---
name: mixmatter
description: >-
  MixMatter restores the original v1 visual runtime as the active authority: bold source-aware editorial print
  reconstruction using disassembly, aggressive recomposition, selective photography, visible halftone/duotone,
  flattened graphic fields, tactile torn-paper collage, source-adaptive color, modernist hierarchy, and strict
  source-text protection. Preserve semantic identity, not the original camera composition. Do not collapse unrelated
  images into one decorative template. Add no new text by default unless the user explicitly supplies exact wording.
---

# MixMatter — v1 Visual Baseline Restored

The canonical runtime lives at `skills/mixmatter/SKILL.md`.

Before performing a MixMatter transformation, read and follow:

1. `skills/mixmatter/SKILL.md` — authoritative runtime instructions.
2. `skills/mixmatter/references/mixmatter-v1.md` — canonical v1 master prompt.
3. `skills/mixmatter/references/quality-rubric.md` — canonical v1-style evaluation and regression guard.

The v2 research documents under `docs/` are retained for historical and research value, but they are not runtime authority for image transformation.

> Photography is source material, not sacred material.

> Reconstruct, do not decorate.

> Preserve semantic identity, not visual completeness.

> Keep the design intelligence consistent, not the decorative skin.

For vague requests such as “process this,” execute directly using the strongest source-specific v1 reconstruction. Do not default to a recurring blue/coral/botanical border treatment. Different source images should be allowed to produce materially different palettes, crops, fragmentation patterns, print states, and collage structures.

Source text remains protected: add zero new text by default, never invent or translate source wording, and use only exact user-supplied wording when added typography is explicitly requested.