Skip to content
Get packs

GKE TPU Dynamic Slices Monitoring

Diagnose, validate and clean up Google Kubernetes Engine TPU slice resources

Source Download skill
Repository
google/skills
Stars
21,782
Licence
Apache-2.0
Files
4 files

What it does

This skill teaches an AI agent how to check the health of TPU slices (groups of Google's AI accelerator chips) running on Google Kubernetes Engine (GKE). Once installed, the agent runs kubectl describe on a slice, reads its status conditions against a lookup table to explain what state it is in and what to do next, and checks that workload manifests (including multi-slice JobSet setups) carry the correct TPU labels and annotations. It can also walk through removing stuck finalizers to force-delete a hung slice, with a warning that this can leave cloud resources orphaned.

Great for

  • Engineers running large AI training or inference jobs on GKE TPU clusters who need faster diagnosis of slice provisioning failures
  • Platform or SRE teams checking slice lifecycle states and manifest configuration
  • Users who want a guided checklist before patching stuck Kubernetes resources

Skip it if

  • Generic GKE node pool creation or non-TPU workloads, which the skill itself says to route to gke-basics or gke-cluster-creation instead
  • Users without a GKE cluster, since it depends on existing cluster access; force-deleting slices with the finalizer patch can orphan VM, network or accelerator resources and should be reviewed by someone who understands the cluster

Try asking

  • “Check why my TPU slice test-slice is stuck in SliceCreationFailed and tell me what to fix”
  • “Verify that my multi-slice JobSet manifest has the right TPU topology annotations and nodeSelectors”
  • “Explain what the ACTIVE_DEGRADED state means for my slice and what logs to inspect”
  • “Help me safely remove stuck finalizers from a slice that is stuck in DEACTIVATING”
Step-by-step guide: how to use GKE TPU Dynamic Slices MonitoringInstall, first run, examples, tips and FAQs

Install in 3 steps

  1. 1

    Download the skill

    The ZIP contains the 4 original files, the Apache-2.0 licence and an attribution note.

    gke-ai-troubleshooting-tpu-dynamic-slices-monitoring.zip
  2. 2

    Add it to your agent

    Claude Code (all projects)

    mkdir -p ~/.claude/skills && unzip gke-ai-troubleshooting-tpu-dynamic-slices-monitoring.zip -d ~/.claude/skills/

    Claude Code / Agent SDK (this project only)

    mkdir -p .claude/skills && unzip gke-ai-troubleshooting-tpu-dynamic-slices-monitoring.zip -d .claude/skills/

    Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.

  3. 3

    Ask for the task

    No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.

Get better results

  • Give the agent your project ID, cluster name, location and slice name up front so it can run the commands without guessing
  • Review any kubectl patch command before approval, since removing finalizers is marked High Risk in the skill
  • The included validate_queries.sh script only checks for a configured project and does no real validation, as it defines no queries

What's inside

  • references/attribution.md
  • references/failure_signatures.md
  • scripts/validate_queries.sh
  • SKILL.md

Redistributed unmodified from google/skills under the Apache-2.0 licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.