Server has 11 tools with complete input schemas and descriptions. Naming follows verb_noun convention (match_, filter_, compute_, grade_, build_, propagate_, query_, db_). Descriptions are detailed (100-300 chars) and explain purpose, alignment to Athena, and parameters. However, output schemas are NOT documented, tools return JSON but LLMs cannot see what fields to expect. Parameter descriptions are present but some lack format/constraint details (e.g., 'event_date' says 'YYYY-MM-DD' but no validation guidance). Error handling is minimal, no recovery guidance or categorization. No tool annotations (readOnlyHint/destructiveHint). Two tools (build_causal_chain, propagate_impact) are WRITE operations but lack confirmation/dry-run patterns.

• Document output schemas for all 11 tools. Include field names, types, and descriptions. E.g., match_event_symbols returns {symbols: string[], count: int, error?: string}.
• Add tool annotations: mark build_causal_chain and propagate_impact with destructiveHint=true; mark all READ tools with readOnlyHint=true; mark query_* tools with idempotentHint=true.
• Implement dry-run mode for WRITE tools. Add optional 'dry_run: boolean' parameter to build_causal_chain and propagate_impact; return preview without persisting when true.
• Enhance error responses with recovery guidance. E.g., 'event_title cannot be empty. Call match_event_symbols with a non-empty title.' Categorize errors: retryable (timeout, rate limit), user-fixable (invalid input), fatal (DB corruption).
• Add min/max constraints to numeric parameters. E.g., car_windows array elements must be integers in range [0, 30]; max_hops must be 1-10; decay_factor must be 0.0-1.0. Document in parameter descriptions.

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

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry