layer-store-assets

Layer Store Assets

Overview

Store art is seen small, beside competitors, and briefly. Each asset has one job at its smallest display size: the icon is recognisable, the capsule shows the logo, the screenshots say what the player does. An asset judged at full resolution on a large monitor has not been judged.

The second rule is consistency. A listing is a set: icon, capsules, and screenshots must read as the same game. Derive every size from one approved master instead of generating each one, because each fresh generation moves the character and shifts the palette.

Store sizes and rules change and differ by store. Take every exact pixel size, and every content rule, from the user or from the store's current published spec. Do not supply them from memory.

Mechanics and the spend gate are in the layer skill, prompt craft in layer-image, reframing and upscaling in layer-image-editing, exact type in layer-typography, and video assembly in layer-video-timeline.

If a sibling skill named here is missing from your available skills, ask the user to install it (npx skills add layerai/skills --skill <name>); unattended, proceed from tool schemas and flag the gap.

Quick reference

Asset	Start from	Tools
Icon	One subject, square, no text	text_to_image with square, then image_editing
Icon variants	The approved icon as init_image	image_editing, one change per variant
Capsule or feature	One wide key art master	reframing to each size, compose_image for the logo
Screenshot set	The user's real gameplay captures	render_text and compose_image, one shared layout
Preview video	Captured gameplay clips	render_video_timeline
Size check	Any finished asset	compose_image downscaled to its display size

Icons

One subject, centred, filling most of the square, against a simple background with strong contrast to it. The usual subject is the game's hero or its signature object, cropped close. Leave out text, small props, and fine detail: none of it survives the smallest size the store displays, and type on an icon reads as noise.

Filter list_base_models with use_case: "text_to_image" and capabilities: {square: true}, adding image_editing: true when the hero must match approved art attached as reference_image.

Variants for an icon test start from the approved icon as init_image on use_case: "image_editing", and change one thing each: background colour, expression, or a seasonal prop. State what stays fixed ("keep the character, pose, framing and outline exactly as they are"). Generate every variant from the approved icon, never from the previous variant, or drift compounds.

Judge at display size

compose_image spends no Creative Units and can shrink anything: a canvas of the display size, one layer with the asset, both edges anchored at 0, and resize_strategy: "fit_and_pad". Look at that output. If the hero's silhouette or the logo is unreadable there, the fix is fewer elements and heavier shapes in the master, not a sharper upscale.

Capsules, headers, and feature graphics

Approve one master key art wider and larger than the largest target, with the subject composed away from the area the logo will occupy. Derive each size from it with use_case: "reframing", the master as init_image, and both width and height set; with either missing and an init image attached, the output takes the master's shape. If the master is too small for the largest size, upscale it (use_case: "upscaling") rather than regenerating.

The logo is the user's file, uploaded once with request_file_upload_url, never generated: models degrade lettering. Place it with compose_image in a box sized for that capsule using fit_and_pad, key art underneath with fit_and_crop. On the smallest capsule the logo is the asset, so give it most of the frame and check it at display size.

Screenshot sets

Screenshots show the real game. Use the user's captures, uploaded once, and put generated or designed art in the frame around them, not in place of them. Check the store's current policy on what a screenshot may show.

A set reads as a set when every frame shares one layout: the same caption band position and height, the same font, colour and size, the same background treatment. Fix that layout once and reuse the same compose_image geometry for every frame:

1. A background plate (generated from the key art, or a flat colour) with fit_and_crop.
2. The capture in a fixed box with fit_and_pad.
3. The caption from render_text with the brand font as font_file_id, on a canvas the size of the caption band, so it is barely scaled when placed.

One caption per frame, a few words, each naming what the player does. Order the set so the first frames carry the pitch, since they are the ones every visitor sees.

Preview video

Build it from captured gameplay with render_video_timeline: short shots, the core loop in the opening seconds, captions inside the safe area, and the logo at the end. Whether generated footage may appear in a preview is store policy; ask before adding any.

Worked example

"We have key art and a logo. Make our Steam capsules and a small set of icon variants."

1. The key art carries a file_id; upload the logo with request_file_upload_url.
2. Ask for the exact capsule sizes from the store's current spec, or take them from the user.
3. list_base_models with use_case: "reframing"; one run per capsule size, key art as init_image, explicit width and height.
4. compose_image per capsule: reframed art with fit_and_crop, logo in its box with fit_and_pad. Then shrink the smallest capsule to its display size and check the logo reads.
5. Icon variants: use_case: "image_editing", approved icon as init_image, three variants that change only the background colour.
6. Before step 3, estimate one reframe per distinct size and one icon edit times three, present the total, and wait for an explicit yes above 20 CUs. Poll every run at poll_interval_seconds.

Common mistakes

• Judging an icon or a capsule at its generated size instead of its display size.
• Text or fine detail on an icon.
• Generating each capsule size separately instead of reframing one master.
• Passing only width or height when reframing.
• Generating the logo instead of compositing the user's file.
• Screenshot captions that change position, font, or size from frame to frame.
• Generated scenes presented as gameplay screenshots.
• Pixel sizes or store rules stated from memory.
• Building icon variants from the last variant instead of the approved icon.