agentleFS
omidnwomidnw/my-skills/skills/ai-image-prompts/SKILL.md

ai-image-prompts

Skill0 starsChanged 13 days agoMITView on GitHub

Write production-ready image or short-video briefs for UI assets and image-brief files. Use for prompt authoring and asset direction; direct image generation belongs to the available image-generation capability, with this skill as a brief companion.

The file

We split this skill into the rules it gives an agent

11 rules. About 0 are common, 9 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈9

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Read the current product design direction and the relevant UI surface.

    line 26
  2. Distinctive:

    Treat generated media as optional enhancement: retain a local CSS/SVG fallback and never depend on a runtime external image URL.

    line 28
  3. Distinctive:

    Never ask the model to bake interface copy, fake data, or a logo into an image.

    line 30
  4. Distinctive:

    Check the existing brief: Search the image brief before creating a new prompt.

    line 37
  5. Distinctive:

    Write the prompt: Use concrete subject, composition, lighting, palette, negative space, and negative constraints.

    line 41
  6. Very specific:

    Add a stable asset id: Use image1, image2, video1, etc., and preserve existing numbering.

    line 44
  7. Distinctive:

    Review the delivered asset: Check for text/logos, artifacts, crop safety, focal-point clarity, contrast under both themes, and file size.

    line 47
  8. Distinctive:

    Integrate safely: Keep approved media local, preserve the fallback, use the existing tokenized overlay, and update the project's …

    line 49
  9. Distinctive:

    Prefer one strong subject and a restrained visual system over a collage.

    line 55
  10. Very specific:

    Use 16:9/21:9 for hero media, 3:2/4:3 for supporting cards, and 9:16 for portrait/social variants unless the surface says otherwise.

    line 58
  11. Distinctive:

    Do not replace a working CSS/SVG fallback with a remote or unapproved asset.

    line 64

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

AI Image & Video Prompts

Purpose

Turn a visual need into a precise, generation-ready brief that matches the product's design system. The agent writes the prompt; the user or an image model generates the asset. The agent then integrates only after the asset is received and reviewed.

When to use

  • A UI surface needs a new hero image, illustration, map treatment, product visual, or atmospheric background.
  • The user asks for an AI-generated image or a short video.
  • A new prompt must be added to an image-brief file.
  • A generated asset is being evaluated for crop safety, theme compatibility, accessibility, or offline packaging.

Preconditions

  • Read the current product design direction and the relevant UI surface.
  • Find and read the project's image brief before creating a parallel prompt.
  • Treat generated media as optional enhancement: retain a local CSS/SVG fallback and never depend on a runtime external image URL.
  • Never ask the model to bake interface copy, fake data, or a logo into an image.

Workflow

  1. Inspect the surface. Identify the route, component, current fallback, focal area, aspect ratio, and overlapping text.
  2. Check the existing brief. Search the image brief before creating a new prompt.
  3. Define the visual job. State purpose, composition, palette, negative space, theme behavior, aspect ratio, and output format.
  4. Write the prompt. Use concrete subject, composition, lighting, palette, negative space, and negative constraints. Keep interface text out of the generated pixels.
  5. Add a stable asset id. Use image1, image2, video1, etc., and preserve existing numbering.
  6. Generate or hand off according to the request. For a prompt-only request, deliver the brief. For requested image generation, use an available authorized image-generation capability; ask the user to generate only when that capability is unavailable or they prefer handoff. Do not send confidential references externally without disclosure authorization or claim an asset exists before delivery.
  7. Review the delivered asset. Check for text/logos, artifacts, crop safety, focal-point clarity, contrast under both themes, and file size.
  8. Integrate safely. Keep approved media local, preserve the fallback, use the existing tokenized overlay, and update the project's component/asset inventory when applicable.

Rules

  • Prefer one strong subject and a restrained visual system over a collage.
  • Reserve clean negative space for localized copy; do not generate text.
  • Describe the product palette instead of scattering arbitrary colors.
  • Use 16:9/21:9 for hero media, 3:2/4:3 for supporting cards, and 9:16 for portrait/social variants unless the surface says otherwise.
  • Short videos must be silent, under 10 seconds unless explicitly requested, stable, and designed to loop.
  • Generated visuals must not contain secrets, internal URLs, private formulas, or unverified business data.
  • Do not replace a working CSS/SVG fallback with a remote or unapproved asset.

Common mistakes

  • Writing a mood-only prompt with no subject placement or crop behavior.
  • Asking for UI text, fake charts, or a logo inside the image.
  • Treating an image prompt as permission to add a non-existent file.
  • Forgetting RTL-safe negative space or light/dark overlay requirements.
  • Generating a long cinematic video when the UI needs a quiet loop.

Validation

Validate only the phase delivered: prompt-only work does not require a received asset or integration build. Generation/integration must additionally pass the media and surface checks below; clearly identify pending phases.

  • The brief contains a stable id, prompt, aspect ratio, and use case.
  • The prompt has explicit negative constraints and no interface copy.
  • The delivered asset is local, optimized, and free of baked-in text/logos.
  • The surface still renders correctly with the fallback when the asset is unavailable, in both LTR and RTL directions.
  • The project's relevant typecheck/build checks pass after integration.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.