Brand Designer

Brand Designer

A systematic toolkit for documenting brand identity. Elicits brand information through structured templates, validates completeness, compiles into reference documents, and generates brand-compliant images.

Quick Start

# 1. Scaffold a new brand
brand-book init ./my-brand --name "Acme Corp"

# 2. Fill in the templates (7 markdown files)
#    Start with strategy.md → visual-id.md → voice.md

# 3. Validate your work
brand-book validate ./my-brand/strategy.md --strict

# 4. Compile output
brand-book compile ./my-brand --artifact brand-card    # One-page summary
brand-book compile ./my-brand --artifact full           # Full brand book

# 5. Generate brand image (optional)
generate brand-card --brand-dir ./my-brand --layout landscape

When to Use This Skill

| Trigger | Artifact | Audience | |---------|----------|----------| | "I need a brand card for press/stakeholders" | Brand card (image + one-pager) | Executives, press, partners | | "I need to document my brand strategy" | Strategy memo | Leadership, marketing | | "I need visual identity documentation" | Visual identity spec | Designers, developers | | "I need brand voice guidelines" | Voice & messaging spec | Writers, content team | | "I need a full brand book for agency handoff" | Compiled brand book | Agencies, partners | | "I need to inventory all my brand assets" | Asset inventory | Everyone |

Workflow

1. Choose artifact type

The decision tree in references/canonical-components.md maps brand maturity to artifact scope:

• Startup ( [--name]` | Scaffold brand directory with 7 templates |

| compile [--artifact] [--format] [--output] | Assemble brand artifact | | validate [--strict] [--json] | Check frontmatter completeness | | preview [--json] | Terminal preview of a brand component |

All subcommands support --json for machine output and --dry-run for preview.

scripts/generate

Image prompt construction for brand-compliant generation. Subcommands:

| Command | Action | |---------|--------| | brand-card [--layout portrait|landscape] | Brand card image prompt | | mockup --type stationery|social|signage|digital | Application mockup prompt | | swatch | Palette visualization prompt | | moodboard [--theme] | Mood board prompt | | logo-bg [--background] | Logo presentation prompt |

Reference Files

| File | What It Contains | |------|-----------------| | references/canonical-components.md | Definitive component hierarchy + maturity decision tree | | references/artifact-hierarchy.md | Every artifact: who consumes it, format, depth, relationships | | references/gold-standard-analysis.md | Structural analysis of Cash App, IBM, Netflix, Bang & Olufsen, Klarna | | references/brand-card-formats.md | Format comparison: PDF vs online vs CLI vs Figma vs image | | references/template-schemas.md | Complete YAML frontmatter schemas for all 7 templates | | references/image-generation.md | Brand-compliant image prompt patterns and reference-image usage |

Artifact Tiers Summary

Tier 0: Brand Card          → image + one-pager        → stakeholders, press
Tier 1: Strategy Memo       → narrative document       → leadership, marketing
Tier 2: Visual Identity     → structured spec          → designers, developers
Tier 3: Voice & Messaging   → guidelines with examples → writers, content
Tier 4: Application         → channel-by-channel rules → producers, agencies
Tier 5: Governance          → framework document       → brand owners, legal
Tier 6: Asset Inventory     → indexed table            → everyone
       └── Compiled Brand Book (Tiers 0-6 combined)    → agencies, partners

Gotchas

Templates are elicitation tools, not final designs

The templates produce markdown documentation, not designed PDFs. The content is structured and correct; visual polish is a post-processing step. If you need a beautifully designed PDF, export the markdown and lay it out in InDesign/Figma.

Color accuracy in generated images

Generated images will approximate brand colors, not match them precisely. The generate swatch command provides hex values in the prompt, but image models interpret color imprecisely. For Pantone-accurate color reference, use the documented values in the spec, not generated images.

Reference-image-aware generation is setup-dependent

The generate script constructs prompts regardless of backend. To use reference images (logos, existing materials) for better results, you need a generation setup that supports image inputs — e.g., ComfyUI, local Stable Diffusion with img2img, or an API that accepts image references. The prompts work without this; they just improve with it.

Brand card sources from other templates

The brand-card template uses source: frontmatter links to strategy, visual-id, and voice templates. This means: compile may show placeholder text if those source templates aren't filled yet. Fill strategy + visual-id before expecting a complete brand card.

The skill produces documents, not art

The image generation capability exists, but the primary output is structured documentation — markdown that can live in a vault, be versioned in git, and be compiled into various formats. The images are the garnish, not the meal.

See Also

• nous-branding — Reference-image generation for Nous Research brand
• cli-builder — Patterns used by the brand-book CLI
• references/image-generation.md — Detailed prompt craft guide

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: magnus919
• Source: magnus919/agent-skills
• License: MIT

Install and usage instructions live in the source repository linked above.