Observability MCP demonstrates solid definition quality with 8 well-named tools following verb_noun conventions (list_*, search_*, query_*, get_*). All tools have descriptions (avg 95 chars, within 10-1024 baseline). Input schemas are complete with proper JSON Schema types, enums, and constraints (e.g., limit 1-1000, level enums). However, output schemas are not documented, responses are inferred from implementation but not formally declared in tool definitions. Parameter descriptions are present but terse (avg 45 chars vs 72 baseline). No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite all tools being read-only. Error handling is minimal, no recovery guidance or actionable error messages visible in source.

• Add formal output schema documentation to all 8 tools. Example: 'Returns {"indices": [{"name": string, "type": string, "doc_count": integer}], "total": integer}'. This enables LLMs to plan downstream calls and extract correct fields.
• Annotate all tools with readOnlyHint=true in tool definitions. This signals to LLMs that these tools are safe to retry and cacheable, improving agent efficiency.
• Enhance error responses with recovery guidance. E.g., 'Elasticsearch query failed: index not found. Try list_log_indices() to see available indices, or use a broader index pattern like "logs-*".'
• Expand parameter descriptions for query_instant and query_range. Add format examples: 'PromQL expression (e.g., rate(http_requests_total[5m])) or ISO 8601 timestamp (e.g., 2024-01-15T10:30:00Z)'.
• Add next_cursor pagination to list_skywalking_services response and document it in the tool description. Cap default limit at 50 to prevent context window overflow.