This server has a single tool (extract_site_links) with critical definition gaps. The tool name is clear and action-oriented (extract_*), but the input schema is severely underdocumented. The input parameter 'url' has a type (string) and a basic description in Japanese, but lacks validation constraints (URL format, length limits, protocol requirements). No output schema is documented anywhere in the visible code, we cannot see what structure is returned or what fields the LLM should expect. The tool description is in Japanese ('公式サイトからheader/footer/navのリンクを抽出し、仮想サイトマップを作成する') and provides intent but is vague on what exactly is returned and when to use this tool vs. alternatives. The code shows heavy scraping logic (BeautifulSoup, Playwright, aiohttp) but the tool definition does not expose this complexity or document constraints (timeout, max page size, SSL handling, retry behavior). No error recovery guidance is visible, if a URL is unreachable or scraping fails, the LLM has no hint what to do next. The visible code includes extensive utility functions for Rakuraku/Google Sheets/Cloud Gym integrations, but none of these appear as registered MCP tools, suggesting they are either helper functions or incomplete tool definitions. Only extract_site_links is registered; the others are inferred from code presence, which violates the 'explicit registration' requirement.

• Document the output schema explicitly. Example: 'Returns an object with {"links": [{"url": string, "text": string, "section": "header|nav|footer"}], "total_links": number, "extraction_time_ms": number, "warnings": string[]}'. Include field types and meanings.
• Expand the tool description to 100-150 characters in English: 'Extracts all hyperlinks from a website header, navigation menu, and footer to build a virtual sitemap. Useful for understanding site structure and discovering related pages. Requires a publicly accessible URL. Returns links organized by section.'
• Add input validation constraints to 'url' parameter: 'Must be a valid HTTP or HTTPS URL (e.g., https://example.com). URLs must be publicly accessible. Max length 2048 chars. Server will timeout after 30 seconds.'
• Add error recovery hints to the description: 'If the URL is unreachable, verify it is correct and the site is online. If scraping times out on large sites, try a smaller subset (e.g., homepage only) by passing a specific page path.'
• If Rakuraku, Cloud Gym, and Google Sheets integration functions are meant to be tools, register them as separate MCP tools with full schema documentation and English descriptions. If they are helpers, remove them from this server or clearly document why they exist but are not exposed.
• Add parameter descriptions in English. Current description for 'url' is in Japanese, ensure all parameter descriptions are in English for global LLM compatibility.
• Document behavior for edge cases: what happens with pages that require authentication, have JavaScript-heavy content, use custom navigation structures, or have no header/footer/nav elements. Return a 'skipped_sections' field if some are missing.

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry