docs(skill): discover Kubernetes gateway workload

[elezar]

Summary

Update Kubernetes troubleshooting guidance so commands target the gateway controller actually installed by Helm. Each standalone diagnostic block now discovers deployment/openshell or statefulset/openshell and errors clearly if neither exists.

Related Issue

Related to #1868. This extracts a generally applicable skill cleanup into its own PR.

Changes

• Resolve the installed gateway controller into GATEWAY_WORKLOAD.
• Use that workload for gateway logs, rollout status, and image inspection.
• Replace troubleshooting guidance that told users to try both controller kinds.
• Keep HA peer troubleshooting and unrelated documentation out of scope.

Testing

• mise run pre-commit passes (partial: Markdown, Helm lint/docs, license, and formatting checks passed; the run then failed because grpc_tools was unavailable to the proto task and z3.h was missing for Rust tests)
• Focused markdownlint passes for the changed skill
• Updated shell snippets pass bash -n
• git diff --check passes
• Unit tests added/updated (not applicable: documentation-only change)
• E2E tests added/updated (not applicable: documentation-only change)

The sync-agent-infra audit found the skill inventory and workflow chains consistent. It also found pre-existing, unrelated architecture-table drift: AGENTS.md omits six current crates (openshell-driver-podman, openshell-prover, openshell-server-macros, openshell-supervisor-network, openshell-supervisor-process, and openshell-vfio). This PR intentionally does not widen scope to fix that drift.

Checklist

• Follows Conventional Commits
• Commits are signed off (DCO)
• Architecture docs updated (not applicable)

[copy-pr-bot]

Auto-sync is disabled for draft pull requests in this repository. Workflows must be run manually.

Contributors can view more details about this message here.