Menu
GPU CI patterns for CUDA compilation caching, manylinux wheels, and multi-arch builds. Use when distribution=pypi-wheel or hardware_targets includes cuda.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Session initialization — run at the start of EVERY session before any other action.
Karpathy-inspired behavioural guidance for non-trivial coding, debugging, review, or refactor work. Use to surface assumptions, keep changes minimal, and define verifiable success criteria.
Manage dependency-aware multi-session workflows in agent_roadmaps/.
Orchestrate build workflows. Environment setup, compilation, and testing.
Validate constraint compliance at any time during development.
Create a new project from the repo_template. Supports Python, C++/CUDA, and Hybrid projects. Template-only skill.
| name | gpu-ci |
| description | Design or review GPU-aware CI for CUDA builds, tests, caches, containers, and Python wheels. Use when hardware targets include CUDA, distribution uses wheels, or CI needs GPU runner gating, multi-architecture coverage, or manylinux validation. |
Read and follow the canonical vendor-neutral skill at
.agents/skills/gpu-ci/SKILL.md. That file is the single source of truth.
