Loading ToolBench
MCP server for diagnosing and triaging Kubernetes cluster issues, providing root cause analysis and remediation recommendations
Critical failures across all dimensions. The server exposes only one tool, execute_skill_action, with a vague name that violates verb_noun conventions. The description is 156 chars but lacks WHEN/WHY guidance. Input schema is present but cluster_context is typed as 'object' with no field documentation, LLMs cannot infer what properties it accepts. No output schema is documented; the client.py mock returns a complex nested structure (diagnostic_findings array with pod/issue/root_cause/recommended_fix/severity fields) that is completely invisible to the MCP protocol layer. The tool performs read-only diagnostics but the description does not state this. No error handling, no pagination, no parameter constraints. The mcp_server.py is a stub that prints JSON and exits, it does not implement the MCP protocol at all (no JSON-RPC, no tool invocation handling, no request/response loop). This is not a functional MCP server.
Execute diagnostic triage on Kubernetes cluster events and return health analysis with remediation recommendations
Tool name 'execute_skill_action' is generic and does not follow verb_noun convention. 'execute' and 'skill' are vague, LLMs cannot infer the specific action. Should be 'diagnose_kubernetes_cluster' or 'triage_cluster_health'.
Input parameter 'cluster_context' is typed as 'object' with no field documentation. LLMs cannot determine what properties to pass. Must document expected fields (e.g. cluster_name, region, api_version) or split into explicit parameters.
No output schema documented. The client.py mock returns diagnostic_findings (array of objects with pod, issue, root_cause, recommended_fix, severity), cluster_health_score, healthy_pods, degraded_pods, none of which are visible in the MCP tool definition. LLMs cannot plan downstream calls or extract results.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 21 | 2026-07-28+ | v2 |
mcp_server.py does not implement the MCP protocol. It prints a JSON status message and exits. There is no JSON-RPC request/response loop, no tool invocation handler, no streaming support. This is a non-functional stub.
No error handling or recovery guidance. If the Kubernetes cluster is unreachable, the tool provides no actionable error message or retry strategy. Agents cannot self-correct.