| name | hero-image |
| description | Generates Midjourney prompt options for blog post hero images on dacharycarey.com. Analyzes the article's thesis and recommends visual metaphors matching the site's established aesthetic. Use when a post needs a hero image.
|
| compatibility | Designed for Claude Code (or similar products) |
| allowed-tools | Read Glob |
| metadata | {"author":"dacharycarey","version":"1.0"} |
Hero Image Prompt Generator
You are generating Midjourney prompt options for hero images on dacharycarey.com, a personal blog covering AI, agent ecosystems, documentation, and software development.
Step 1: Read the article
If the user provided a file path, read it. Otherwise, ask which post needs a hero image.
Read the full article. You need to understand the thesis, central metaphor, and emotional register to generate effective prompts.
Step 2: Read recent hero images for visual context
Read 3-5 recent hero images from static/images/ to calibrate against the current visual style. Choose images from posts close in date to the target article. The images are viewable directly (they are image files).
Step 3: Identify the core visual concept
Before writing any prompts, identify:
- The article's central tension or thesis — what is the piece about at a conceptual level?
- A physical metaphor — what real-world object, scene, or situation captures that concept? The best hero images on this site are conceptual/metaphorical, not literal illustrations of the topic.
- The emotional register — is the article investigative? Cautionary? Optimistic? Frustrated? The image should match.
State these three things to the user before generating prompts.
Step 4: Generate 3-5 prompt options
Generate 3-5 Midjourney prompts that represent different visual approaches to the concept. Vary the approaches:
- At least one photorealistic scene (objects on a desk, an environment, a person in silhouette)
- At least one conceptual/miniature (the yarn-ball style — a single striking metaphor against a clean background)
- At least one split composition or diptych if the article has a contrast or before/after structure
Visual style rules for this site
These are derived from the established hero image aesthetic on dacharycarey.com:
Composition:
- Objects as protagonists more often than people
- When people appear, they are in profile, silhouette, or from behind — secondary to the object/concept
- Shallow depth of field in photorealistic scenes
- Clean negative space in conceptual/miniature compositions
Lighting and color:
- Moody, warm lighting. Amber and golden tones dominate.
- Dramatic shadows, cinematic quality
- Muted, desaturated palettes with one warm accent color (amber glow, red markup, golden light, purple UV)
- Dark or neutral backgrounds preferred
Content:
- No text, words, letters, or labels in the image
- No UI mockups or screenshots
- No stock-photo energy (no smiling people at laptops, no handshakes, no abstract geometric patterns)
- No overt AI imagery (no glowing brains, no neural networks, no circuit boards unless the article is specifically about hardware)
- Don't overuse the magnifying glass as metaphor
What works on this site (reference examples):
- Redlined document with scissors on a moody desk (editorial evaluation)
- Cobwebbed filing cabinet with warm glow (staleness/decay)
- Child pulling a massive yarn ball (complexity vs. human scale)
- Split-screen desk: mystical/alchemical vs. scientific/microscope (vibes vs. data)
- Conveyor belt of news items funneling into a glowing tornado (information processing)
- Woman with magnifying glass examining a data wall (investigation/scrutiny)
Midjourney flags
Every prompt must end with:
--ar 4:3 --no text, words, letters, labels --s 50
The site uses --ar 4:3 due to the theme implementation, so images must work at that size. We can omit the no text flag if there is a single word that may generate well, but overall, should avoid it, as Midjourney is still bad at words.
Step 5: Present options
Present each prompt with:
- A short title (2-4 words describing the visual concept)
- The full Midjourney prompt
- Why it works (one sentence connecting the visual to the article's thesis)
Let the user choose. If they want variations on a specific option, generate 2-3 riffs on that concept with different compositions, lighting, or metaphor angles.
Step 6: Update front matter
Once the user has generated their image and saved it to static/images/, update the post's image: field in the front matter to point to the new file. The naming convention is:
/images/<slug>-hero.jpg
Where <slug> matches the post's URL slug (e.g., ai-content-pipelines-verification-gap-hero.jpg).
