primatrix.ai
Official@primatrix
Optimizes TPU pretraining performance and GKE cluster resource management through structured profiling, cluster orchestration, and rigorous software development lifecycle methodologies.
Agent Skills by primatrix.ai
Showing 28 vetted skills indexed across 1 GitHub repositories.
primatrixxprof-profiling-analysis
Analyze TPU/GPU performance with operator breakdowns, memory profiles, and A/B comparisons.
primatrixagent-recap
Generate daily/weekly summaries of Claude and Codex session history and optionally sync them to GitHub Issues.
primatrixprofile-anatomy
Parse TPU pretraining profile data and document its schema.
primatrixmemory-profile
Analyze TPU pretraining profiles to determine HBM peak occupancy and buffer allocation.
primatrixcomm-analysis
Analyze communication primitives in TPU pretraining profiles for bandwidth utilization and overlap.
primatrixcompute-breakdown
Analyze TPU pretraining compute efficiency from XSpace/XPlane profile data.
primatrixbeaver-engine
Manage GitHub issue lifecycles with Project V2 fields and Issue Types.
primatrixspec-document-reviewer
Evaluate RFC drafts for completeness, fact traceability, and internal consistency.
primatrixgke-tpu
Plan, render, and apply GKE TPU Job manifests on TPU v6e/v7x.
primatrixlint-fix
Check and fix Python code style issues with isort, ruff, black, and codespell.
primatrixapply-resource
Manage GKE TPU Nodepool resources via XPK commands.
primatrixexec-remote
Execute Python scripts on remote GPU or TPU clusters via SkyPilot.
primatrixdeploy-cluster
Automate SkyPilot-managed TPU cluster deployment on Google Kubernetes Engine.
primatrixusing-git-worktrees
Create isolated Git workspaces using native tools or git worktree fallbacks.
primatrixtest-driven-development
Implement Test-Driven Development with a red-green-refactor cycle.
primatrixsystematic-debugging
Guide software debugging through a four-phase root cause analysis process.
primatrixusing-superpowers
Establish rules for invoking Superpowers skills across platforms.
primatrixdispatching-parallel-agents
Dispatch independent investigation tasks to parallel agents without shared state.
primatrixexecuting-plans
Execute detailed plans with integrated review checkpoints.
primatrixfinishing-a-development-branch
Automate branch completion decisions with conditional merge, PR, or cleanup actions.
primatrixbrainstorming
Guide structured brainstorming sessions to propose designs and obtain user approval.
primatrixwriting-plans
Generate implementation plans with task decomposition and documentation for complex tasks.
primatrixrequesting-code-review
Create structured code review requests with Git commit context.
primatrixreceiving-code-review
Guide developers in responding to code review feedback with technical verification.
Frequently Asked Questions About primatrix.ai
FAQPage SchemaWhat specific performance tasks can be performed on TPU clusters?▼
Users can analyze TPU pretraining profiles to evaluate compute efficiency, HBM peak occupancy, buffer allocation, and communication primitive bandwidth utilization using XSpace and XPlane data.
Which infrastructure environments are supported for cluster deployment?▼
The registry supports planning, rendering, and applying GKE TPU Job manifests for v6e and v7x hardware, alongside remote execution on GPU or TPU clusters via SkyPilot.
What are the prerequisites for managing GKE TPU resources?▼
Deployment requires an active Google Kubernetes Engine environment, configured XPK command access, and SkyPilot integration to manage nodepool resources and remote cluster execution.