WindChimeRan/claude-dotfiles

no-ai-slop

Edit drafts into sharper, more human writing while preserving the writer's personal voice, or detect AI-slop patterns without rewriting. Use when the user wants a draft clearer, more direct, more opinionated, or less AI-sounding, or asks whether writing reads…

原文语言：英语

writing-style

Technical writing style guide for editing blog posts, research write-ups, documentation, and public-facing responses (paper rebuttals, review replies, public comments). Use when reviewing, editing, or drafting technical prose — especially when asked to…

原文语言：英语

refactor-pr-check

Checklist for preparing or reviewing a refactoring PR — a change that moves, extracts, renames, or reorganizes code with behavior preserved. Use when you have written or are about to open a refactor/cleanup/extract/move PR, or are reviewing one, and want it…

原文语言：英语

html-metal-explainer

Generate educational HTML reports that explain Metal GPU kernel internals, optimizations, and parallelism for the vllm-metal project. Use when the user asks to explain a kernel concept, compare kernel implementations, document an optimization proposal, or…

原文语言：英语

jekyll-post-preview

Visually verify a Jekyll / al-folio blog post — embedded HTML/CSS/JS widgets, figures, tables, math — by building the site and screenshotting it in light AND dark with headless Chrome, then reading the images back to check them. Use this whenever you edit a…

原文语言：英语

memstat-bench

Capture a per-second memory timeline (wired / used / active / pressure) during a vllm-metal benchmark and plot it. Supports single-run and two-run side-by-side comparison. Use when the user needs to see how a serving run behaves in Metal-wired memory over…

原文语言：英语

paper-rebuttal

Strategic patterns for writing strong author rebuttals to academic peer reviews. Use when drafting, planning, or revising a response to reviewer comments on a conference or journal paper — especially ML/NLP venues (NeurIPS/ICML/ICLR/ACL/COLM/EMNLP/CVPR/etc.).…

原文语言：英语

benchmark

Run vLLM serving benchmarks on the vllm-metal project. Handles server startup, benchmark execution, branch comparison (git checkout), and generates PR-ready markdown with GitHub summary/detail blocks and matplotlib comparison figures. Use when the user asks…

原文语言：英语

bibtex-collector

Collect BibTeX entries from Google Scholar using the Chrome browser extension. Use when the user asks to find, collect, or add BibTeX entries for papers referenced in LaTeX files. CRITICAL - this skill enforces zero-hallucination by only using verbatim BibTeX…

原文语言：英语