dotfiles

Reduce AI detection rate in writing (Chinese & English). Use when de-AIing, humanizing, or auditing text for AI traces in any format (LaTeX, Typst, Markdown, plain text).

以鲁迅的语气、用词、逻辑和修辞写作或改写文章。支持杂文、散文、评论等体裁，基于35篇学术论文的系统研究成果。Use when writing in Lu Xun's style or converting text to Lu Xun's voice.

分析 git 变更，自动拆分为多个 Conventional Commits 提交

Use when porting ROS1 packages (roscpp, rospy, msg/srv/action, launch, CMakeLists.txt, package.xml) to ROS2 Jazzy, or reviewing mixed ROS1/ROS2 code for API drift.

Tampermonkey userscript development assistant. Use when writing, reviewing, debugging, or optimizing Tampermonkey/Greasemonkey userscripts for browser automation and page customization.

Profiles CUDA kernels with Nsight Compute CLI, imports .ncu-rep reports, and turns NCU metrics into structured bottleneck and code-level analysis. Use when the user asks to diagnose CUDA kernel performance, interpret Nsight Compute output, compare optimization runs, or trace Nsight Compute findings back into local CUDA/C++ kernel code.

Audit and improve English LaTeX academic papers. Use when writing, reviewing, or compiling IEEE, ACM, Springer, NeurIPS, and ICML papers.

Chinese LaTeX academic thesis assistant (PhD/Master). Use when writing, reviewing, outlining, compiling, or de-AIing Chinese LaTeX theses.

Unified paper audit for Chinese and English papers. Use when reviewing paper quality, pre-submission checks, or doing adversarial reviews.

Typst academic paper assistant (Chinese and English). Use when writing, compiling, or improving Typst papers for conference submissions.

Scan document files (tex/md/typst) for figure prompts in comments, generate figures via AI, and insert them. Supports custom output directories or defaults to ./images/.

Generate and edit images using Nano Banana Pro (Gemini 3 Pro Image Preview) via NewAPI reverse proxy for Poe. Use when the user asks to generate, create, draw, make, edit, modify, change, alter, or update images. Also use when user references an existing image file and asks to modify it in any way (e.g., "modify this image", "change the background", "replace X with Y"). Supports text-to-image generation and image-to-image editing with configurable resolution (1K, 2K, 4K), aspect ratios, and optional web search for real-time information. The model can create detailed, context-rich visuals, precisely edit or restyle input images with exceptional fidelity, and generate legible text in images in multiple languages.