デジタル庁デザインシステム (DADS) の公開情報をローカル stdio 型 MCP サーバーとして提供する非公式 TypeScript 実装
Server has well-structured tool definitions with consistent naming (verb_noun pattern), complete input schemas using Zod validation, and descriptions in Japanese. However, output schemas are not formally documented, error handling lacks recovery guidance, and parameter descriptions could be more detailed about constraints and use cases. All 7 tools follow the same quality baseline: clear names, present descriptions (avg ~60 chars), and validated inputs. No critical gaps, but several medium-impact improvements needed for production readiness.
DADSのカラートークン一覧を取得します
指定コンポーネントの仕様を取得します
指定セクションのガイドライン詳細を取得します
スペーシングトークン一覧を取得します
タイポグラフィ仕様を取得します
デジタル庁デザインシステム(DADS)のガイドラインをキーワード検索します
2色のコントラスト比をWCAG基準で検証します
Output schemas not documented. Tools return text content but LLMs cannot infer structure of returned data (e.g., does get_color_tokens return a table, JSON, or markdown?). This forces LLMs to parse unstructured output.
Error handling lacks recovery guidance. errorResult() returns error codes (E_GUIDELINE_NOT_FOUND, E_COMPONENT_NOT_FOUND) but does not suggest next steps. E.g., 'Section not found' should suggest 'Try search_guidelines() with a keyword.'
Parameter descriptions lack constraint details. E.g., 'sectionId' description says 'セクションID (例: foundations-color-system)' but does not explain the pattern rule (^[a-z0-9-]+$) in the description text itself. LLMs cannot read JSON Schema patterns.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 58 | <=2025-11-25 | v2 |
get_spacing_tokens has empty input schema (no parameters). Description does not explain what data is returned or when to call it. Minimal context for LLM selection.
No pagination or result limits documented. search_guidelines could return many results; no limit parameter or guidance on max results. Large unstructured responses waste tokens and degrade LLM reasoning.