Four tools with clear verb-noun naming (search_docs, call_api, inspect_spec, convert_spec). All have descriptions and input schemas. However, output schemas are undocumented, LLMs cannot predict return structure. Parameter descriptions are present but generic (e.g., 'File path or URL' lacks format/constraint details). No error handling guidance. Missing safety/permission documentation for destructive write operations. Tool composition is sound (search→call pattern), but output chaining requirements are not explicit.

• Document output schemas for all four tools. Example: search_docs returns {results: [{id, method, path, summary, parameters, responses}], total_count}. call_api returns {status_code, headers, body, elapsed_ms}.
• Add error recovery guidance to call_api: 'If the endpoint returns 4xx, check parameter names and types via search_docs. If 5xx, retry after 5 seconds. If timeout, the API may be overloaded.'
• Expand parameter descriptions with format constraints. E.g., 'input: File path (absolute or relative) or HTTPS URL to OpenAPI 3.0/3.1 JSON/YAML, Postman v2.1+ JSON, HAR 1.2, or GraphQL introspection JSON.'
• Add explicit state-modification language to convert_spec: 'WRITE: Generates a new MCP server directory with TypeScript source, SQLite index, and package.json. Overwrites existing output directory if present.'
• Document search_docs pagination: 'Returns up to `limit` results (default 5, max 20). If more endpoints match, include next_cursor in response for pagination. Total endpoint count returned as total_count field.'