Skip to content

TheGoat395/Codex-Skills / skills/web-imagegen-prompting

web-imagegen-prompting

Verified

Write image prompts with crop/placement intent.

FileKindSize
SKILL.mdskill md1.3 KB
agents/openai.yamlconfig250 B
references/web-imagegen-prompting-guide.mdmarkdown3.4 KB

references/web-imagegen-prompting-guide.md

View on GitHub
# Web Imagegen Prompting Guide

## Purpose

- Make generated imagery usable as website media instead of vague atmospheric filler.

## Use When

- The project lacks strong visual assets and generated imagery is acceptable.
- A hero, product, venue, portrait, texture, or editorial media system needs art direction.
- The current AI image prompt creates generic, unusable, cropped, or low-specificity visuals.

## Do Not Use When

- Real product, person, venue, legal, medical, or portfolio evidence must be shown accurately.
- The user has supplied brand-approved assets that should be used instead.
- The page can be stronger with typography, layout, or real screenshots rather than generated imagery.

## Discovery Questions

- What should the image prove or reveal on the page?
- What is the subject, setting, material, camera angle, crop, lighting, and mood?
- Which viewport crops must work: wide desktop, laptop, tablet, mobile, social preview?
- What must not appear: fake logos, incorrect product details, text in image, extra hands, stock-photo cues?

## Decision Tree

- Define the asset role before writing a prompt.
- Name subject, setting, camera, lens feel, lighting, material, composition, palette, and negative constraints.
- Request enough margin around the subject for responsive cropping.
- Generate variants by direction, not random style words.
- Inspect the output in the actual layout before accepting it.

## Implementation Rules

- Do not use vague words like premium or cinematic without specifying visible choices.
- Avoid text, logos, UI labels, or exact product details inside generated images unless they can be corrected.
- Prefer real screenshots, real photography, or supplied assets for factual proof.
- Make the image composition serve the page hierarchy and text placement.
- Keep prompts brand-specific and avoid generic stock-photo language.

## Useful Patterns

- Hero prompt with subject left/right negative space for overlaid type.
- Texture prompt for subtle material surface, not attention-grabbing decoration.
- Portrait prompt with environmental context and editorial crop.
- Product scene prompt with real usage context and inspectable object.

## Anti-Patterns

- Dark blurred background image with no subject.
- Generated luxury room that could belong to any brand.
- Image has text that looks broken or fake.
- Prompt asks only for 'modern premium cinematic website hero'.

## QA Checklist

- Place the image in the page and inspect desktop/mobile crops.
- Check for unwanted text, artifacts, distorted anatomy, fake UI, or wrong product details.
- Confirm contrast behind overlay text.
- Add accurate alt text or mark decorative as empty alt.

## Acceptance Criteria

- The generated asset has a clear subject and page role.
- It survives responsive cropping and supports readable typography.
- It feels specific to the brand/site rather than generic AI ambience.

## Scoped execution

Apply the [shared web contract](../../website-operating-rules/references/scoped-web-contract.md) once when execution crosses phases; reuse it if already read. Load only the specialist guidance relevant to the requested surface and deliverable.

## Official Source Anchors

- MDN img element: https://developer.mozilla.org/en-US/docs/Web/HTML/Reference/Elements/img
- MDN responsive images: https://developer.mozilla.org/en-US/docs/Web/HTML/Guides/Responsive_images
- W3C WAI alt decision tree: https://www.w3.org/WAI/tutorials/images/decision-tree/
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 322kUpdated Claude CodeCodexVerifiedInstall