Server has 3 tools with basic schemas and descriptions, but significant gaps limit production readiness. Tool naming is inconsistent (create-user vs get_github_repos vs create-random-user). Descriptions exist but are minimal (10-50 chars, below the 194-char baseline). Input schemas use Zod but lack parameter-level descriptions for most fields. The create-random-user tool uses deprecated sampling/createMessage pattern. Error handling is generic ('An error occurred') without recovery guidance. No output schemas documented. Tool annotations present but incomplete.

• Standardize tool names to verb_noun format: rename 'create-user' → 'create_user', 'create-random-user' → 'create_random_user' for consistency with 'get_github_repos'.
• Add parameter descriptions to create-user: 'name' (required, 1-100 chars), 'email' (required, valid email format), 'address' (optional, max 200 chars), 'phone' (optional, E.164 format).
• Replace generic error messages with actionable guidance: 'Failed to create user: email already exists. Try search_users(email) to find existing account.' Include the invalid value and constraint violated.
• Document output schemas in tool descriptions: 'Returns {id: number, name: string, email: string, address: string, phone: string}' for create-user; 'Returns array of {name, url, description, stars, language}' for get_github_repos.
• Remove server-side sampling call in create-random-user. Either accept random user data as a parameter, or move LLM integration to the client layer. Current pattern violates stateless request handling.

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

• Relies on Sampling (deprecated) - integrate directly with the LLM provider API