Brand

Brand

Treat brand as a governed system of decisions. Preserve approved identity and record intentional changes.

Route the request

• Brand foundation: voice, audience, positioning, values, and messaging.
• Visual identity: logo use, color, typography, imagery, and layout behavior.
• Asset governance: naming, folders, versions, approval, and provenance.
• Compliance review: compare an artifact with approved rules and report deviations.
• Update: change one controlled source, then propagate and verify dependents.

Read only the relevant file under references/. Use templates/brand-guidelines-starter.md when the user wants a new durable guideline document.

Workflow

1. Locate the source of truth.

   • Prefer the user's existing guideline, tokens, logo files, and approved examples.
   • Separate confirmed rules from inferred patterns.

2. Define the audience and communication job.

   • Identify who should think, feel, or do what.
   • Keep positioning, promise, proof, and tone distinct.

3. Build or update the system.

   • Use semantic names for colors and tokens.
   • Define voice with observable do/don't rules and examples.
   • Specify logo clear space, minimum size, backgrounds, and prohibited treatments.
   • Define asset ownership, approval status, version, and source provenance.

4. Apply consistently.

   • Reuse approved tokens and assets.
   • Adapt tone to channel without changing brand character.
   • Record deviations required by accessibility, platform, or legal constraints.

5. Validate.

   • Run node "/scripts/validate-asset.cjs" for supported asset checks.
   • Use references/consistency-checklist.md and references/approval-checklist.md for manual review.
   • Check contrast and readability even when an approved color combination is supplied.

6. Synchronize only when requested.

   • node "/scripts/sync-brand-to-tokens.cjs" --dry-run previews brand-to-token changes.
   • Review the diff before writing generated token files.

Output contract

State:

• source-of-truth files used;
• confirmed rules and inferred recommendations;
• changes or violations by severity;
• affected assets or channels;
• validation performed;
• approvals still required.

Guardrails

• Do not silently rebrand an existing product.
• Do not overwrite master assets or approved guidelines without authorization and a backup.
• Do not infer legal trademark clearance.
• Do not weaken accessibility to preserve a visual rule; propose a compliant variant.
• Do not hardcode an AI provider, model, or host-specific skill path.