This server presents 18 tools with significant definition quality gaps. While 3 tools (semantic_search_tool, emotion_timeline_tool, topic_clusters_ml_tool) have comprehensive schemas and descriptions meeting A-grade baselines, the remaining 15 tools have critical deficiencies: 12 tools are defined in archived/old files with no visibility into actual schema implementations, and 3 consent/health tools lack input schema documentation. Naming is inconsistent, tools mix verb_noun (semantic_search_tool, imsg_health_check) with noun-heavy patterns (imsg_relationship_intelligence, imsg_anomaly_scan). Only the 3 ML tools in active src/ have complete parameter descriptions; the others are inferred from archive files with no schema evidence. Error handling is absent across all tools, no recovery guidance, categorization, or actionable messages. Output schemas are undocumented for all 18 tools. The server lacks composition patterns: no clear tool chain design, no pagination for list-like tools, and no batch variants for repeated operations.

• Move all 15 archived tools from archive/old_main_versions/main_complete.py into active src/ modules with explicit FastMCP tool registration. Ensure each tool decorator includes name, description, and complete input_schema with types and descriptions for all parameters.
• Document output schemas for all 18 tools. Use Pydantic models or JSON Schema to define return types. Example for semantic_search_tool: return {results: [{message_id, score, timestamp, redacted_preview}], total_count, next_cursor}.
• Add error handling with recovery guidance. For each tool, document common failure modes and next steps. Example: 'Database not found: ensure ~/Library/Messages/chat.db exists. Or pass db_path=/alternate/path.db'. Categorize errors as retryable (timeout) vs user-fixable (bad parameter) vs fatal (permission denied).
• Clarify contact_id contract: document that it is a hashed identifier returned by imsg_contact_resolve. In imsg_contact_resolve, return both display_name and hashed_contact_id so downstream tools can use the hash without ambiguity.
• Add pagination to list-like tools: imsg_conversation_topics, imsg_cadence_calendar, imsg_network_intelligence should accept limit (1-100), offset, and return {items, total_count, next_offset}.
• Rename analysis tools to use action verbs: emotion_timeline_tool → analyze_emotion_timeline, imsg_sentiment_evolution → track_sentiment_evolution, imsg_response_time_distribution → analyze_response_time_distribution, imsg_cadence_calendar → get_cadence_heatmap, imsg_anomaly_scan → detect_communication_anomalies.