ABOUTME: Generate editorial cover images from article context using Gemini image generation

ABOUTME: Distills articles into hand-drawn sketch visual metaphors with ink stroke style

Cover Image Skill

Generate editorial cover images using Gemini's image generation. Distill article content into a visual metaphor rendered as a hand-drawn sketch.

Workflow

1. Understand the Article

Read the article/context provided by the user. Identify the core concept, tension, or insight that could become a visual metaphor.

2. Craft the Prompt

Build a prompt using this structure:

Black and white hand-drawn sketch, loose ink strokes, technical architectural drawing style.
[visual metaphor distilled from article context]. High contrast, minimalist, white background,
editorial tech illustration. --ar 16:9

Visual metaphor guidelines:

• Translate abstract concepts into concrete imagery (e.g., "microservices" -> "separate buildings connected by bridges")
• Prefer architectural/technical metaphors over literal depictions
• Keep the scene simple: one strong central image, not a collage
• Avoid text, labels, or UI elements in the image

3. Show Prompt for Approval

Present the crafted prompt to the user. Wait for confirmation before generating.

4. Generate

uv run ~/.claude/skills/_generate_image.py "<approved prompt>" --ar 16:9 -o cover.png

5. Report

Report the saved file path and suggest reviewing the image.

Example

Article about: "Why we migrated from monolith to microservices"

Prompt:

Black and white hand-drawn sketch, loose ink strokes, technical architectural drawing style.
A massive ancient stone building cracking apart, its fragments floating and reorganizing into
a constellation of small interconnected structures, linked by thin precise lines.
High contrast, minimalist, white background, editorial tech illustration. --ar 16:9

Notes

• Requires GEMINI_API_KEY or GOOGLE_API_KEY in environment
• Default output: cover.png in current directory
• Default aspect ratio: 16:9 (landscape, editorial)
• Uses shared script: ~/.claude/skills/_generate_image.py