hatch-pet

Provide a name, description, or reference images and the skill generates a base character, nine animation rows, and sixteen look directions. It then assembles the 8x11 atlas, runs QA, and packages it with spriteVersionNumber 2.

A v2 pet atlas is a 1536x2288 PNG or WebP image with an 8-column by 11-row grid of 192x208 cells. Rows 0-8 hold standard animation states and rows 9-10 hold 16 clockwise look directions.

Yes, the skill runs a brand discovery step that researches official sources for palette, tone, and mascot cues before generation. The resulting brief guides the pet design without copying logos or trademarked marks.

Existing 8x9 atlases are validated and extended with generated look-direction rows, while 8x11 atlases can have failed rows regenerated. Approved standard rows are preserved and the result is repackaged as v2.

All visual generation is delegated to the imagegen skill so it controls the built-in image path and fallback rules. The hatch-pet scripts only perform deterministic work like extraction, mirroring, assembly, and validation.

Effects must be attached to the pet silhouette, opaque, and free of chroma-key colors. Detached sparkles, shadows, motion lines, text, and floating symbols are rejected because they break transparency cleanup and component extraction.