← DesignlyCONTENT HISTORY

Update to Designly

Snapshot Sep 30, 2026 · 23:14 UTC · version 4.4.0

Collection source: not recorded for this historical snapshot.

WHAT CHANGED · RULE-BASED ANALYSIS

First saved snapshot

No earlier snapshot is available to establish a change.

Compare saved observations

Download comparison JSON
Full technical diff · 0 changed fields
Full snapshot data
{
  "name": "taste-engine",
  "description": "Taste extraction and transferable visual rule mixing engine. This skill should be used when deconstructing visual reference images into transferable design principles, building Taste Profiles, mixing multiple references by assigned design jobs, or preventing plagiarism and AI-slop anti-patterns.",
  "included_files": [
    {
      "relative_path": "agents/openai.yaml",
      "size_in_bytes": 393
    },
    {
      "relative_path": "assets/large-logo.svg",
      "size_in_bytes": 773
    },
    {
      "relative_path": "assets/small-logo.svg",
      "size_in_bytes": 773
    },
    {
      "relative_path": "assets/taste-mix.template.json",
      "size_in_bytes": 350
    },
    {
      "relative_path": "assets/taste-profile.template.json",
      "size_in_bytes": 2327
    },
    {
      "relative_path": "schemas/taste-mix.schema.json",
      "size_in_bytes": 1032
    },
    {
      "relative_path": "schemas/taste-profile.schema.json",
      "size_in_bytes": 4943
    },
    {
      "relative_path": "scripts/taste_lint.py",
      "size_in_bytes": 1810
    },
    {
      "relative_path": "scripts/taste_merge.py",
      "size_in_bytes": 2458
    },
    {
      "relative_path": "scripts/test_taste_lint.py",
      "size_in_bytes": 757
    },
    {
      "relative_path": "scripts/test_taste_merge.py",
      "size_in_bytes": 1512
    }
  ],
  "skill_md_contents": "---\nname: taste-engine\ndescription: Taste extraction and transferable visual rule mixing engine. This skill should be used when deconstructing visual reference images into transferable design principles, building Taste Profiles, mixing multiple references by assigned design jobs, or preventing plagiarism and AI-slop anti-patterns.\n---\n\n# Taste Engine\n\nTaste Engine converts raw reference images into evidence-backed, transferable design rules and structured Taste Profiles. It prevents literal copying by abstracting principles (lighting ratios, color harmonies, focal geometry, surface finishes) and assigning dedicated jobs to each reference.\n\n---\n\n## 1. Core Workflow\n\n1. **Reference Deconstruction**:\n   - Analyze reference image across 5 core dimensions:\n     - **Lighting**: Key angle, fill ratio, shadow softness, color temperature.\n     - **Palette**: Dominant, secondary, and accent ratios (e.g. 60-30-10 rule).\n     - **Texture & Finish**: Material response (matte, satin, metallic, specular).\n     - **Spatial Geometry**: Grid anchors, visual weight distribution, negative space.\n     - **Mood & Tension**: Atmosphere, grain, contrast level.\n\n2. **The 4-Step Rule Extraction Pipeline**:\n   - `Evidence` (what is physically visible in the image)\n   - `Observation` (structural or aesthetic pattern)\n   - `Transferable Rule` (generalizable principle applicable to new briefs)\n   - `Constraint / Anti-Rule` (what to strictly avoid to prevent unintended slop)\n\n3. **Job-Based Reference Mixing**:\n   - Never blend references indiscriminately.\n   - Assign distinct, non-overlapping design jobs:\n     - Reference A -> **Lighting Job** (e.g. 45° soft directional rim lighting).\n     - Reference B -> **Composition Job** (e.g. dynamic diagonal off-center anchor).\n     - Reference C -> **Color Harmony Job** (e.g. muted warm earth tones).\n\n4. **Originality & Anti-Plagiarism Guard**:\n   - Strictly discard subject matter, character identity, and proprietary artistic trademarks from the reference.\n   - Never copy unique composite artwork verbatim.\n\n5. **Output Contract**:\n   - Emits structured `TasteProfile` JSON and `DesignSignalPacket` containing `taste_state`.\n\n---\n\n## 2. Tools & Scripts\n\n- Lint Taste Profile: `python3 scripts/taste_lint.py <taste-profile.json>`\n- Merge References: `python3 scripts/taste_merge.py <taste-mix.json>`\n\n---\n\n## 3. Cross-Skill Neural Connections & References\n\n### Peer & Downstream Skills\n- [Reference Memory](../reference-memory/SKILL.md) — Persistent REF-ID lookup and taste history\n- [Brand Intelligence](../brand-intelligence/SKILL.md) — Checking reference rules against brand locks\n- [Composition Director](../composition-director/SKILL.md) — Applying reference geometric anchors to grid\n- [Photography Director](../photography-director/SKILL.md) — Translating lighting extractions into optics\n- [Campaign Canon](../campaign-canon/SKILL.md) — Cross-referencing visual style with canon cases\n- [Designly Director](../designly-director/SKILL.md) — Orchestrator and conflict resolver\n\n### Schemas & References\n- [Taste Profile Schema](schemas/taste-profile.schema.json) — Local profile schema\n- [Taste Mix Schema](schemas/taste-mix.schema.json) — Mixing contract schema\n- [Taste Engine Guide](../../shared/references/taste-engine.md) — Extraction principles\n- [Reference Analysis](../../shared/references/reference-analysis.md) — Analytical steps\n"
}

SHA-256: 5d174d0f45dcd458fca7b90ed8e88515ead6facdd5d8b16de75a3ad26f6a4e18