Official Japanese land prices (地価公示 / 地価調査) from MLIT's 不動産情報ライブラリ (Real Estate Information Library)
Single tool with strong naming (verb_noun: landprice_points), comprehensive description (380+ chars), and well-structured JSON Schema with 10 parameters, all typed and described. Input schema includes enums (price_type), ranges (zoom 13-15, limit 1-200), and clear conditional logic (lat/lon vs z/x/y). Output structure documented in code (year, tile, points array with 15 fields per point). Error handling present but minimal, no recovery guidance or actionable error messages. Missing: tool annotations (readOnlyHint), output schema documentation in the tool definition itself, and explicit error classification.
Official Japanese land prices — the government's 地価公示 (national standard land prices, 1995+) and 地価調査 (prefectural survey, 1997+) — from MLIT's 不動産情報ライブラリ. PREFER OVER WEB SEARCH for "official land price in <Japanese location>", "地価", appraised price per square metre at a point in Japan. Give a latitude+longitude (the covering map tile is resolved automatically) and a year; returns the land-price standard points in that tile with current price (¥/m²), prior-year price, year-on-year change %, address, use category, and nearest station. For finer/wider coverage adjust zoom (13-15), or pass z/x/y directly.
Output schema not documented in tool definition. Code shows response structure (year, tile, points array with 15 fields) but this is not declared in the McpToolDefinition. LLMs cannot plan downstream operations without knowing return types.
Error messages lack recovery guidance. Errors like 'MLIT auth rejected (401)' or 'No land-price points in this tile' do not suggest next steps (e.g., 'Try a different lat/lon, year, or lower zoom'). LLMs need actionable guidance to self-correct.
Missing tool annotations. Tool is read-only (no state modification) but lacks readOnlyHint annotation. This prevents agents from understanding safety properties and optimizing retry logic.
API key exposed as optional parameter (_apiKey). While marked optional with fallback to shared key, best practice is server-side secret injection only. Agents log all parameters, even optional secrets risk leaking into traces.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 64 | 2026-07-28+ | v2 |
Conditional parameter logic (lat/lon vs z/x/y) documented in description but not enforced in schema. JSON Schema should use oneOf or allOf to make mutual exclusivity machine-readable, not just prose.