codefox-repo/brand-studio-codefox/banner-banner

Banner

A read-only, instruction-grade producer. The agent reads this as procedure; the configured text-to-image backend (or inline SVG/HTML) renders the pixels.

Inputs

• Structured brand weight: palette, typography direction, references, avoid list,

and the domain's accepted samples.

• Deliverable spec: id + size. Common banner sizes (align to multiples of 16 for

gpt-image): GitHub social preview 1280×640, X header 1500×500, LinkedIn cover 1584×396, YouTube channel art safe area ~1546×423, OG image 1200×630.

Output

• Generator path: produce one tight prompt that places the wordmark/headline

in a safe area (never cropped), keeps the brand palette/typography, respects the avoid list, and frames the product as the subject — then hand it to the text-to-image backend.

• Exact-text path: when the banner is mostly a wordmark + tagline on a brand

field, prefer a deterministic SVG/HTML lockup so text is pixel-crisp, using the brand palette and type direction directly.

Conventions, not technique

Borrow layout/hierarchy/spacing conventions from referenced design skills; do not copy any specific skill's implementation unless the user asks. Keep headlines off the frame edge, maintain generous safe margins, and never invent sponsor/brand marks not in the brand.