scenes-gathered-zine-v1-3 — Claude Code skill: photo to zine poster

What scenes-gathered-zine-v1-3 does

The skill takes one photo and builds a 3:5 poster from it. The photographic part stays truthful, while a larger illustration field grows out of shapes already present in the source. Dense foliage, branches, crowds and fine texture get compressed into a few large quiet masses instead of being traced.

Inside is a set of rules the agent follows. It first reads the photograph: 1–2 core subjects, spatial invariants, the dominant gesture, a visual-weight map, and one or two source shapes that can carry across photography, illustration and colour. Then it compiles an image-generation prompt field by field — canvas, attention geometry, photo-to-illustration allocation, abstraction map, chromatic structure, torn-paper handoff, micro-text. By default you get the generated image plus a short creative rationale; the final prompt comes only when you ask for it.

Default abstraction is medium: roughly 60–80% of small descriptive detail is dropped, repeated forms merge into single masses, and realistic shading gives way to flat ink, broken contour or halftone. One high-chroma hue works as compositional structure rather than decoration, and a visibly hand-torn fibrous edge marks where photography becomes paper. Lower the abstraction when a face, building or location would otherwise stop being recognisable — that is worth checking before the poster goes anywhere near print.