Server has 8 tools with significant quality gaps. Tool naming is inconsistent (example_prompt, onetrust_ingest_* vs onetrust_*). Descriptions vary widely in quality, some are detailed (onetrust_ingest_assessments at ~280 chars) but others are vague (onetrust_cookie_domain_data at ~50 chars). Most critically, input schemas are either missing or inferred from code rather than explicitly visible in the source. Parameters like 'action' in onetrust_data_mapping accept 40+ enum values but lack proper schema constraints. No output schemas documented. Error handling is absent from descriptions. Security concerns: params_json accepts arbitrary JSON strings, risking injection. Tool composition is poor, onetrust_data_mapping bundles 40+ unrelated operations into one tool.

• Split onetrust_data_mapping into 8-10 focused tools: create_data_category, delete_data_category, create_data_element, delete_data_element, create_inventory, update_inventory, delete_inventory, link_inventory_to_parent. Each tool should have a single, clear responsibility.
• Replace params_json free-form strings with explicit, typed parameters. E.g., onetrust_ingest_assessments should accept 'size' (integer, 1-1000, default 100) and 'offset' (integer, default 0) instead of '{"size":100}'. Document each parameter's type, range, and purpose.
• Document output schemas for every tool. Specify returned fields, types, and structure. E.g., 'Returns: {assessments: [{id, name, template_id, person_id}], total_count, next_offset}'. This enables LLM chaining and data extraction.
• Expand vague descriptions to 100-200 chars. E.g., 'Manage OneTrust cookie domain data operations' → 'Retrieve cookie domain metadata and scan results. Use this to understand which domains have been scanned and their cookie inventory. Requires domain_id or domain_name parameter.'
• Add error handling guidance to each tool description. E.g., 'Returns 404 if assessment not found, try listing assessments first. Returns 403 if insufficient permissions, contact OneTrust admin. Retryable on 5xx errors.'

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

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