12 tools with basic schemas and descriptions, but significant gaps in parameter documentation and output schema clarity. Most tools have 1-4 parameters with type info, but descriptions are minimal (10-50 chars). No error handling guidance, no output schemas documented, no tool annotations (readOnlyHint/destructiveHint). Naming is verb-noun compliant (pods_list, pods_delete, resources_create_or_update). Risk classifications present but not integrated into descriptions. No pagination support visible for list operations despite potential for large result sets.

• Document output schemas for all tools. Example for pods_get: {pod_name: string, namespace: string, status: string, containers: [{name: string, image: string, ready: boolean}], created_at: ISO8601}. This enables LLMs to chain calls and extract required fields.
• Add pagination to list operations: pods_list_in_namespace(namespace, limit=20, offset=0) → {pods: [...], total_count: int, has_more: boolean}. Cap results at 50 items per call to prevent context window exhaustion.
• Enhance parameter descriptions with constraints and examples. E.g., 'namespace: Kubernetes namespace name (alphanumeric + hyphens, 1-63 chars). If omitted, uses current context.' and 'port: TCP port to expose (1-65535, optional; creates Service and Route if > 0)'.
• Add confirmation/dry-run support for destructive tools. pods_delete(namespace, name, dry_run=false) should return 'Would delete pod X in namespace Y' when dry_run=true, letting agents preview before executing.
• Integrate Risk classifications into tool annotations. Mark pods_delete and resources_delete with destructiveHint=true, pods_list_* with readOnlyHint=true, and pods_run with idempotentHint=false in the tool schema.