| name | image-generation |
| description | Image-direction prompts for the content pipeline. Generate featured-image prompts, build stock image search queries, run the image-director reasoning that picks which sections get visuals (image_gen vs pexels, style, and prompt/query for each) plus one featured hero image, and caption rendered images with vision-derived alt text.
|
| license | Apache-2.0 |
| metadata | {"category":"image_generation","prompts":[{"key":"image.featured_image","output_format":"text","description":"Generate a featured-image prompt for a blog post topic in a chosen art style"},{"key":"image.inline_illustration","output_format":"text","description":"Generate an inline section-illustration prompt in a chosen art style"},{"key":"image.search_queries","output_format":"json","description":"Generate stock-image search queries for a topic"},{"key":"image.decision","output_format":"json","description":"Image-director reasoning prompt — picks which sections get visuals, image_gen vs pexels, style, and prompt/query for each, plus one featured hero image"},{"key":"image.caption_alt_text","output_format":"text","description":"Vision alt-text instruction — describe only what is actually visible in the image, one factual sentence under {budget} characters. Used by services/image_captioner.py (the caption_images stage + the alt-text backfill script)."}]} |
Image generation skill
The prompts the pipeline uses to direct visuals for a post — a featured
hero prompt, inline-section illustrations, stock search queries, the
image-director reasoning that decides which sections get images, and the
vision alt-text captioner that describes the rendered result. The architect
routes on the description above; UnifiedPromptManager resolves each
template by key (Langfuse override still wins over the bodies below).
Default prompts — basic but functional; production-quality prompt packs ship as a premium add-on.
image.featured_image
Write a single Stable Diffusion XL image prompt for a magazine-style editorial cover illustration.
Image subject: {subject}
Art style: {style} — {style_tags}
Depict a concrete, specific subject or scene that represents the image subject above (a recognizable object, place, or visual metaphor), rendered fully in the "{style}" art style. Commit to that style's medium, palette, and composition. Do NOT default to a generic glowing-circuit board or abstract floating-data backdrop, and do not lock every image to teal/cyan — vary the focal subject, composition, and color treatment so it reads differently from a typical tech illustration. Faceless silhouettes are fine; no identifiable faces, no hands, no text or words in the image.
Output ONLY the image prompt, 1-2 sentences, nothing else.
image.inline_illustration
Write a single Stable Diffusion XL image prompt for a blog section illustration.
Section subject: {search_query}
Art style: {style}
Depict a specific, concrete scene for the section subject, rendered fully in the "{style}" art style — commit to that style's medium and palette rather than a generic tech render or screenshot. No people, no identifiable faces, no hands, no text or words. Vary the composition so it doesn't look like every other section image.
Output ONLY the image prompt, 1 sentence, nothing else.
image.search_queries
Generate 3 image search queries as JSON array for the topic: {topic}
image.decision
You are an image director for a {category} content site. Analyze this article and decide what images would make it more engaging.
ARTICLE TOPIC: {topic}
CATEGORY: {category}
SECTIONS (title + a short excerpt of the actual text):
{section_list}
AVAILABLE IMAGE SOURCES:
- "image_gen": AI-generated images. Best for: abstract concepts, mood imagery, artistic visualizations, conceptual scenes. Styles: blueprint, dramatic, minimal, isometric, macro, editorial.
- "pexels": Stock photography. Best for: real-world objects, environments, workspaces, hardware close-ups, materials. Avoid shots of people.
RULES:
1. Pick {max_images} sections that would benefit most from a visual (skip sections that are mostly code)
2. For each, decide: image_gen or pexels? What style? What specific image?
3. Also decide on 1 featured image (the hero/header image for the article)
4. Be specific in your prompts — describe the exact scene, not vague concepts
5. NEVER depict people, hands, faces, or human figures in ANY image — the brand style is objects, hardware, and environments only. Also never put text, words, or letters in AI-generated images — and never ask for a "diagram" or "chart" specifically, since diffusion models render those with garbled fake labels; describe the underlying object or scene instead.
6. Ground each image's subject in the section's excerpt — depict what that section actually discusses, not a generic {category} image.
Output ONLY valid JSON (no markdown, no explanation):
{{
"featured": {{
"source": "image_gen" or "pexels",
"style": "style_name",
"prompt": "detailed image prompt or search query",
"reasoning": "why this image works for the hero"
}},
"inline": [
{{
"section": "exact section title",
"source": "image_gen" or "pexels",
"style": "style_name",
"prompt": "detailed image prompt or search query",
"reasoning": "why this visual helps this section"
}}
]
}}
image.caption_alt_text
Write alt text for this image. Describe ONLY what is actually visible — factual, concise, one sentence, under {budget} characters. Do NOT begin with 'image of' or 'photo of'. Do NOT invent details that aren't visible.
