MCP server for Kubernetes operations
Server has 8 tools with consistent naming (verb_noun pattern) and basic descriptions. All tools have Zod schemas with type definitions and parameter descriptions. However, descriptions are terse (avg ~60 chars, well below the 194-char baseline), lack context on WHEN to use tools, and omit guidance on error recovery. Output schemas are not documented, LLMs cannot predict response structure. No pagination support despite list_* tools. Error handling is minimal (no recovery guidance, no categorization). Security: apply_manifest accepts arbitrary YAML without validation or dry-run confirmation, creating risk for destructive operations.
Apply a Kubernetes manifest (YAML) to the cluster, creating or updating resources
Get Kubernetes events, optionally filtered by namespace
Get logs from a pod, optionally specifying container and tail line count
List deployments in a Kubernetes namespace
List all namespaces in the Kubernetes cluster
List pods in a Kubernetes namespace with optional label selector filter
List services in a Kubernetes namespace
apply_manifest accepts arbitrary YAML without validation, dry-run, or confirmation. Destructive operations (delete, replace) can execute immediately without safeguards.
Descriptions are terse (avg 50 chars vs 194-char baseline). Missing context on WHEN to use each tool, prerequisites, and how they differ. E.g., list_deployments vs list_pods distinction unclear.
No output schemas documented. LLMs cannot predict response structure (JSON fields, types, nesting). Responses are JSON strings, structure is opaque to the agent.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 56 | 2026-07-28+ | v2 |
Scale a Kubernetes deployment to a specified number of replicas
List tools (list_pods, list_deployments, list_services, list_namespaces, get_events) lack pagination (limit, offset, next_cursor). Large result sets will blow context window.
Error handling is absent. No recovery guidance, error categorization, or actionable messages. Stack traces or raw API errors will confuse LLMs.