release-blog-features

Diff two release branches to generate a structured PyTorch blog feature draft.

3|29|Updated Apr 16, 2025
One-click install
npx skills add https://github.com/meta-pytorch/torch-release-notes --skill release-blog-features
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: release-blog-features
Source: https://github.com/meta-pytorch/torch-release-notes/tree/main/.claude/skills/release-blog-features
Command: npx skills add https://github.com/meta-pytorch/torch-release-notes --skill release-blog-features

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill helps content teams generate a structured, blog-ready PyTorch release feature draft by diffing two branches and organizing findings into a Highlights section and per-component blocks with stability tags.

Core Features & Use Cases

  • Branch-diff driven highlights: automatically extract notable changes and attach PR references.
  • Per-component sections: generate organized subsections for Dynamo / Inductor, Distributed, Export, MPS, ROCm, XPU, CPU, CUDA, and other areas.
  • Stability tagging: annotate entries with [Stable], [Beta], [Prototype], or [API-Unstable].
  • Automated draft structure: produce a header, Highlights, and per-component sections in blog format.
  • Uncertainty markers: insert TODO notes for ambiguous items and point to PRs/docs when available.

Quick Start

Diff two release branches (base and target) to produce a ready-to-paste blog Features draft organized by component and stability.

Frequently Asked Questions about release-blog-features

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I generate a PyTorch release blog draft from branch diffs?

To generate a PyTorch release blog draft from branch diffs, compare a base and target release branch to extract user-facing changes into a structured format with Highlights and per-component sections. The resulting draft includes PR references and stability tags.

What is the best way to organize PyTorch release features by component for a blog?

Organizing PyTorch release features by component is done by automatically categorizing diff changes into subsections for Dynamo, Inductor, Distributed, Export, MPS, ROCm, XPU, CPU, and CUDA. This ensures the blog draft is structured by distinct framework areas.

How are stability levels like Beta or Prototype handled in PyTorch release notes?

Stability levels like Beta or Prototype are handled by automatically annotating each feature entry in the blog draft with tags such as [Stable], [Beta], [Prototype], or [API-Unstable]. This clearly communicates the readiness of each change.

Can I automatically link PyTorch pull requests when drafting release highlights?

You can automatically link PyTorch pull requests when drafting release highlights. The workflow extracts notable changes from the branch diff and attaches the corresponding PR references and documentation links directly to each feature entry.

What happens to undocumented or ambiguous changes when generating a PyTorch release draft?

When undocumented or ambiguous changes are found during PyTorch release draft generation, the workflow inserts TODO markers. It ensures only documented features are surfaced while flagging ambiguous items for manual review and points to available PRs.