The server has 5 tools with solid naming conventions (all verb_noun style: list_, get_, apply_, delete_, post_). Tool descriptions exist but are minimal (9-64 chars, well below the 194-char baseline for production tools). Tool annotations are present and correct. The server demonstrates good engineering fundamentals but lacks the documentation depth required for reliable LLM interaction, parameter descriptions are present in the schema builders but tool-level guidance about prerequisites, return types, and error recovery is missing.

• Add complete input schema definitions to pkg/mcp/server.go for get_resource, delete_resource, and post_resource. Ensure each schema includes type, required fields, and descriptions for all parameters.
• Expand tool descriptions from 9 - 64 characters to 100 - 250 characters. For each tool, explain: (1) what it does, (2) when to use it (e.g., 'Use this instead of list_resources when you need a single resource'), (3) prerequisites (e.g., 'Requires a valid namespace for namespaced resources'), (4) common error cases and recovery.
• Document output schemas for all tools. Add a 'returns' field to each tool definition specifying the response structure. For example, list_resources should document that it returns {items: [{...}], continue: string, total: number}.
• Add actionable error guidance. In get_resource description, add: 'If the resource is not found, verify the namespace, resource type, and API group are correct.'
• Clarify pagination defaults and limits in list_resources. Add: 'Default limit is 20. Maximum limit is 500. Passing limit=0 returns all items (use with caution for large result sets).'
• For delete_resource, implement a two-step pattern: provide a dry-run mode or require explicit confirmation. Document this in the tool description: 'This is a destructive operation. Set dry_run=true to preview the deletion before executing it.'
• Document the 'group' parameter more clearly. Add: 'For core API resources (v1), omit group. For others (e.g., apps, networking.k8s.io), specify the group. Use list_api_groups to discover available groups.'