[Feature Request] Add statusOnly parameter to resources_get to reduce context usage

[pixelsoccupied]

When diagnosing issues with Kubernetes resources, I typically only need the .status section (conditions, phase, readiness, error messages). The current resources_get tool returns the entire resource, which can be very large — especially for CRDs/operators (e.g., ClusterVersion, ArgoCD Application, MachineSet) where .spec, .metadata.managedFields, and annotations can be hundreds of lines.

This wastes LLM context window tokens on information that isn't needed for diagnosis, and can cause truncation on large CRs.

Happy to open up a PR

[Cali0707]

Hey @pixelsoccupied - I've been playing around with the token tracking in our eval suite (using mcpchecker) to find ways to reduce the token counts. One thing we found that works is described in #877

Do you think this would work for you?