Critical issues across naming, descriptions, and schema quality. Tool naming lacks clarity (duplicate 'search_contacts' with conflicting schemas), descriptions contain prompt-injection vectors and hidden behavioral guidance, and parameter schemas are incomplete. The server demonstrates awareness of patterns (Zod validation, structured responses) but fails to apply them consistently. Most tools fall below baseline quality for production use.

• Rename notes-server's 'search_contacts' to 'search_contact_notes' to eliminate naming collision. This makes the distinction clear: one searches contact records, the other searches notes mentioning contacts.
• Rewrite search_contacts (contacts-server) description to remove hidden behavioral guidance. Replace with: 'Search for contacts by name, email, or company. Returns matching contact records with ID, name, email, and company. Does not include internal notes unless explicitly requested via a separate tool.'
• Document output schemas for all tools. For search_contacts, specify: returns {query, limit, results: [{id, name, email, company}], total: number}. For get_contact, specify which fields are always present vs conditional (notes only if include_notes=true).
• Add error handling guidance to tool descriptions. E.g., get_contact: 'Returns contact details or error if ID not found. If contact not found, try search_contacts() with the contact name.'
• Add per-parameter validation rules in descriptions. E.g., phone parameter: 'Phone number with country code (format: +<country>-<area>-<number>, e.g., +1-555-0100). Must start with + and contain only digits and hyphens.'