package (and optionally register it in a brand portfolio). Use when the user says "start a new brand", "create a brand", "set up a brand project", "new brand package", "init brand", "brand from scratch", or before any other brand work when no brand package exists yet. Creates brand.yaml + folder structure so every other brand skill reads and writes one coherent, queryable brand instead of regenerating from chat.

SKILL.md

3.4 KB, as published. Nobody here has run it

Brand Init

You set up the persistence layer for a brand: a structured package on disk that every other brand skill reads first and writes into. A brand becomes a queryable artifact (brand.yaml), not ephemeral chat output. Full spec: ../../references/brand-package-spec.md.

When to run

Starting any new brand and no package exists yet (no brand.yaml in ./, ./brand/, or brands/<slug>/).
Migrating a legacy .agents/brand-context.md into a proper package.

If a package already exists, do NOT re-init — load it and continue with the relevant skill.

Steps

1. Establish the essentials

Ask only for what you need to scaffold (the deep capture happens in brand-context):

Brand name (required)
One-liner — one sentence on what it is (the recall headline). If unknown yet, leave blank.
Where it lives — pick the mode:
- ./brand/ — in-situ, you're branding the current project (default)
- brands/<slug>/ — portfolio mode, you manage several brands
- <product>/brand/ — the brand belongs to a product repo
Portfolio? — if managing multiple brands, register this one in brands/registry.yaml.

2. Scaffold the package

Run the bundled script (dependency-free). It lives at scripts/brand.sh relative to this skill's directory:

# In Claude Code, ${CLAUDE_SKILL_DIR} expands to this skill's directory. In other
# agents, resolve scripts/brand.sh relative to where this SKILL.md was loaded.
bash ${CLAUDE_SKILL_DIR}/scripts/brand.sh init \
  --name "Acme" --one-liner "…" --out brand --date "$(date +%F)" \
  --register brands/registry.yaml      # omit --register for a single in-situ brand

This creates <out>/brand.yaml + <out>/assets/ and (if --register) upserts the registry. Pass --date explicitly when you want a reproducible run.

3. Hand off

Tell the user the package is created and what's next: run brand-context to capture the brand DNA into context.md, then naming / brand-strategy / etc. — each writes into this package and flips its flag in brand.yaml.

Querying the portfolio

# ${CLAUDE_SKILL_DIR} = this skill's directory in Claude Code; elsewhere use scripts/brand.sh
# relative to where this SKILL.md was loaded.
bash ${CLAUDE_SKILL_DIR}/scripts/brand.sh list --registry brands/registry.yaml
# slug · name · one-liner · path · [status]

Use this to answer "what brands do we have / where do they live / what's the one-liner" without opening each package.

Rules

The registry is a SSOT file, never agent memory. Brand locations + one-liners are data → they belong in registry.yaml (git-tracked, shareable), not an LLM's per-machine memory.
Don't pre-create empty section files. Only brand.yaml + assets/ at init; each skill creates its own *.md when it runs.
Never overwrite an existing package. The script refuses; respect that.