primatrix/apply-resource
plugins/exec-remote/skills/apply-resource/SKILL.md
Manages GKE TPU clusters using xpk. Creates, deletes, and lists TPU Nodepool resources on Google Kubernetes Engine. Multi-user safe - always queries GKE for real-time cluster state.
development
Updated Apr 16, 2026
$ install --global
skillsauthnpx skillsauth add primatrix/skills apply-resourceInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
Security Scan Results
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
3
Scanners Passed
9
Scanners in report
Clean
TrivyContainer and dependency vulnerability scanner
95%
Clean
SemgrepStatic code analysis for vulnerabilities
95%
Clean
mcp-scan (Snyk)Model Context Protocol security validation
95%
Skipped
Snyk (dep)Open source security scanning
50%
Skipped
Socket.devSupply chain security analysis
50%
Skipped
VirusTotalMulti-engine malware detection
50%
Skipped
CrowdStrikeAdvanced threat intelligence
50%
Skipped
OSV-ScannerOpen Source Vulnerability database check
50%
Skipped
OWASP Dep-Check
50%
Last scanned: Apr 16, 2026, 4:50 AM4.2s4 files scanned
SKILL.md
- name:
- apply-resource
- description:
- Manages GKE TPU clusters using xpk. Creates, deletes, and lists TPU Nodepool resources on Google Kubernetes Engine. Multi-user safe - always queries GKE for real-time cluster state.
- argument-hint:
- [create|delete|list] [interactive-params...]
GKE TPU Cluster Management Skill
This skill manages TPU clusters on Google Kubernetes Engine (GKE) using xpk.
Multi-User Safe: This skill does NOT use local caching. All cluster information is queried directly from GKE in real-time, ensuring accuracy in multi-user scenarios where clusters may be created, modified, or deleted by other users.
Prerequisites
Before using this skill, ensure the following tools are installed:
- Google Cloud SDK (gcloud): Install guide
- Run
gcloud auth loginto authenticate - Check:
gcloud auth list
- Run
- Kubectl: Install guide
- Install auth plugin:
gke-gcloud-auth-plugin - Check:
kubectl version --client
- Install auth plugin:
- Xpk: Install guide
- Check:
xpk --help
- Check:
Defaults
The following defaults apply unless the user explicitly overrides them:
| Parameter | Default |
|----------------|----------------------------|
| PROJECT_ID | tpu-service-473302 |
| CLUSTER_NAME | sglang-jax-agent-tests |
| ZONE | asia-northeast1-b |
| NUM_SLICES | 1 |
Use these values directly — do NOT ask the user to confirm or re-enter them unless they specify otherwise.
Required Parameters
When creating a cluster, the following parameters are required:
- PROJECT_ID: GCP project ID (default:
tpu-service-473302) - CLUSTER_NAME: Name of the GKE cluster (default:
sglang-jax-agent-tests) - TPU_TYPE: TPU accelerator type (e.g.,
v6e-16,v6e-4,v4-8) — must be specified - NUM_SLICES: Number of TPU slices to provision (default:
1) - ZONE: GCP zone (default:
asia-northeast1-b)
If these parameters are already known from an upstream caller (e.g., exec-remote or deploy-cluster), use them directly — do NOT re-ask the user. Only prompt interactively when this skill is invoked standalone and the user wants to override defaults.
Operations
1. Create Cluster
Creates a new GKE cluster with TPU nodepool using Pathways.
Command:
xpk cluster create-pathways \
--cluster $CLUSTER_NAME \
--num-slices=$NUM_SLICES \
--tpu-type=$TPU_TYPE \
--zone=$ZONE \
--spot \
--project=$PROJECT_ID
Interactive Flow:
- Prompt user for CLUSTER_NAME
- Prompt user for TPU_TYPE
- Prompt user for NUM_SLICES
- Prompt user for ZONE
- Validate TPU availability in the specified zone
- Check if cluster name already exists in GKE (prevents conflicts)
- Execute cluster creation
2. Delete Cluster
Deletes an existing GKE cluster.
Command:
xpk cluster delete \
--cluster $CLUSTER_NAME \
--zone=$ZONE \
--project=$PROJECT_ID
Interactive Flow:
- Query GKE for all live clusters
- Display cluster list with current status
- Prompt user to select a cluster
- Verify cluster still exists (multi-user safety check)
- Confirm deletion
- Execute cluster deletion
3. List Clusters
Lists all managed clusters.
Command:
xpk cluster list
Real-time Query:
- Queries GKE directly using
gcloud container clusters list - Shows current status, location, and node pool information
- No local cache - always up-to-date
4. Describe Cluster
Shows detailed information about a specific cluster.
Command:
xpk cluster describe \
--cluster $CLUSTER_NAME \
--zone=$ZONE
TPU Availability Handling
When a specified ZONE doesn't support the requested TPU_TYPE:
- Automatically fetch available zones from TPU regions documentation
- Suggest alternative zones that support the TPU type
- Allow user to select a different zone or cancel
Multi-User Safety Features
This skill is designed for multi-user environments:
- No Local Caching: All cluster information is queried from GKE in real-time
- Existence Checks: Before creating, checks if cluster name already exists
- Pre-deletion Verification: Before deleting, verifies cluster still exists
- Location Validation: Automatically uses actual cluster location if different from user input
- Concurrent Operation Safe: Multiple users can operate simultaneously without conflicts
Parameter Passing
This skill can be invoked in two ways:
- Standalone (
/apply-resource create): Prompt user for all required parameters interactively. - From pipeline (called by
deploy-clusterorexec-remote): Parameters (CLUSTER_NAME, TPU_TYPE, NUM_SLICES, ZONE) are already known — use them directly without prompting.
Error Handling
- Missing prerequisites: Check if xpk, gcloud, kubectl are installed
- Authentication errors: Verify
gcloud auth list - Zone unavailability: Automatically suggest alternative zones
- Quota errors: Display quota information and suggest solutions
- Network errors: Retry with exponential backoff
Useful Resources
- TPU Architecture
- TPU Regions and Zones
- XPK Cluster Documentation
Related Skills
primatrix/memory-profile
development
VerifiedTrustedCommunity
Use when analyzing TPU pretraining HBM occupancy from a profile directory — locates the static HBM peak (the same number TensorBoard's Memory Viewer shows), enumerates every buffer alive at the peak schedule moment with size / HLO instruction / opcode / op_name, and rolls the alive set up by opcode and op_name. Reads compile-time `*.hlo_proto.pb` (BufferAssignmentProto) as the primary source; runtime `*.xplane.pb` allocator events are a secondary, often-truncated signal.
SKILL.mdUpdated May 27, 2026
primatrix/compute-breakdown
testing
VerifiedTrustedCommunity
Use when analyzing TPU pretraining compute efficiency from xplane.pb — produces source-line-aggregated HLO duration tables, layer-scoped breakdowns, non-compute (padding/cast/copy) audits, and v7x roofline shortfall vs theoretical peak. Reads schema documented by profile-anatomy.
SKILL.mdUpdated May 25, 2026
primatrix/plugins/tpu-perf/skills/comm-analysis
tools
VerifiedTrustedCommunity
--- name: comm-analysis description: Use when analyzing communication on a TPU pretraining profile — extracts every comm primitive (async + sync, TC + SparseCore), attributes axes via HLO replica_groups, computes per-row NCCL bus BW vs per-axis peak ICI BW (peak_link × k_torus_dims × directions_per_dim; TPUv7x: 200 GB/s bidir per link on a 3D torus; util% requires `--mesh-spec` with topology), and reports per-step compute/comm overlap. Builds on profile-anatomy. --- # Communication Analysis **
SKILL.mdUpdated May 25, 2026
primatrix/profile-anatomy
documentation
VerifiedTrustedCommunity
Use when reading TPU pretraining profiles (xplane.pb, trace.json.gz) — describes the on-disk layout, the XSpace/XPlane/XLine/XEvent/XStat hierarchy, and provides reference scripts that future tpu-perf skills can read as schema documentation.
SKILL.mdUpdated May 24, 2026