| name | image-generation-workflow |
| description | How to get images for landing pages, social posts, ad creative, email templates, and other visual deliverables. Two paths: founder-loop (queue prompts in `image-prompt-queue.md`, founder generates in Claude.ai/Midjourney) and API-loop (call `infrastructure/scripts/ generate-image.py` for automation, $5/day portfolio cap). Faceless rule applies absolutely. Use whenever your work product needs an image, illustration, or campaign visual.
|
Image Generation Workflow
Claude.ai's image generation requires human-in-the-loop. Midjourney requires Discord. Both can't be called from agent code directly. So the portfolio uses two paths depending on urgency and automation needs.
When to use
Trigger on any of these moments:
- Landing page hero needs a visual
- Email template needs a header/banner
- Social post needs a 1080×1080 or 9:16 asset
- Book cover or product cover needs design
- Newsletter needs an inline illustration
- Ad creative for paid testing (Phase 2+)
When NOT to use
- A logo / wordmark (that's Brand & Marketing's identity work, not Designer's image gen)
- A photograph of a real person (faceless rule prohibits)
- Stock-photo-replacement that requires paid licensing without Chief Accountant approval
- Image-heavy content where you should ship with placeholders first and add images last
Faceless rule (absolute, always)
❌ NEVER prompt for "founder" or "Alex Antoniou" or any real-name reference
❌ Photorealistic humans that resemble founder
❌ Founder's actual face/voice/likeness in any form
✅ Stylized human figures (silhouettes, illustrations, abstract figures)
✅ Photorealistic humans NOT resembling founder
✅ Animation, motion graphics, abstract art
✅ Document/archive aesthetic
✅ Vintage poster aesthetic
Path A: Founder-loop (preferred when not time-critical)
The founder runs prompts in Claude.ai or Midjourney; agents queue prompts in advance.
Workflow
- Designer drafts prompt following the format below
- Append to
companies/<co>/design/image-prompt-queue.md (one prompt per block)
- CEO surfaces in next Q4hr brief with count of new prompts
- CoS includes count in next FOUNDER_ACTIONS.md with a one-line summary
- Founder generates in Claude.ai or Midjourney (~5-10 min/batch)
- Founder drops images in
companies/<co>/design/assets/
- Designer wires images into pages, replacing styled placeholders
Prompt block format
## Prompt #N — <short label> — <YYYY-MM-DD>
**Intended use:** {landing page hero / email header / 1080×1080 social / book cover / etc.}
**Dimensions / aspect:** {16:9 1920×1080 / 1:1 1080×1080 / etc.}
**Tool target:** {Midjourney / Claude.ai / either}
**Brand notes:** {palette refs, typography overlay if any, tone}
**Faceless check:** {confirm: no human faces required, OR stylized only}
**Rights:** {confirm: no copyrighted IP, no celebrity likenesses unless stylized}
**Prompt:**
"{Full prompt text — direction, style, lighting, composition, mood, references, negative prompts}"
Path B: API-loop (when automation needed)
Use infrastructure/scripts/generate-image.py to call OpenAI / Grok / Anthropic APIs. Costs are tracked.
Usage
python3 /Applications/DrAntoniou\ Projects/AgentCompanies/infrastructure/scripts/generate-image.py \
--provider {openai|grok|anthropic} \
--prompt "Vintage 2016 campaign poster style, document archive aesthetic, sealed envelope" \
--out /Applications/DrAntoniou\ Projects/AgentCompanies/companies/concise/design/assets/trump-hero-1.png \
--company concise
The script:
- Loads
.env for the relevant API key
- Calls the API
- Writes the resulting image to
--out
- Appends a JSONL line to
infrastructure/scripts/.image-gen-costs.jsonl with: timestamp, provider, prompt-hash, $cost, company, output-path
Provider preference
Default order (subject to availability):
- OpenAI DALL-E 3 — most reliable, $0.040/standard 1024×1024 image
- Grok (xAI image gen) — fallback, similar cost profile
- Anthropic image gen — verify availability before each call; not yet GA at last check
Cost cap
Portfolio-wide image-gen API spend is capped at $5/day until founder raises. The script enforces this — if today's accumulated spend in .image-gen-costs.jsonl is already over $5, the script aborts with a clear error.
If you genuinely need to exceed: surface to CoS via your standup with cost projection. CoS escalates to founder.
Don't bottleneck on images
If a landing page needs an image you don't have, ship the page with a styled placeholder. Use:
- Gradient block with brand colors
- Brand-color rectangle with subtle pattern
- Simple SVG icon or geometric shape
Mark the slot in code with <!-- IMAGE QUEUED: prompt-N -->. Replace with real image once it arrives via Path A or Path B. Don't block deploy on imagery.
Quality checklist before queueing
Before adding to queue or running API:
Related skills
agent-autonomy-doctrine — when to act vs escalate (image-gen API cap is one such gate)goal-first-briefing-format — how to surface image-prompt-queue counts in CEO briefs
Severity
This is a revenue-adjacent skill. Most landing pages need at least one image. Most email templates need a header. Without a working image-gen workflow, the portfolio's revenue thesis stalls. Use this skill correctly = revenue moves; ignore it = pages ship with placeholders forever.
See BIBLE.md § 8 for the policy statement and shared/personas/designer.md for the canonical Designer doctrine.
