/suede-image

name: suede-image
description: "Suede AI marketing image production: choosing generation, editing, templates, or real screenshots; prompts and layout specs; hero, social, banner, directory, and Open Graph images; product mockups from real UI; export sizing, compression, and image SEO (alt text, file names, dimensions, structured data). Use when the user needs a marketing image or an image-production workflow. NOT FOR: paid-ad creative systems (use suede-ad-creative), video (use suede-video), app-store screenshots and listing strategy (use suede-aso), or social post strategy (use suede-social)."
metadata:
  version: 2.1.0

Suede AI Marketing Image Production

Produce marketing imagery as a rights-aware, placement-specific system: choose the right production method, protect canonical brand assets, preserve real product truth, and verify the exported result. Use generation models and design tools to create efficient hero, social, mockup, banner, and preview workflows without fabricating interfaces or provenance.

Before Starting

**Check for product marketing context first:** If `.agents/product-marketing.md` exists (or `.claude/product-marketing.md`, or the legacy `product-marketing-context.md` filename, in older setups), read it before asking questions. Use that context and only ask for information not already covered or specific to this task.

Gather this context (ask if not provided):

1. Image Goal

• What type of image? (Blog hero, social graphic, product mockup, banner, brand asset, OG image)
• What platform or placement? (Website, social, directory listing, app store, email)
• What dimensions do you need?

2. Production Approach

• Do you have existing brand assets? (Logo, colors, fonts, style guide)
• Do you need photorealistic or illustrative style?
• Is this a one-off or a template for repeated use?

3. Technical Context

• Which image, browser, design, or local conversion tools are currently callable?
• What is the approved maximum cost and data-handling boundary?
• Do you need the image optimized for web performance?

Do not ask the user to paste API keys or secrets into the conversation.

---

Choosing Your Approach

First discover the current production surface. Inspect callable tools and connected accounts; do not assume a named model, provider, API, plugin, or design app is available. Then choose among these methods:

| Approach | Best For | Candidate surface | |----------|----------|-------------------| | **Generation** | Original concepts and scenes | A callable image-generation tool | | **Editing** | Authorized changes to supplied images | A callable editor with image-input support | | **Template design** | Brand-consistent recurring assets | An authorized design app or local template | | **Screenshot + overlay** | Truthful product showcases | Callable browser capture plus local layout | | **Licensed media** | Existing photography or illustration | User-owned library or verified license source |

---

AI Image Generation

Use generation only after the current tool and authority gates pass.

Capability and authority gate

1. Confirm a generation or editing tool is callable in the current session. 2. Check its current official documentation for model availability, accepted inputs, output sizes, editing/reference support, safety restrictions, retention, commercial-use terms, and pricing. Record the source and check date. 3. Confirm rights to every uploaded logo, screenshot, photo, font, and reference image. Do not upload confidential or personal material outside its approved boundary. 4. Calculate the maximum cost for the requested attempts and get explicit approval before using a paid account or exceeding an already approved budget. 5. Confirm whether the user's request authorizes generation only, editing of supplied files, overwriting a source, or publication. These are separate gates.

Provider names and model versions are volatile. Examples such as OpenAI, Google, Black Forest Labs, Ideogram, Midjourney, Recraft, and self-hosted diffusion are research candidates, not routing instructions or capability claims.

Selection criteria

• For text-heavy assets, prefer a deterministic overlay or design template; test

any verified in-image text capability before committing to it.

• For repeated brand work, prefer locked templates and approved assets over a

claimed consistency feature.

• For edits, use a tool whose current documentation and callable schema confirm

image input and the required edit mode.

• For vectors, require a real vector export and inspect its paths; a raster image

labeled as vector is not sufficient.

• For product UI, capture the live authorized interface rather than generating it.
• For volume, compare verified cost, rate limits, review time, and output quality

on a small test batch.

If no suitable renderer or editor is callable, deliver a production-ready prompt, layout spec, asset manifest, rights checklist, and export checklist. State clearly that no image was generated; do not route the user to an unavailable tool as though execution occurred. Use these exact headings:

## Prompt
<Subject + Setting + Style + Lighting + Composition + Technical, one block,
ready to paste; note any text that must be a deterministic overlay instead>

## Layout Spec
Canvas WxH + ratio | safe margins | focal point | text zones with max character
counts | logo placement and clear space | color values

## Asset Manifest
Asset | origin (owned / licensed / captured / to be generated) | rights basis
and holder | attribution or expiry | file path or "to source"

## Rights Checklist
- [ ] Every uploaded or referenced asset has a confirmed rights basis
- [ ] No real person, endorsement, or product interface is fabricated
- [ ] Brand mark is the approved file, unmodified, or omitted
- [ ] Paid-tool cost approved, or no paid tool used

## Export Checklist
Destination | dimensions | format | quality target | file