style

# style — design isms Query the visual and technical knowledge of 49 design isms. The same dataset that renders the site is the source of truth — read the JSON directly, never copy it. ## Executable query route When the `design-isms` MCP is attached, use its single `execute_code` tool. Discover operations with `actions.find()` / `actions.describe(name)`; `design.search`, `design.get`, `design.recipes`, `design.compose`, and `design.brief` return data. Request only the needed view (`summary`, `guide`, `code`, `full`), preserve search version/completeness/cursor when projecting, and never present clipped code as complete. Composition is a value; pass it unchanged to `design.brief({composition})` for Markdown. The Node CLI `scripts/design-query.mjs` provides the same operations as NDJSON. If MCP is not attached, keep the direct JSON-reading route below; do not install or register a server automatically. See `docs/PLUGIN.md` for the local stdio configuration. ## Data location All paths are relative to the plugin root (`${CLAUDE_PLUGIN_ROOT}/assets/data/`). - `isms.json` — array of 49 isms. Fields: `id`, `name`, `nameKr`, `tagline`, `description`, `descriptionEn`, `history`, `keywords[]`, `palette[]` (hex), `examples[]` ({name,url}, usually 10), `images[]`, `prompts[]`. - `dev-guides.json` — object keyed by ism id (49 entries, 1:1 with isms.json): - `layout`: `grid`, `columns`, `gutter`, `margins`, `spacing`, `symmetry`, `geometry` - `typography`: `fontPairing`, `sizeHierarchy`, `lineHeight`, `letterSpacing`, `weightStrategy` - `color`: `usage`, `bgFg`, `contrast` - `motion`: `easing`, `duration`, `hover`, `scroll`, `transition` - `dos[]`, `donts[]`, `implementation` (object: `summary`, `components`, `build`, `checks`) - Companion catalogs: `color.json` (25 role-based palettes), `typography.json` (20 font pairings), `layout.json` (25 layout patterns), `motion.json` (20 motion recipes). Not keyed by ism id, but offer them when the user wants concrete patterns. ## Intent → Action | User intent | Read | Return | |-------------|------|--------| | "what is X" / "tell me about X ism" (match id, name, nameKr) | `isms.json` entry | tagline + description + history + palette[] + keywords[] + examples[] | | "X color palette" | `isms.json[id].palette` + `dev-guides.json[id].color` | hex values + `usage`/`bgFg`/`contrast` role assignment | | "X font pairing" | `dev-guides.json[id].typography` (+ `typography.json`) | `fontPairing` + `sizeHierarchy` + `lineHeight`/`letterSpacing`/`weightStrategy` | | "X grid / layout numbers" | `dev-guides.json[id].layout` | `grid`/`columns`/`gutter`/`margins`/`spacing`/`symmetry`/`geometry` | | "X motion" | `dev-guides.json[id].motion` | `easing`/`duration`/`hover`/`scroll`/`transition` | | "recommend a style for my project" (keywords, mood) | whole `isms.json` → match `keywords[]` + `tagline` | 3–5 candidates: id + nameKr + tagline + why it matches | | "compare A vs B" | both entries side by side | palette / typography / layout / motion diff | | "full build guide for X" | whole `dev-guides.json[id]` | layout + typography + color + motion + `dos`/`donts` + `implementation` | ## Matching rules - Ids are lowercase kebab-case (`minimalism`, `brutalism`, `art-deco`). Match `nameKr` for Korean queries and `name` for English ones, then use the real `id`. - Loading all 49 at once is fine. For recommendations weight `keywords[]` and `tagline` first; `description` and `history` are secondary signals. ## Output guidance - Give hex values inline with their role (background / body / accent), no code fence. - Present font pairings as real CSS `font-family` values. - Give grid and spacing as drop-in CSS values (`24px`, `5vw`, `repeat(12, 1fr)`). - Always cite the source (`isms.json[id]` / `dev-guides.json[id]`) in the response. - Reply in the user's language; keep field names, hex values, and CSS verbatim.