Scoring was not performed

• Expand each tool description from 15-30 chars to 100-150 chars. Format: 'Lists available Kubernetes [resource type]. Returns array of objects with id, name, and status. Use this to discover available [resources] before calling other tools. Supports optional filtering by namespace (pass namespace parameter) and pagination (limit, offset).'
• Add parameter descriptions and type definitions. Example for namespaces_list: parameters should include 'namespace' (string, optional, description: 'Filter to single namespace'), 'limit' (integer, 1-500, default 50), 'offset' (integer, default 0).
• Document output schemas explicitly. Example: 'Returns {"items": [{"name": string, "status": string, "age": string, "creation_time": ISO8601}], "total": integer, "has_more": boolean}'.
• Add error guidance to descriptions. Example: 'Returns 403 if kubeconfig lacks cluster-admin permission. Returns 404 if context not found, call configuration_contexts_list first to discover available contexts.'
• Implement proper pagination for list tools. Add limit (1-500, default 50) and offset parameters. Document the returned total count and has_more flag so LLMs can iterate correctly.
• For events_list and namespaces_list, add filter parameters (e.g., 'resource_type', 'warning_level') with enum constraints rather than free-form strings to prevent hallucinated filter values.
• Consider splitting generic targets_list into more specific tools: get_current_context, list_contexts (discovery). This improves LLM intent matching per the tool-chain pattern.