@pytorch

pytorch-building covers compiling executorch from source across platforms, including python package installation, c++ runtime builds, model runners, cross-compilation for ios/android, and backend-specific configurations. routes users to the appropriate build path based on their target.

add-uint-support provides step-by-step guidance for expanding pytorch operator type coverage to unsigned integers via AT_DISPATCH macro updates, with clear decision paths between two implementation methods.

at-dispatch-v2 guides conversion of pytorch's legacy AT_DISPATCH macros to the v2 format, providing step-by-step transformation rules, type group mappings, and pattern examples for aten kernel porting.

at-dispatch-v2 converts legacy pytorch AT_DISPATCH macros to the v2 api with step-by-step mapping of type groups, lambda wrapping, and argument reordering for aten kernels and cuda implementations.

pyrefly-type-coverage incrementally removes type-checking suppressions from python files by identifying and deleting file-level ignore directives, then running lintrunner to validate coverage gains.

pytorch-triaging-issues routes github issues to oncall teams, applies module and type labels from an allowlist, closes questions with forum redirects, and flags high-priority items for human review. pt2 issues receive special handling requiring module labels before marking triaged.

pytorch-pr-review evaluates pytorch pull requests for code quality, test coverage, security, and backward compatibility issues that automated checks miss. the skill is philosophically sound but lacks procedural specificity and does not address failure modes or recovery paths.

metal-kernel covers implementing native Metal kernels for PyTorch operators on Apple Silicon via c10/metal infrastructure rather than MPSGraph, with workflows for new operators and MPSGraph migration.

skill covers pytorch docstring conventions using sphinx/rest format with sections for signature, description, arguments and examples, though the actual file content appears truncated or incomplete.

pytorch-setup activates a conda environment and runs an install script for executorch development, with optional huggingface integration via pinned commit.

Add unsigned integer (uint) type support to PyTorch operators by updating AT_DISPATCH macros. Use when adding support for uint16, uint32, uint64 types to…

Review PyTorch pull requests for code quality, test coverage, security, and backward compatibility. Use when reviewing PRs, when asked to review code changes,…

Guide users through creating Agent Skills for Claude Code. Use when the user wants to create, write, author, or design a new Skill, or needs help with SKILL.md…

Export a PyTorch model to .pte format for ExecuTorch. Use when converting models, lowering to edge, or generating .pte files.

Profile ExecuTorch model execution. Use when measuring performance, analyzing operator timing, or debugging slow models.

Write Metal/MPS kernels for PyTorch operators. Use when adding MPS device support to operators, implementing Metal shaders, or porting CUDA kernels to Apple Silicon. Covers native_functions.yaml dispatch, host-side operators, and Metal kernel implementation.

Write docstrings for PyTorch functions and methods following PyTorch conventions. Use when writing or updating docstrings in PyTorch code.

Migrate a file to use stricter Pyrefly type checking with annotations required for all functions, classes, and attributes.

Guide users through creating Agent Skills for Claude Code. Use when the user wants to create, write, author, or design a new Skill, or needs help with SKILL.md files, frontmatter, or skill structure.

Debug AOTInductor (AOTI) errors and crashes. Use when encountering AOTI segfaults, device mismatch errors, constant loading failures, or runtime errors from aot_compile, aot_load, aoti_compile_and_package, or aoti_load_package.

aoti-debug — an installable skill for AI agents, published by pytorch/pytorch.

triaging-issues — an installable skill for AI agents, published by pytorch/pytorch.