Image generate

Skill m-binimran/design-pack/skills/image-generate

Connector-first Claude Code pack for graphic designers: Figma, Canva, Adobe Firefly, Nano Banana, Higgsfield, Unsplash, Brand.dev + skills, design guardrails (contrast/licensing/asset hygiene), loops & review agents.

Install

npx -y skills add m-binimran/design-pack --skill image-generate

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

One thing to look at

3 stars3 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

Generate images with an AI connector (Nano Banana / Gemini, Higgsfield Soul, or Firefly) using a strong prompt framework, the right model, aspect ratio, and iteration. Use when creating original imagery, concepts, or hero art.

SKILL.md

1.5 KB, as published. Nobody here has run it

image-generate

Pick the model, write a real prompt, generate, curate, iterate. Record the license.

Pick the connector/model

Nano Banana (Gemini): fast, multi-reference composition (up to 14), editing, up to 4K - great default.
Higgsfield Soul / Flux / Seedream: photoreal, character consistency, cinematic looks.
Adobe Firefly: when you need commercially-safe training + Adobe ecosystem. (See mcp/CONNECTORS.md. If none is connected, say so - don't fake an image.)

Prompt framework

Subject + action + setting + style + composition + lighting + colour + mood + camera/lens + quality. State aspect ratio and any reference images. Be specific; vague prompts give generic output.

Process

Generate a small set (2-4 variations). 2. Curate the strongest. 3. Iterate the prompt (fix hands/text/ artifacts, adjust comp/colour). 4. Upscale / set final aspect ratio. 5. Record model + license.

Output

The chosen image(s) from the connector (confirm they returned), the final prompt, and the model + license note.

Guardrails

Don't describe an image as generated unless the connector returned it (truth-telling + review-gate).
Record AI source + usage rights (the license-guard hook). Curate and fix - don't ship raw first output.