Two tools with reasonable naming and descriptions, but significant gaps in schema completeness and parameter documentation. send-email has a well-structured input schema with detailed parameter descriptions and helpful constraints (e.g., 'You MUST ask the user' for sensitive fields). list-audiences has no input schema visible and minimal documentation. Output schemas are not documented for either tool. Error handling and recovery guidance are absent. The server follows basic tool definition patterns but lacks production-grade rigor in schema completeness and error guidance.

• Document output schema for send-email: return {success: boolean, messageId: string, scheduledAt?: string, error?: string}. Include messageId so agents can track sent emails.
• Document output schema for list-audiences: return {audiences: [{id: string, name: string, createdAt: string}], total: number}. Include audience IDs for downstream tool calls.
• Add error handling guidance to send-email: 'If email validation fails, return {success: false, error: "Invalid recipient: must be a valid email address"}. If rate limited, return {success: false, error: "Rate limit exceeded. Retry after 60 seconds.", retryable: true}'.
• Expand list-audiences description to explain when to call it: 'Retrieves all audiences from Resend. Call this first if the user mentions an audience name but you need the audience ID for other tools.'
• Move agent-facing instructions ('You MUST ask the user') from tool descriptions to the agent system prompt. Tool descriptions should focus on WHAT the tool does, not HOW the agent should invoke it.