Single tool with critical gaps in description quality and schema documentation. The tool name 'search_vitepress_docs' is reasonably action-oriented (verb_noun convention), but the description is problematic: it contains prescriptive instructions ('Extract up to five keywords each English and native language, and define all of them as single words') rather than explaining what the tool does for the LLM. The input schema is minimal, only 'keywords' array with a basic description. No output schema is documented anywhere in the source code. The description is 177 characters, which falls within the acceptable baseline range (p10=34, p90=392), but the content is not LLM-optimized; it reads as user guidance rather than tool documentation. This violates the pattern:tool-description pattern which requires descriptions to explain WHAT the tool does, WHEN to use it, and any prerequisites. No parameter constraints (enums, min/max), no error handling guidance, no output structure definition.

• Rewrite the tool description to explain the tool's purpose: 'Search VitePress documentation for the specified keywords and return relevant sections with links. Useful when the user is looking for documentation on a feature, API method, or configuration option.' Remove prescriptive instructions about keyword extraction.
• Document the complete output schema in a comment or docstring. Example: 'Returns an array of objects with: { section_title: string, content: string, url: string, relevance_score: number }. Maximum 10 results per call.'
• Add constraints to the 'keywords' parameter schema: minimum array length (1), maximum array length (5 as the description claims), and optionally a pattern or description format. Add a note that keywords should be single words or short phrases.
• Implement and document error handling. Example responses: 'No documentation found for keywords: [list]. Try simpler keywords or check the table of contents.' Or if too many keywords: 'Maximum 5 keywords allowed; received 7.'
• Add a language parameter (if supported) with enum values: ['english', 'native'] or similar, rather than requiring the user to extract keywords in multiple languages.
• Verify and document the search behavior: does it return paginated results? Are there limits on result count? What does 'relevance' mean? This information must be in the description and output schema.

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

• Relies on Stateful initialize / Mcp-Session-Id (removed; protocol is stateless) - make each request self-contained