Three tools with comprehensive descriptions and well-structured schemas. All tools have clear, detailed descriptions (194-250 chars, above baseline). Input schemas use Zod validation with proper types and enums. However, output schemas are not documented, responses are JSON-stringified without field-level documentation. Parameter descriptions are strong but lack explicit format/constraint guidance in some cases (e.g., avg_turns_per_session has no min/max bounds). Error handling returns generic error text without recovery guidance. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite all three tools being read-only operations.

• Document output schemas for all three tools. Specify the structure of the returned persona object, assessment result, and upgrade guide (e.g., fields like 'acd_level', 'score_breakdown', 'gaps', 'implementation_steps'). This enables LLMs to extract and chain results.
• Add bounds to numeric parameters: avg_turns_per_session (min: 1, max: 100), and clarify that concurrent_users in interaction_context is the authoritative flag (remove redundancy).
• Enhance error handling with recovery guidance. Instead of 'Error generating persona: ...', return structured errors: {'error': 'invalid_interaction_context', 'message': 'session_type must be one of: single, multi-session, continuous', 'recovery': 'Check interaction_context.session_type'}.
• Add tool annotations to all three tools. Use readOnlyHint: true to signal that these tools do not modify state and are safe to call multiple times.
• Add format constraints to free-form text parameters. E.g., 'brand_voice (max 500 chars, plain text or markdown)', 'domain_knowledge (max 1000 chars, technical description)'.