blog-image

Invoke the generate command with your topic, and the skill builds a 6-component prompt covering subject, lighting, and composition before calling gemini_generate_image. It returns the image path, alt text, and a coverImage frontmatter snippet at 16:9.

Run the setup_image_mcp.py script with your Google AI API key via environment variable or key file. It writes the MCP configuration to ~/.claude/settings.json by default, then verify with validate_image_setup.py.

OG and social card images use 16:9 at exactly 1200x630 pixels, which the skill enforces by calling set_aspect_ratio before generation and cropping with ImageMagick during post-processing.

IMAGE_SAFETY comes from a non-configurable server-side output filter that sometimes blocks benign prompts. The skill auto-rephrases with positive framing and retries up to three times before reporting the failure.

Yes, the edit command enhances your instruction and calls gemini_edit_image, or uses gemini_chat for multi-turn refinement. Editing preserves session style consistency and is preferred when the image is mostly correct.

When invoked directly, the skill tells you to run the setup command. When called internally from blog-write or blog-rewrite, it returns silently so the calling workflow continues with stock photos.