mixmatter · diff

git:20260918.543b1d1 to git:20260930.7c220a9

15 added, 67 removed. Audit A to A.

---

name: mixmatter

description: >-

- Mixmatter 2.1 art-directs and reconstructs user-supplied imagery while preserving the established Mixmatter

- visual DNA. It reads semantic identity and visual structure before transformation, forms a source-specific direction,

- reconstructs through flattened editorial composition, source-aware cropping, selective halftone/duotone, graphic

- fields, controlled tactile collage, and strict source-text protection, then preserves successful decisions across

- revisions. Add no new text by default unless the user explicitly supplies exact wording or explicitly authorizes

- generated typography. v2.1.0.

+ 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 2.1

- The canonical skill implementation lives at `skills/mixmatter/SKILL.md`.

+ # MixMatter — v1 Visual Baseline Restored

- When this repository-level skill entry is loaded, **read and follow `skills/mixmatter/SKILL.md` as the authoritative instructions before performing a Mixmatter transformation**.

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

- Also use these canonical references when relevant:

+ Before performing a MixMatter transformation, read and follow:

- - `skills/mixmatter/references/mixmatter-v2-runtime.md` — art-direction, preservation, hierarchy, and revision behavior.

- - `skills/mixmatter/references/mixmatter-v1.md` — established Mixmatter visual reconstruction language. V2 must not replace or dilute it.

- - `skills/mixmatter/references/quality-rubric.md` — evaluation and regression guidance.

- - `skills/mixmatter/references/hierarchy-stabilization.md` — hierarchy formation for repetitive, sparse, diffuse, or ambiguous sources.

- - `docs/system/MIXMATTER_CORE_CONSTITUTION.md` — non-negotiable product identity.

- - `docs/system/MIXMATTER_VISUAL_GRAMMAR.md` — research-derived visual grammar.

+ 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.

- ## Non-negotiable identity

+ 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.

- > Every visible intervention requires a structural or semantic cause.

- > Expand Mixmatter's intelligence, not its aesthetic identity.

- ## Interaction summary

- Mixmatter should feel simple in conversation.

- - **Clear request:** execute directly.

- - **Vague request:** inspect the actual image, choose the strongest source-specific direction, and execute unless a real ambiguity would materially change the result.

- - **Revision:** preserve successful crop, locks, hierarchy, identity, and useful material decisions; change only the requested or diagnosed axis.

- - **Alternative:** return to the original source image rather than repeatedly transforming a previous result unless the user explicitly asks to build on that result.

- Direction, Structure, and Intensity remain useful internal concepts, but they are not dependent on a custom host UI. They can be inferred from natural language or supplied explicitly by the user.

- The user should not need to know internal concepts such as `READ → UNDERSTAND → PROTECT → DIRECT → RECONSTRUCT → MATERIALIZE → CRITIQUE → REVISE`.

- ## Hierarchy continuity

- Do not inherit weak source hierarchy by default. If the source is dominated by repeated similar subjects, diffuse attention, or substantial low-information fields, use `hierarchy-stabilization.md` to form a stronger dominant / secondary / quiet organization before reconstruction.

- Repeated fields should gain unequal visual roles without losing group identity. Large quiet fields should receive a compositional job rather than arbitrary filler. Do not use rigid percentage or object-count thresholds as universal rules.

- ## Visual continuity

- The established Mixmatter system remains the visual authority:

- - strongly flattened, surface-first editorial composition,

- - source-derived crop and scale shifts,

- - selective photographic retention,

- - visible but selective halftone and duotone,

- - graphic planes and silhouettes,

- - controlled cut/torn-paper collage,

- - compressed/interrupted depth,

- - active low-information fields,

- - contemporary print energy rather than generic retro styling.

- Do not turn Mixmatter into a generic photo editor, style marketplace, cinematic image model, or parameter-heavy design suite.

- ## Text continuity

- Default behavior remains conservative:

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

- - add zero new text unless the user explicitly supplies exact wording or explicitly authorizes generated copy,

- - preserve identity-critical source text when feasible,

- - never invent approximate source wording,

- - never translate source text by default,

- - never create bilingual duplicates by default,

- - if exact source text cannot be preserved reliably, obscure/crop it rather than hallucinate it.

+ 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.

- For the full system, always defer to `skills/mixmatter/SKILL.md` and its references.

+ 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.