从本地工具到AI代理基建：本周开源项目的实用主义与硬核实验

收录数按本期首次达标收录的项目统计。

• Caho1/open-claude-uiTypeScript · 1,210 Stars

  UI组件复用、逆向分析学习

  Stars 统计更新于 2026/10/19 09:45:41

  查看原始资料

  作者描述

  基于 REA 深度拆解并高保真还原 Claude 官方桌面端 UI。还原 30 个独立 React 组件与 8 项硬核交互物理算法，附带 3000+ 行逆向分析手册。

  编报时的 AI 中文解读

  高保真还原 Claude 桌面端 UI 的组件库与动效引擎，提供 30 个生产级 React 组件和 8 项交互物理算法，附带 3000+ 行逆向分析手册，适合前端开发者学习或复用。

  编报时的 README 依据

  | **系统与协作拓扑** | `ContainerResponsiveBar.tsx` | 基于 `@container` 的 4 阶流体压缩梯队 (>=576px / <480px / <384px / <320px)，杜绝单字折行。 | | **系统与协作拓扑** | `useClaudeRovingFocus.ts` | W3C WAI-ARIA 规范：方向键二维与一维网格漫游导航、Home/End 边界快速跳跃、RTL 翻转。 | | **系统与协作拓扑** | `useRateLimitBackoff.ts` | 429/503 截断指数退避（`t = min(t_max, t_base * 2^a) * (1 ± J)`）、全抖动与熔断。 | --- ## 🚀 快速上手 (Quick Start) ### 1. 引入样式表与 Tailwind 配置 在你的项目中引入 CDS 核心主题样式表： ```css /* 引入全局主题变量与关键帧 */ @import "open-claude-ui/code-examples/cds-theme.css"; ``` 在你的 `tailwind.config.js` 中扩展主题： ```javascript

  cacheWriteTokens: 16000, maxContextWindow: 200000 }} /> <ChatComposer onSend={(text) => console.log('Send message:', text)} /> </div> </ScrollFadeContainer> </div> ); } ``` --- ## 🎨 视觉故事书预览 (Interactive Storybook) 无需启动繁重的本地服务器，只需双击打开自带的 HTML 文件即可体验全量高保真组件： ```bash open docs/claude-components-storybook.html ``` 故事书包含： * 陶土色彩体系与暗黑模式即时切换 * 全量按钮尺寸与悬浮反馈 * 999px 全圆角滑动分段胶囊 * 二阶弹簧动效实时演示 * 附件磁贴与代码复制衰减状态机 --- ## 🤖 AI Agent 协同与开发规范 (Agent Guidelines)

  <a href="https://github.com/morluto/rea"><img src="https://img.shields.io/badge/Reverse%20Engineered%20with-REA-ea580c?logo=github" alt="Reverse Engineered with REA" /></a> <a href="assets/README.md"><img src="https://img.shields.io/badge/assets-clean--room%20compliance-success" alt="Asset Compliance" /></a> <a href="https://www.typescriptlang.org/"><img src="https://img.shields.io/badge/TypeScript-5.5+-blue?logo=typescript" alt="TypeScript" /></a> <a href="https://react.dev/"><img src="https://img.shields.io/badge/React-18+-61DAFB?logo=react" alt="React" /></a>

  <a href="https://tailwindcss.com/"><img src="https://img.shields.io/badge/TailwindCSS-3.4+-38B2AC?logo=tailwind-css" alt="TailwindCSS" /></a> <a href="https://www.framer.com/motion/"><img src="https://img.shields.io/badge/Framer_Motion-11+-0055FF?logo=framer" alt="Framer Motion" /></a> </p> --- ## 📖 简介 (Introduction) `open-claude-ui` 是基于对现代桌面端 AI 客户端进行深度拆解与逆向分析构建的高保真开源 UI 组件库与动效引擎。 整个逆向推导、反混淆静态分析与 ASAR 资源解构，均依托于先进的多平台逆向工程框架 **[REA (Reverse Engineer Anything)](https://github.com/morluto/rea)** 完成——通过 REA 的静态 AST 解析、闭包代码反混淆与运行时跟踪能力，穿透生产级压缩包，精准还原了底层工程物理常数与数学模型。

  不同于市面上常见的盲猜或近似复刻，本项目所有组件规范、物理阻尼常数、着色器关键帧及音频管道均**严格依据真实前端架构反编译验证**提炼而成： * **温润的人文质感 (Editorial Warmth)**：纯正陶土橙（Clay `#d97757`）、4 层微对比表面阶梯（Surface 0-3）与视网膜高分屏发丝细边补偿（Hairline Ring）； * **严苛的交互物理学 (Interaction Physics)**：4 套生产级物理弹簧参数调校、4px 指针移动死区、CJK 输入法合成守卫、7 档渐隐非线性 Alpha 遮罩； * **高阶视听与系统架构 (Audiovisual & Architecture)**：基于原生 Web Audio API 的零延迟合成提示音、以光标为锚点的画廊跑道平移缩放、macOS 桌面毛玻璃透明融合与 Cowork 多代理协同拓扑。 > ⚖️ **版权与合规声明 (DMCA & Assets Policy)**： > 本开源项目**不公开分发任何专有二进制资产**（包括商业字体 `.ttf` 与录音 `.mp3`）。本项目提供原生系统级字体回退栈与 Web Audio API 纯代码合成音效，完全自包含、合规安全。详情参见 [assets/README.md](assets/README.md)。 ---

  查看 README 缓存打开 GitHub
• 0xProbabillity/house-controlSwift · 626 Stars

  3D家庭控制、灯光控制、空调控制

  Stars 统计更新于 2026/10/19 09:45:44

  查看原始资料

  作者描述

  Your home as a live 3D model in the Mac menu bar and on iPhone — lights and ACs via Smart Life, Home Assistant, Philips Hue and Apple Home.

  编报时的 AI 中文解读

  将家庭平面图呈现为实时3D模型，在Mac菜单栏和iPhone上控制灯光与空调，支持Smart Life、Home Assistant、Philips Hue和Apple Home设备。

  编报时的 README 依据

  <p align="center"> <img src="docs/icon.png" width="112" alt="House Control icon"> </p> <h1 align="center">House Control</h1> <p align="center"> Your home as a live 3D model — in the Mac's menu bar, around its notch and on your iPhone —<br> controlling its lights and air conditioners through Smart Life, Home Assistant, Philips Hue and Apple Home. </p> <p align="center"> <a href="#download"><b>Download</b></a> &nbsp;·&nbsp; <a href="#build-and-run"><b>Build it yourself</b></a> &nbsp;·&nbsp; <a href="#support-the-project"><b>Support the project</b></a> </p> <p align="center">

  <img src="docs/hero.png" alt="The panel, an air conditioner sheet and a colour light sheet"> </p> ## What it does - **Lives in the menu bar.** Click the house icon and a panel drops down; click anywhere else and it is gone. - **Your floor plan in 3D.** Rooms, walls, doors, windows and furniture are built from one JSON file. Lamps light the scene the way they light the room; an AC that is running blows visible air. - **Lights.** Tap to toggle, drag across a tile to dim, open a light for brightness, white temperature and colour.

  <img src="docs/connections.png" width="820" alt="The editor's Connections: Smart Life, Home Assistant, Philips Hue and Apple Home"> </p> Right-click the icon → **Edit Home… → Connections**, and connect what you have: | Platform | How | Details | | --- | --- | --- | | **Smart Life / Tuya** | Scan a QR code with the Smart Life app. No Tuya developer account | [docs/smart-life.md](docs/smart-life.md) | | **Home Assistant** | Found on your network by itself; paste a long-lived access token | [docs/connections.md](docs/connections.md#home-assistant) |

  | Click the menu bar icon | Open or close the panel | | Right-click the menu bar icon | Edit Home, Load Floor Plan, Reveal Plan File, Reveal Device Log, Show Island, Pair iPhone, AI Access, Reload, Quit | | Click a lamp in the 3D view, or its tile | Toggle it | | Drag sideways across a light tile | Dim it | | Click the sliders icon on a light tile | Brightness, white temperature, colour | | Click an AC, or its tile | Open its sheet: temperature, mode, fan speed | | Click a room | Zoom in on it and list only its devices | | Drag the 3D view / scroll over it | Rotate / zoom |

  you are holding — so the paired Mac puts the code on its screen for you. Without a Mac, share the code to any other screen. The session is kept in the iPhone's Keychain and never leaves the device. ## Make it your home Right-click the icon and choose **Edit Home…** <p align="center"> <img src="docs/editor.png" width="820" alt="The floor plan editor"> </p> - Drag rooms to move them and drag a corner to resize; everything snaps to 5 cm. - Where two rooms touch, the wall between them appears by itself. Walls facing the camera are cut away.

  the app. House Control ships signed ad hoc instead, and earns trust the open-source way: the code is all here, the release is built in public, and you can [build it yourself](#build-and-run) in a couple of minutes. ### Remove it Quit it (right-click the menu bar icon → **Quit**), delete **House Control** from Applications and, to remove your plan and connections too, `~/Library/Application Support/HouseControl/`. ## Requirements | | | | --- | --- | | macOS | 13 Ventura or newer | | Xcode Command Line Tools | Swift 5.9 or newer — `xcode-select --install` | | Node.js | 20.19+ or 22.12+ |

  | For the iPhone app | Xcode 16 or newer, [XcodeGen](https://github.com/yonaskolb/XcodeGen) (`brew install xcodegen`), iOS 17+ | ## Build and run ```bash git clone https://github.com/0xProbabillity/house-control.git cd house-control ./scripts/build-app.sh open build/HouseControl.app ``` The script builds the web UI with Vite, the native shell with SwiftPM, and assembles an ad-hoc signed `build/HouseControl.app`. The first build downloads the npm packages; later builds take a few seconds.

  (`UNSIGNED=1 ./scripts/release-mac.sh`), attests its provenance and opens a **draft** release with the disk image and its SHA-256. Check it, then publish the draft. Running the workflow by hand from the Actions tab builds the same image as a downloadable artifact without a release. With an Apple Developer ID of your own, `ASC_KEY_ID=… ASC_ISSUER_ID=… ./scripts/release-mac.sh --publish` builds a signed and notarized image locally instead; the script lists what it needs. ## Limitations - The iPhone app is not on the App Store (it is on TestFlight for the author's testers).

  - The floor plan is edited on the Mac only. The iPhone receives it, and only while both are on the same network. - Rooms are axis-aligned rectangles. An L-shaped room is two rooms. - Apple Home devices are switched by the iPhone app only: macOS gives HomeKit to App Store apps only. - Smart Life devices are reached through Tuya's cloud, not over the local network, so they need an internet connection (Home Assistant and Hue work on your network alone). - Infrared air conditioners are one-way: the app shows what was last sent, not what the unit is doing. ## Support the project <picture>

  查看 README 缓存打开 GitHub
• StoopwenpueNq/jackie-fmri-simPython · 468 Stars

  脑状态重建

  Stars 统计更新于 2026/10/19 09:45:44

  查看原始资料

  作者描述

  jackie-fmri-sim

  编报时的 AI 中文解读

  这是一个研究原型，基于真实神经影像和生理记录，重建一只狗（Jackie）的实时情绪和脑区状态，可查询任意文本刺激并返回状态向量和行为表型。

  编报时的 README 依据

  Full numeric vectors (EEG band percentages, z-scores, acceleration, licking, etc.) are stored as `Session` dataclasses inside `scan_data.py`. --- Example session: > best friend dog #42 best friend dog → affection HR=112.0 arousal=0.412 {'joy': 0.61, ...} > i 1.6 random street dog #43 random street dog → anxiety HR=124.0 arousal=0.783 ... > social random street dog → anxiety HR=124 fMRI=assets/image - 2026-10-14T115850.011.jpg best friend dog → affection HR=112 fMRI=assets/image - 2026-10-14T115911.524.jpg

  # jackie-brain-twin **Digital twin of the canine subject “Jackie” (Canis lupus familiaris)** Reconstructed from multi-session functional MRI, simultaneous EEG, physiological telemetry and behavioural annotation.    --- ## Table of Contents 1. [Project Summary](#project-summary) 2. [Scientific Background](#scientific-background) 3. [Included fMRI Slices](#included-fmri-slices)

  neighbor dog → calm HR=98 fMRI=assets/image - 2026-10-14T115921.681.jpg ## Installation ```bash git clone https://github.com/<your-username>/jackie-brain-twin.git cd jackie-brain-twin python -m venv .venv source .venv/bin/activate # Windows: .venv\Scripts\activate pip install -r requirements.txt # currently only stdlib required

  查看 README 缓存打开 GitHub
• google-research/harness2Python · 6,697 Stars

  代理优化、基准测试

  Stars 统计更新于 2026/10/19 09:55:32

  查看原始资料

  编报时的 AI 中文解读

  Harness² 是一个用于递归代理优化的框架，通过任务级和优化级两层递归，在保持模型权重不变的情况下，自动改进代理的系统提示、工具、记忆等组件，适用于需要提升代理在开放世界中处理未见任务能力的场景。

  编报时的 README 依据

  alternatives, and the [Codex executor](docs/SETUP.md#codex-executor). ## Run a benchmark The following example installs LAB and runs one task, then evaluates and reports its result. Model-backed runs incur API costs; start with a small task selection. ```bash harness2-setup lab harness2-check --bench lab harness2 --bench lab --k 1 --mode parallel \ --only-domains antitrust-competition --max-tasks 1 harness2-judge --bench lab --k 1 --mode parallel harness2-report --bench lab --k 1 --mode parallel ``` Sequential recursion runs the same commands with `--mode sequential`. `--k` is

  the task-level recursion budget; see [Experiments](#experiments). Judging and reporting must use the same benchmark, executor, subset, `--k`, `--mode`, and seed as the refinement run. Add `--dry-run` to `harness2` to preview task selection without model calls. | Benchmark | Tasks | Install | Guide | |---|---|---|---| | JobBench | Professional work | `harness2-setup jb` | [JobBench](docs/BENCHMARKS.md#jobbench) | | LAB | Legal work | `harness2-setup lab` | [LAB](docs/BENCHMARKS.md#lab) |

  <img src="assets/components.png" alt="Eight harness components grouped by runtime role: system prompt, guardrails, memory, and skills construct context; tool guidance and scripts define the tool interface; subagents orchestrate delegated work; plugins control middleware." width="100%"> </a> </p> ## Install Use **Linux, Python 3.12, and [uv](https://docs.astral.sh/uv/getting-started/installation/)**. From the repository root: ```bash uv venv --python 3.12 .venv source .venv/bin/activate uv pip install -e '.[render]' harness2-demo --output runs/demo --k 2 ```

  The local demo runs both recursion modes with deterministic editors and real Python executions. It requires no model credentials. For model-backed benchmarks, install **Node.js 20+, the Google Cloud CLI**, and the [benchmark dependencies](docs/SETUP.md#requirements). LAB needs Pandoc and Poppler; WorkBuddy-Bench needs Docker. Then configure your models: ```bash cp .env.example .env # Edit .env with your Vertex AI project and model IDs. gcloud auth application-default login ``` See the [setup guide](docs/SETUP.md) for model settings, authentication

  Executor network access is unrestricted, and Codex runs with its sandbox disabled. Public benchmark rubrics may be accessible online. Before evaluation, configure [network restrictions](docs/EXPERIMENTS.md#web-access) to prevent access to benchmark answers and grading criteria. ## Results and resume Results are saved under `runs/`, organized by benchmark and subset where applicable. They include task summaries, deliverables, adopted harnesses, reflection notes, and evaluation results. Repeat the same command to resume. Use a new `HARNESS2_RUNS_DIR` when changing

  查看 README 缓存打开 GitHub
• KnifeLemon/FileeC# · 4,240 Stars

  文件格式转换

  Stars 统计更新于 2026/10/19 09:55:32

  查看原始资料

  作者描述

  Free, open-source file converter. Convert 180+ formats offline by drag & drop — images, PDF, Word, Excel, PowerPoint, HWP/HWPX, video, audio, e-books and archives. No upload, no Office needed.

  编报时的 AI 中文解读

  Filee 是一款免费开源的本地文件格式转换工具，支持通过拖拽方式离线转换超过180种格式，涵盖图片、PDF、Office文档、视频、音频、电子书和压缩包等，无需上传文件或安装Office等额外软件。

  编报时的 README 依据

  <p align="center"> <a href="https://filee.sh"><img src="docs/media/banner-en.jpg" alt="Filee: Drag. Drop. Converted." width="100%"></a> </p> <p align="center"> <b>A free, open-source file converter you use by dragging.</b><br> Hold a key (<kbd>Ctrl</kbd>, <kbd>Option</kbd> on macOS), drag files and drop them on a format in the donut that opens at your cursor. </p> <p align="center"> <a href="https://filee.sh"><b>Website</b></a> · <a href="https://github.com/KnifeLemon/Filee/releases/latest"><b>Download</b></a> · <a href="#install">Install</a> ·

  ## Features - **A donut at the cursor**: hold a key, drag files and drop them on a format. Only formats that fit the files appear. - **One toolbar per file type**, arranged by drag and drop, with presets for quality, size, DPI, PDF pages, video and audio, output folder and file names. - **Many files at once**, in parallel. Routes with several steps (HWP → HWPX → DOCX) are found automatically. - **Watch folders** convert every file that lands in them. - **Other ways in**: the right-click menu, a keyboard shortcut, the drop zone and the [`filee` command](#command-line).

  <a href="https://github.com/KnifeLemon/Filee/releases"><img src="https://img.shields.io/github/downloads/KnifeLemon/Filee/total?color=1D9E75" alt="Downloads"></a> <img src="https://img.shields.io/badge/Windows-10%20%7C%2011-3C3489" alt="Windows 10 | 11"> <img src="https://img.shields.io/badge/macOS%20%7C%20Linux-preview-534AB7" alt="macOS | Linux preview"> <a href="LICENSE"><img src="https://img.shields.io/badge/license-MIT-6A62C9" alt="MIT license"></a> </p> <p align="center"><b>English</b> · <a href="README.ko.md">한국어</a> · <a href="README.zh-CN.md">简体中文</a></p> <table> <tr>

  Setup can add the Explorer menu, start at sign-in, the `filee` command and optional engines. - **macOS 14+** (preview): unpack the `osx-arm64` or `osx-x64` package and run `./install.sh`. The build isn't notarized yet, so allow the first start in System Settings → Privacy & Security → **Open Anyway**. - **Linux x64/ARM64** (preview): unpack the package and run `./install.sh` (no root, needs Python 3). Drag gestures need X11. On Wayland, use the drop zone and the file manager's right-click menu. Filee shows new releases in its menu and updates itself, keeping your settings and engines.

  Marked ones need an optional engine, downloaded only when needed: ¹ Ghostscript, ² LibreOffice, ³ Pandoc, ⁴ calibre, ⁵ FFmpeg ([details](docs/ENGINES.md)). ## Command line The `filee` command uses the same engines and presets as the app (Setup or `install.sh` adds it). ``` filee convert photo.heic --to jpg filee convert *.png --to webp --quality 80 -o converted filee watch D:\Inbox --to pdf --move-originals ``` Every command and option: [filee.sh/cli](https://filee.sh/cli/) or `filee help`. ## Development Build with the .NET 10 SDK on Windows, macOS or Linux: ```bash

  查看 README 缓存打开 GitHub
• jhaizhou-ops/d2r-mobileC · 3,854 Stars

  游戏运行

  Stars 统计更新于 2026/10/19 09:55:32

  查看原始资料

  作者描述

  Open-source iOS / macOS compatibility layer for Diablo II: Resurrected with the Reign of the Warlock expansion. Runs your own Switch copy natively on iPhone, iPad and Mac — full touch controls, 60 fps, new endgame modes. Bring your own game files. 开源的《暗黑破坏神 II：狱火重生》术士君临 iOS / macOS 兼容层

  编报时的 AI 中文解读

  这是一个开源兼容层，让用户将自己拥有的《暗黑破坏神 II：狱火重生》Switch版（含术士君临扩展）在iPhone、iPad和Mac上原生运行，提供完整触控操作、60帧性能和新增终局内容。

  编报时的 README 依据

  # D2R Mobile **An open-source iOS compatibility layer for Diablo II: Resurrected with the Reign of the Warlock expansion.** It brings the Nintendo Switch release you bought, expansion included, to iPhone and iPad in full: the original game code runs natively, every control is rebuilt for touch, the frame rate holds at 60 fps, and you are in a game one second after the main menu. This is not a tech demo that "sort of runs". It is a mobile version polished for daily play, from your first character all the way through the endgame.

  <img src="https://github.com/jhaizhou-ops/d2r-mobile/releases/download/assets/en_pad_inventory.jpg" width="49%" alt="Navigating the stash with a controller, button prompts at the bottom"> </p> ## New ways to play, on top of the original Finishing the original game is only the beginning. Without touching the original rules, we added a full set of endgame content that keeps you farming. ### Greater Rift

  Platforms: iPhone / iPad (iOS 26 and later) · Game version: Switch release with Reign of the Warlock, 3.3.94012 · English / 中文 · License: MIT · **Bring your own game files** 中文: [README.zh-CN.md](README.zh-CN.md) <p align="center"> <img src="https://github.com/jhaizhou-ops/d2r-mobile/releases/download/assets/hero_en.jpg" width="100%" alt="The game filling the whole edge-to-edge screen of an iPhone: Harrogath on Hell, with the red portal to the Greater Rift"> </p> <p align="center">

  Both routes start with **the files of the game you bought**: the Switch release of Diablo II: Resurrected plus the Reign of the Warlock expansion, updated to 3.3.94012. This project is built for the version with that expansion; using it without the expansion is not supported and has not been verified. How to prepare them is in [docs/GAME_FILES.md](docs/GAME_FILES.md). | | With a Mac | Without a Mac | |---|---|---| | Computer | An Apple silicon Mac with Xcode installed | Windows / Linux / a Mac without Xcode |

  | Method | A full build from source | Download a "shell" with no game content, use the packaging tool to add your own game content, then install it with a sideloading tool | | Status | The main route; everything above comes from it | Works end to end on Windows, from making the package to playing on a device, and the game program produced on the two systems is byte-for-byte identical; the one step nobody has tried yet is signing with a sideloading tool — try it and tell us | | Instructions | The three steps below | [docs/NO_XCODE.md](docs/NO_XCODE.md) |

  Download our prebuilt "shell" package (`D2RMobile-shell.ipa`), which contains no game content, from this repository's [Releases page](https://github.com/jhaizhou-ops/d2r-mobile/releases), add your own game files, make the install package on Windows or Linux, then sign and install it with a sideloading tool. On this route only, because the computer making the package has no Apple shader compiler, the **first launch** on the phone spends about 7 minutes compiling shaders (once, with a progress bar); a package built with Xcode has no such step. The steps, and how far each one has been verified,

  are in [docs/NO_XCODE.md](docs/NO_XCODE.md). ## Known limitations | Item | Status | |---|---| | Online play | Out of scope; fully offline | | Game packages in `.nsp` format | Not verified (only `.xci` was tested) | | Signing and installing with a sideloading tool (Sideloadly, AltStore and the like) | Not verified; open an issue and tell us how it went | | Making the package on Linux | Not verified (verified on Windows and Mac) |

  查看 README 缓存打开 GitHub
• Saffsanity/sillSwift · 3,431 Stars

  远程桌面、跨设备协作

  Stars 统计更新于 2026/10/19 09:55:32

  查看原始资料

  作者描述

  Any window from your Mac, live on your iPhone or iPad. Touch, Apple Pencil, trackpad and keys. Free, open source, no servers.

  编报时的 AI 中文解读

  Sill 是一款免费开源工具，可将 Mac 上的任意窗口或整个桌面实时显示到 iPhone 或 iPad 上，并支持通过触摸、Apple Pencil、触控板或键盘进行远程交互。适用于需要在移动设备上访问 Mac 应用或桌面的用户。

  编报时的 README 依据

  - Your Mac and your device on the same network, joined by a USB cable, or near each other with Direct Wireless Connection on. Away from home, your own VPN, such as Tailscale, with Remote Access on. ## How it works - **Sill for Mac** lives in the menu bar. When your device picks a window, or the whole desktop, Sill captures it with ScreenCaptureKit and encodes it as HEVC with VideoToolbox, on the hardware video encoder in your Mac (or, while that is busy, in software). It captures nothing while no device is connected.

  Sill for Mac is a free, notarized download from [getsill.app](https://getsill.app/download). Sill for iPhone and iPad is free on the [App Store](https://apps.apple.com/app/id6816359860), and you can also build it from source (below). ## Screenshots These are the iPad screenshots from the App Store; on iPhone and iPhone Duo, Sill has layouts of its own.

  Screen.](design/screenshots/ipad-6.jpg)](design/screenshots/ipad-6.jpg) | ## Requirements - A Mac with Apple silicon, on macOS 14 or later. - An iPhone or iPad with iOS 17 or iPadOS 17 or later. - On your Mac, Sill asks for Screen Recording, to show your windows, and Accessibility, to click, scroll and type for you and to move and size the window it shows. On macOS 15 or later, also allow Local Network. On your iPhone or iPad, Sill asks for Local Network, to find your Mac, and for the camera when you scan your Mac's pairing code (you can type the code instead).

  - **The virtual display** (off by default) moves a window you stream onto a display of its own on your Mac, so it keeps updating when other windows cover it. It uses a private macOS API, which is one reason Sill for Mac comes from the website and not from the Mac App Store. Good to know: Sill shows one window, or the whole desktop, at a time, and it does not play sound from your Mac. Only the iPhones and iPads you paired can connect, unless you turn off Require pairing in Settings › Devices; then any device on your network can, still encrypted. Pairing at home came with Sill

  查看 README 缓存打开 GitHub
• FelipePrado0/gow3_pcC++ · 3,393 Stars

  游戏移植、游戏增强

  Stars 统计更新于 2026/10/19 09:55:32

  查看原始资料

  作者描述

  God of War III Remastered (PS4) for Windows and Linux (experimental AppImage). Original x86-64 code runs directly on the CPU with a custom runtime; graphics use shadPS4's Vulkan video core and cuesta4's fixes. Launcher, in-game menu, controller icons and up to 240 FPS. Requires your own game dump. No game files included.

  编报时的 AI 中文解读

  将《战神3重制版》PS4版移植到Windows和Linux的专用工具，非通用模拟器，需用户自备游戏文件，支持高帧率、修改器和性能调节。

  编报时的 README 依据

  2. Extract it to a folder of your choice, for example `C:\Games\gow3-windows`. Do not run it from inside the zip. 3. Open **`God of War III.exe`**. If Windows shows "Windows protected your PC", click **More info**, then **Run anyway**: the program is not digitally signed. 4. Click **Choose the game folder**. It takes you to the **Game** page. Next to **Game folder**, click **Browse…** and select the folder with `eboot.bin`. 5. Back on **Home**, the cover art and "CUSA01623 · version 01.02" appear. Adjust **Quick

  4. On **Game**, choose the game folder. Adjust **Display**, **Performance** and **Patches** if you like, then press **PLAY**. To play with the saved settings without opening the launcher's window (for a shortcut or Steam): ```powershell python launcher/gow3_launcher_win.py --play ``` Or without the launcher at all: ```powershell $env:GOW3_GAME_DIR = 'D:\path\CUSA01623' python run.py ``` Settings are kept in `gow3.ini` (the game and the in-game menu) and in `%APPDATA%\gow3-launcher\settings.json` (the launcher). Saves, the shader cache and

  numbers are the game's own use, not the whole system's. ## Download and play **You need** - Windows 10 (1903 or newer) or Windows 11, 64-bit. - A graphics card with Vulkan 1.3 and an up-to-date driver. - Your own copy of God of War III Remastered: **CUSA01623 with the 01.02 update**, as a folder that contains `eboot.bin`, `sce_module` and `sce_sys`. No game files are included in this download. **Steps** 1. Download `gow3-windows.zip` from the [Releases page](https://github.com/FelipePrado0/gow3_pc/releases/latest).

  - Black screen at start: **Advanced → Clear shader cache**, then start again. - The game does not open: check the **Log** page, or send `user\last_run.log` with your report. ## Requirements - Windows 10 (1903 or newer) or Windows 11, 64-bit. - A graphics card with Vulkan 1.3 and a current driver. - Your copy of the game: CUSA01623 with the 01.02 update applied (the folder with `eboot.bin`, `sce_module` and `sce_sys`). - The patches are applied only to the `eboot.bin` they were checked against byte by byte

  (sha256 `d85c8135d330c3b601bf5dc3f1dd86bd6119fb8a9fce372bed2c9785f9a79299`), and the in-game hooks (cheats, multipliers, health bar) only where the code they change has the expected bytes. Otherwise the game starts without them, and the log and the menu say why. - To build from source: [MSYS2](https://www.msys2.org) and [Python 3](https://www.python.org) with Pillow (`pip install pillow`, used for the game's icon and cover art). ## Building from source Development happens on the `gow3` branch; `main` receives tested merges.

  查看 README 缓存打开 GitHub
• onehumanbeing/esp32-iphone-agentC · 3,130 Stars

  iPhone自动化、AI辅助控制

  Stars 统计更新于 2026/10/19 09:55:32

  查看原始资料

  作者描述

  Let an AI model use a real iPhone with a $7 ESP32: AirPlay eyes, USB HID hands, any OpenAI-compatible vision model. No Mac, no developer account, no jailbreak.

  编报时的 AI 中文解读

  这是一个让AI模型通过ESP32-S3和AirPlay控制真实iPhone的项目，无需Mac、开发者账号或越狱。适用于希望自动化iPhone操作的研究者或开发者。

  编报时的 README 依据

  helper app and an Apple developer account (a free account's signature expires every 7 days). This project uses the same paths a person uses — the screen and a pointer — so the phone needs nothing installed. | | ESP32 iPhone Agent | XCTest / WebDriverAgent tools | |----------------------------------|---------------------------|-------------------------------| | Apple developer account | **Not needed** | Needed (to sign the runner) | | Xcode | **Not needed** | Needed |

  > **Set up with an AI coding agent.** Paste this into Claude Code, Codex or a similar agent on the computer the > board will connect to: > > ```text > Clone https://github.com/onehumanbeing/esp32-iphone-agent and follow its README to set it up on this machine: > flash the ESP32-S3 plugged in over USB, install UxPlay and the server, write server/.env for my model API, start > the server and give me the web UI address. Ask me for the Wi-Fi name/password and the model API key. > ``` ### 1. Flash the board

  4. In the web page, open **Manual controls → Calibrate**, then click exactly on the pointer in the live screen. ### 4. Give it a task Type a goal (for example *"Open Settings and go to Wi-Fi"*) and press **Start**, or click one of the examples. ## Web UI - **Live phone** — click to tap, drag to swipe, scroll to scroll; manual Home, Back, typing and keys. - **Live log** — every plan, thought and action as it happens, with timings. - **Step wall** — the screenshot the model decided on for every step, with its tap or swipe drawn on top, plus the

  action it checks whether the screen actually changed and tells the model, so a missed tap is noticed and retried instead of built upon. Repeated actions trigger a re-plan; a dropped link or a bad reply is recovered from. ## Hardware | Part | Notes | |---|---| | **ESP32-S3 board** | Tested with a bare [Seeed Studio XIAO ESP32S3](https://wiki.seeedstudio.com/xiao_esp32s3_getting_started/) (~$7, no camera add-on needed). Any ESP32-S3 board whose USB-C port is wired to the S3's native USB works. |

  | **iPhone** | USB-C models plug in directly. Lightning models need a Lightning to USB Camera Adapter. The phone powers the board. | | **Computer** | macOS or Linux, on the same network as the phone; runs UxPlay and the server. | | **Vision model** | Any OpenAI-compatible chat API that accepts images. Running a model locally is optional. | <p align="center"><img src="docs/images/setup.png" alt="A XIAO ESP32S3 plugged into an iPhone, next to the web UI" width="100%"></p> Use a USB-C hub with power pass-through if the phone should keep charging while the board is plugged in. ## Quick start

  - [ ] Correct the pointer only when needed, to make taps faster - [ ] More example tasks across popular apps ## Limitations - iOS does not mirror some screens (passcode entry, Apple Pay, DRM video); the agent cannot see them. - Mirroring stops when the phone locks; keep auto-lock off while the agent works. - Keyboard input is ASCII only. - Re-calibrate after changing AssistiveTouch tracking speed. - On Linux machines with several network interfaces, limit Avahi (mDNS) to the interface on the phone's network if the phone does not list the receiver. ## Responsible use

  查看 README 缓存打开 GitHub
• Arekejoker/qwen21-nsfw-pose-skillPython · 2,855 Stars

  图像编辑、姿势控制

  Stars 统计更新于 2026/10/19 09:55:32

  查看原始资料

  作者描述

  Qwen-Image 2.1 NSFW/姿势出图 skill：作者级提示词配方 + NSFW LoRA + 136 姿势 VNCCS 人偶控制；适用 Claude Code / Codex / Cursor / OpenClaw

  编报时的 AI 中文解读

  为 Qwen-Image 2.1 本地出图提供包含 NSFW/姿势生成、作者级提示词配方、136 姿势库和图形界面的技能包，适用于 Claude Code、Codex、Cursor 等智能体及手动使用。

  编报时的 README 依据

  # qwen21-nsfw-pose [](https://github.com/Arekejoker/qwen21-nsfw-pose-skill/actions/workflows/validate.yml) **Qwen-Image 2.1 本地 NSFW / 姿势出图完整配方** — 作者级提示词结构 + NSFW LoRA + 136 个 3D 姿势库 + VNCCS 人偶精确姿势控制。 基于 ComfyUI + `qwen_image_2.1_int8_convrot` 底模（与 Civitai 作者同款底模），复刻并验证过 Civitai 作者样图（2026-10-12）。 - 适用：Claude Code / Codex / Cursor / OpenClaw（agent 技能）+ 人类手动使用 - 内容：可跟跑的作者案例、强制配方、姿势大胆度阶梯（L0–L3）、完整依赖与安装指南 --- ## 效果展示 ### 本配方输出：同一海边场景的三级对照（L0 → L1 → L2）

  完整元数据（提示词/参数/seed/完整工作流）：`downloads/samples/phase5_meta.json`。 ## 三条路线 | 需求 | 路线 | 命令 | | --- | --- | --- | | 新 NSFW/性感图（文字出图） | ① T2I + NSFW LoRA（必开） | `scripts/qwen_pose_test.py --mode t2i` | | 已有角色图换姿势（保脸/保服装） | ② Edit | `scripts/qwen_pose_test.py --mode edit` | | 指定 3D 姿势库里的精确姿势 | ④ VNCCS PoseStudio | `scripts/run_vnccs_ui.py --pose <name>` | 姿势怎么选（大胆度阶梯 L0–L3、用户要求→姿势族映射表）：见 `SKILL.md` §4，或 `scripts/run_vnccs_ui.py --list-poses` 列出全部 136 个姿势。 ## 配方要点（摘要） - `er_sde` + `beta`，CFG 2.5，50 步，1152×1728 竖版 - **NSFW LoRA 永远开**（默认 `femaledVaginus v1.3b` @ 1.0）

  `MIUProject/MySuperPoseRepo`）+ 55 个手工精选，共 **136 个**。 --- ## 快速开始（作者案例跟跑） 前置安装见 [install/setup.md](install/setup.md)。装好后一条命令复刻作者淋浴样图（~5 分钟）： ```bash python3 scripts/qwen_pose_test.py \ --mode t2i --name author_case_shower \ --prompt "$(cat prompts/taws_v13_shower.txt)" \ --seed 80964 \ --steps 50 --cfg 2.5 --sampler er_sde --scheduler beta \ --width 1152 --height 1728 \ --lora "qwen2_1-femaledVaginus-nsfw-v1_3-fine_L22-31.safetensors" \ --lora-strength 1.0 \ --out outputs/author_case/shower.png ``` 对比对象：`downloads/samples/taws_v13_shower_author.jpg`。

  - 提示词必须按作者结构写满 7 要素、300–1500 字符、含 LoRA 触发词 - 弱化配方（无 LoRA/短提示词/减步数）前必须告知用户代价，不许默默减弱 完整规则与证据：`SKILL.md`。 ## 目录结构 ``` ├── SKILL.md # agent 技能（完整规则） ├── install/setup.md # 安装指南（模型/LoRA/节点/姿势库） ├── scripts/ # 出图脚本、案例脚本、media_lock.py、validate_package.py ├── workflows/ # qwen21_t2i.json / qwen21_edit.json / VNCCS_PoseStudio_QI2.1.json ├── prompts/ # 3 篇作者原提示词 + 参数 ├── references/ # 效果展示图（L0/L1/L2 海边对照）+ 默认人物图 ├── downloads/samples/ # 作者原图 + 作者元数据 + 原始工作流 JSON └── pose_library/ # 55 个手工精选姿势（安装时拷入 ComfyUI） ``` ## 说明

  - 底模与 LoRA 版权归原作者所有（HF `Comfy-Org/Qwen-Image-2.1`、Civitai 2986844 / 2958918 / 2988215） - NSFW 示例图仅供配方演示；仓库为公开仓库，介意图片的先收藏后阅读

  查看 README 缓存打开 GitHub
• liseami/CCBagaRust · 2,653 Stars

  Claude身份管理、环境验证、网络设置

  Stars 统计更新于 2026/10/19 09:55:32

  查看原始资料

  编报时的 AI 中文解读

  CCBaga 是一款面向旅居海外用户管理本地 Claude 身份与环境的桌面工具，提供身份归档、浏览器指纹轮换、系统设置调整和网络环境验证等功能。

  编报时的 README 依据

  <a href="#development">Development</a> · <a href="CONTRIBUTING.md">Contributing</a> </p> </div>  **Intended use:** CCBaga is intended solely for people residing or traveling abroad to manage their own Claude identities. Use that violates Anthropic's official terms of service is prohibited.

  CCBaga brings local maintenance and environment setup into one focused desktop workflow. Inspect Claude data before removing it, rotate your companion browser's fingerprint and restart it, align system settings with a selected region, and compare real browser readings with the machine and network exit. Two applications work together: **CCBaga-A** owns native operations; **CCBaga-B** runs an isolated Fingerprint Chromium profile. Browser measurements return to the host through a local session. Changing a setting and verifying its effect are separate steps. ## Features <table> <tr>

  **One region selector. Four everyday actions.** Easy setup puts browser cache cleanup, fingerprint rotation, timezone/language reset and identity archiving on the same page. The sidebar keeps advanced controls in Network, Claude identities, Browser and Computer. The interface supports **English and Chinese**, with shared YAML catalogs. Switching the interface language does not change the computer or browser language. The bilingual **Usage guide** covers common actions and data boundaries. See the [full guide](docs/usage.md). ## Installation

  | CCBaga-A | `CCBaga-A.app` | Identity management, system settings and environment verification | | CCBaga-B | `CCBaga-B.app` | Companion browser with a dedicated profile | Download **one DMG**; it contains both applications. There is no separate browser download. The installer uses the same pixel-art style as this README. 1. Open the disk image and drag **both applications** into Applications. 2. Launch CCBaga-A. It starts the companion browser automatically. 3. Choose English at the bottom of the sidebar if needed.

  **Platforms:** macOS 12 or later on Apple Silicon; Linux x86_64 with GTK 3 and WebKitGTK 4.1; Windows 10/11 x86_64. Intel macOS and Linux ARM64 are not supported. Linux x86_64 packages include DEB, RPM and AppImage built on Ubuntu 22.04 (glibc 2.35), plus a separate package for current Arch/pacman systems. All include the companion browser. See [Linux setup](docs/linux.md) for installation and platform limits. Official versions are published exclusively by the project owner, **liseami**.

  The success page appears only when the measured checks match, the report reaches the cleaner and no measurement error remains. **DNS and WebRTC leak checks are not independently verified.** A green result describes the measured checks, not a guarantee of anonymity or account acceptance. ### Network setup Open **Network** for three steps: **Get VPS → Get residential proxy → Install Skill**. Select **General**, **Codex**, **Cursor**, **OpenCode** or **Copilot**, then copy the installation prompt into your local coding agent. The prompt follows the app's English/Chinese setting.

  查看 README 缓存打开 GitHub
• storytold/figcraftRust · 2,645 Stars

  界面设计、原型制作、协作设计

  Stars 统计更新于 2026/10/19 09:55:32

  查看原始资料

  作者描述

  FigCraft: interface design and prototyping; an open-source, clean-room reimplementation of collaborative design workflows in pure Rust

  编报时的 AI 中文解读

  FigCraft 是一个用纯 Rust 重新实现的 Figma 开源替代品，提供界面设计、原型制作和协作功能，支持离线使用，适合设计师和开发者。

  编报时的 README 依据

  Every menu item, tool and drag is a command, so the same engine runs from the UI, the CLI, a JSON control channel or an MCP server. </td> </tr> </table> <br> ## A look around <table> <tr> <td width="50%" valign="top">

  figcraft-cli render landing.figcraft -o landing.png --scale 2 figcraft-cli export landing.figcraft -o out/ --format svg # Run a script of commands without a window figcraft-cli run build-card.json --out card.figcraft # Every command, with its parameters figcraft-cli commands --filter autoLayout # Let an agent drive it over MCP (headless, or attached to the running app) figcraft-cli mcp figcraft-cli mcp --attach 47800 ```

  Frames, auto layout, components and variants, vector networks, styles, variables and prototypes,<br> in a native app that opens instantly and works offline. Open source, and yours. </p> <p align="center"> <img alt="100% Rust" src="https://img.shields.io/badge/100%25-Rust-b7410e?style=flat-square&logo=rust"> <img alt="macOS · Windows · Linux · FreeBSD · Web" src="https://img.shields.io/badge/macOS%20%C2%B7%20Windows%20%C2%B7%20Linux%20%C2%B7%20FreeBSD%20%C2%B7%20Web-native-0EA5C6?style=flat-square">

  - **Never crashes:** no `unwrap`, `expect` or `panic` in shipped code, enforced by clippy; commands run behind a guard so a bug fails one action, not your session. - **Clean-room:** implemented from public documentation and observed behaviour only. No proprietary code, file formats, icons or images. ## Get started ```sh git clone https://github.com/storytold/figcraft cd figcraft cargo run --release -p figcraft # the desktop app cargo run --release -p figcraft-server # the collaboration server (docs/server.md) cargo test --workspace # the test suite

  > **Status:** FigCraft is in **alpha**: about 49% of the way to replacing Figma for real work, by our own estimate. Nearly all of Figma's main menu is wired ([`docs/parity-checklist.md`](docs/parity-checklist.md)) and the core editor works end to end, but FigCraft opens `.fig` files only best effort, has only a first preview of multiplayer (live sessions hosted from the app or a self-hosted server, with per-person undo and share links to edit or to view; no accounts, teams or cloud yet), runs its own WebAssembly plugins rather than Figma's, and hasn't been checked against real Figma

  查看 README 缓存打开 GitHub
• feder-cr/git-aiGo · 2,300 Stars

  版本控制

  Stars 统计更新于 2026/10/19 09:55:32

  查看原始资料

  作者描述

  Git designed for AI agents.

  编报时的 AI 中文解读

  gitai 是一个面向 AI 编码代理的开源版本控制系统，实现了 Anthropic 描述的 Scion 设计，支持服务器端历史查询、无合并队列的变更落地、自动冲突解决等功能，适合需要高效协作的代理工作流。

  编报时的 README 依据

  Details: [ARCHITECTURE.md](doc/graft/ARCHITECTURE.md) and [STORAGE.md](doc/graft/STORAGE.md). ## Quick start Requirements: Linux or macOS, Go 1.26, a C toolchain, Meson and Ninja (the build compiles a bundled Git). ```shell git clone https://github.com/feder-cr/git-ai.git cd git-ai make build # server, CLI and tools into _build/bin examples/graft/quickstart.sh # a server, a cache daemon, two agents and a land, in a temporary directory ./_build/bin/graft # the CLI and its commands ```

  To run your own server, adapt [`examples/graft/config.toml`](examples/graft/config.toml) and start it with `_build/bin/gitaly serve config.toml`. To put the repository state in a database and object storage, add a `[graft.storage]` section: start in `shadow` mode, import your repositories, then switch to `authoritative` ([STORAGE.md](doc/graft/STORAGE.md)). Inside the code the extensions are called **Graft**: the `[graft]` configuration section, the `graft` CLI, `internal/graft` and `doc/graft`. ## Documentation

  - [Agent security](doc/graft/AGENT-SECURITY.md), [benchmarks](doc/graft/BENCH.md), [fuzzing](doc/graft/FUZZING.md) - [Limitations and roadmap](doc/graft/LIMITATIONS.md) - Design documents of each phase: [PLAN.md](doc/graft/PLAN.md) to [PLAN6.md](doc/graft/PLAN6.md) ## Status Version 0.1.0, the first public release. Everything above is implemented and covered by unit, end-to-end, multi-node, crash and fuzz tests. Interfaces may still change before 1.0. Known gaps (for example garbage collection of packs in the storage core) are listed in [LIMITATIONS.md](doc/graft/LIMITATIONS.md).

  查看 README 缓存打开 GitHub
• HeyPuter/everythingJavaScript · 2,221 Stars

  开源软件替代、软件合集

  Stars 统计更新于 2026/10/19 09:55:32

  查看原始资料

  作者描述

  We're rebuilding and open-sourcing the world’s software from scratch.

  编报时的 AI 中文解读

  这是一个旨在从零开始重建并开源世界各类专有软件的大型项目，以单体仓库形式汇集了办公、协作、创意工具、工程、生产力、金融、商业、开发者工具和游戏等多个领域的应用。

  编报时的 README 依据

  The Everything Project is an effort to rebuild and open-source the world’s software from scratch. From Office to Slack and Photoshop to AutoCAD, we are recreating free, open-source alternatives to all proprietary software. This is a mono-repository containing all the applications and tools developed under the project. [Office](#office) · [Collaboration](#collaboration) · [Creative Tools](#creative-tools) · [Engineering](#engineering) · [Productivity](#productivity) · [Finance](#finance) · [Business](#business) · [Developer Tools](#developer-tools) · [Games](#games) <br> ## Getting started

  ```sh git clone https://github.com/HeyPuter/everything.git cd everything/src/[app-name] npm install npm run dev ``` <br> ## Office <table width="100%"> <tr><th width="296" align="left">Application</th><th width="420" align="left">Alternative to</th><th width="150" align="left">Links</th></tr>

  查看 README 缓存打开 GitHub
• Gathin23/tv-vibehackJava · 2,084 Stars

  电视去广告、启动器替换

  Stars 统计更新于 2026/10/19 09:55:32

  查看原始资料

  作者描述

  Your TV, minus the ads. Debloat any Android TV over ADB and give it a clean, tvOS-inspired launcher. Built in an afternoon with Claude Code.

  编报时的 AI 中文解读

  这是一个用于精简Android TV/Google TV系统并安装类tvOS风格启动器的工具，通过ADB禁用预装应用，提供可逆的恢复脚本，适合希望去除电视广告和定制主屏幕的用户。

  编报时的 README 依据

  # tv-vibehack **Your TV, minus the ads.** Debloat any Android TV / Google TV over ADB and give it a clean, tvOS-inspired launcher. Hand the prompt to Claude Code or any coding assistant, answer a few questions about which apps you use, and it does the rest on the real TV. | Stock home screen | After | |---|---| |  |  | Everything is reversible: packages are disabled, never uninstalled, and a restore script puts the stock launcher back. ## Quick start

  Needs JDK 17 and the Android SDK (platform 34, build-tools 34.0.0). On macOS: `brew install openjdk@17 android-commandlinetools && sdkmanager "platforms;android-34" "build-tools;34.0.0"`. ## Talking to the TV by hand ```bash python3 -m venv tools/venv && tools/venv/bin/pip install -r tools/requirements.txt export TV_HOST=192.168.1.50 # your TV's IP tools/venv/bin/python -I tools/tv.py --probe # model, Android version, firmware tools/venv/bin/python -I tools/tv.py 'pm list packages' # one command

  printf 'cmd1\ncmd2\n' | tools/venv/bin/python -I tools/tv.py --batch tools/disable.sh examples/xiaomi-mi-tv-4a/debloat.txt # apply a list tools/restore.sh examples/xiaomi-mi-tv-4a/debloat.txt # undo it, restore stock launcher ``` ### Gotchas we hit - **macOS: `adb connect` says "No route to host"** while the port is clearly open. Recent macOS treats the Google-signed `adb` as its own app under Local Network privacy and silently blocks it. `tools/tv.py` avoids the binary entirely.

  1. On the TV: Settings → Device Preferences → About → press OK on **Build** seven times → back → Developer options → turn on **USB debugging** (and Network/ADB debugging if your TV shows it). Keep the TV on and on the same Wi-Fi as your computer. 2. Open Claude Code (or your assistant of choice) on your computer and paste the prompt from [`PROMPT.md`](PROMPT.md), filling in your TV model. 3. Accept the "Allow USB debugging?" prompt on the TV when told to, then answer the checklist questions (which apps you use, which four go in the dock).

  - **No "wireless debugging" toggle** on some TVs (Xiaomi among them). USB debugging alone opens network ADB on port 5555. - **Chained `pm` commands** (`pm disable a; pm disable b`) may silently stop after the first. Run one per shell call; `--batch` does this. - **Large pushes hang** through the Python client. Use `tools/b64push.py` for APKs. - **Disable the stock launcher only after the new one is installed**, or the Home button lands on a factory test app. - **Third-party launchers** often charge for basics like a custom wallpaper. Hence `launcher/`.

  - **No on-screen keyboard** after switching launchers (seen on Xiaomi): the default input method pointed at a Baidu IME that isn't installed. Fix with `tools/tv.py 'ime set com.google.android.inputmethod.latin/com.android.inputmethod.latin.LatinIME'`. ## Supported TVs Anything running Android TV or Google TV: Xiaomi/Redmi, TCL, Hisense, Sony, OnePlus, Realme, Nvidia Shield and most others. Not Samsung Tizen, LG webOS or Amazon Fire TV (different OS, no ADB launcher replacement). Tested so far on the models in [`examples/`](examples/). ## Contributing

  查看 README 缓存打开 GitHub
• deeplethe/solid-glassTypeScript · 2,068 Stars

  界面主题、AI界面组件

  Stars 统计更新于 2026/10/19 09:55:32

  查看原始资料

  作者描述

  A design system for AI applications. Material, physical, tactile, technical.

  编报时的 AI 中文解读

  Solid Glass 是一套为 AI 应用设计的视觉主题，基于 shadcn/ui 构建，提供厚重、有质感的玻璃质感界面风格，适用于需要清晰可读的 AI 交互界面。 它提供单一样式表，可无缝集成到现有 shadcn/ui 项目中，无需修改组件，并附带多个专为 AI 代理界面设计的组件。

  编报时的 README 依据

  the classes it already carries, so the components you have need no changes. Start from a shadcn/ui project (React 19, Tailwind CSS 4, the Base UI components): ```bash npx shadcn@latest init ``` Add the stylesheet, from the site or straight from this repository: ```bash npx shadcn@latest add https://deeplethe.github.io/solid-glass/r/solid-glass.json npx shadcn@latest add deeplethe/solid-glass/solid-glass ``` Import it after Tailwind and shadcn's stylesheet, and delete the `:root` and `.dark` colour variables that `shadcn init` wrote (Solid Glass defines the same names): ```css

  @import "tailwindcss"; @import "tw-animate-css"; @import "shadcn/tailwind.css"; @import "./solid-glass.css"; ``` Dark is the material as designed; light is there too. The [installation page](https://deeplethe.github.io/solid-glass/docs/installation) has the rest. Solid Glass's own components install the same way: ```bash npx shadcn@latest add https://deeplethe.github.io/solid-glass/r/status-glyph.json ``` | Component | What it is | |---|---| | `status-glyph` | Three points of light that say a turn is going, and what it is doing. |

  查看 README 缓存打开 GitHub
• ahwootech/BRUTALC# · 1,904 Stars

  游戏开发、引擎开发

  Stars 统计更新于 2026/10/19 09:55:33

  查看原始资料

  作者描述

  A free, open source C# framework for making games, or your own game engine. Direct access to native APIs and Vulkan, with the low-level libraries around them. The foundation under Kitten Space Agency and RimWorld's Waylith engine.

  编报时的 AI 中文解读

  BRUTAL 是一个免费、开源的 C# 框架，用于制作游戏或构建自己的游戏引擎。它提供对原生 API 和 Vulkan 的直接访问，并包含数学、内存管理、并发等底层库。适合希望拥有高度控制权、不依赖 Unity 等通用引擎的开发者。

  编报时的 README 依据

  At its core is a custom interop layer giving C# direct access to native APIs and libraries. Around that we've built mathematics, memory management, strings, collections, job scheduling, logging, localization, and bindings for Vulkan, GLFW, Dear ImGui, FMOD, RakNet, Jolt and Box2D. BRUTAL doesn't make the architectural decisions for you. You write your own stack. You can use whatever NuGet packages you want, including third-party ones.

  It's called BRUTAL for a reason. In some ways it's much harder to use than Unity, Unreal or Godot. In other ways, particularly if you're doing something unusual, it's considerably easier. You actually have to know what you want to do first. ## Why You are only as independent as your tools are.

  Each top-level folder in this repository is one library, published as a NuGet package. Reference only the ones you need. | Library | What it is | | --- | --- | | `Brutal.Core` | Base types, attributes, platform detection, pointer helpers, Roslyn analyzers | | `Brutal.Interop` | The custom interop layer that gives C# direct access to native APIs | | `Brutal.Memory` | Unmanaged memory management and allocators | | `Brutal.Collections` | Collections that don't allocate | | `Brutal.Strings` | String building and handling without allocation |

  - **A foundation, like a kernel.** Think of the Linux kernel and something like Ubuntu. It's not an exact comparison, but it gets the idea across. We provide the foundation and you build your own systems on top. ### What BRUTAL is not BRUTAL is not trying to be a general-purpose engine. There is no editor, no scene graph and no renderer, and we're not adding them. We push back on efforts to add more, because adding everything is how the industry ended up with the kind of engines we don't want to use. Pull requests that drift away from a clear purpose will be rejected.

  | `Brutal.Monitor.Common`, `Brutal.Monitor.Host`, `Brutal.RenderDoc`, `Brutal.RenderDoc.ImGui` | Runtime monitoring and RenderDoc capture integration | | `Brutal.NativeFileDialog`, `Brutal.NotCurses`, `Brutal.ZLib` | Native file dialogs, terminal UI, compression | Third-party notices for each library live in its `LICENSES` folder, `Brutal.ImGui/LICENSES` say. ## Getting started ### Prerequisites - [.NET 10 SDK](https://dotnet.microsoft.com/download) - A Vulkan 1.3 capable GPU with current drivers (if you render)

  - The [Vulkan SDK](https://vulkan.lunarg.com/) if you want validation layers while developing - Windows or Linux ### Your first window Create a project and add Core and Glfw: ```sh dotnet new console -n HelloBrutal cd HelloBrutal dotnet add package Ahwoo.Brutal.Core dotnet add package Ahwoo.Brutal.Glfw ``` Replace `Program.cs`: ```csharp using Brutal.GlfwApi; using Brutal.Numerics; if (!Glfw.Init()) throw new Exception("Failed to initialize GLFW"); Glfw.WindowHint(GlfwWindowHint.ClientApi, 0); var window = Glfw.CreateWindow(new GlfwWindow.CreateInfo { Title = "Hello BRUTAL",

  查看 README 缓存打开 GitHub
• Jakeschincariol/instagram-agent-skillPython · 1,880 Stars

  内容创作辅助、爆款研究、文本优化

  Stars 统计更新于 2026/10/19 09:00:30

  查看原始资料

  作者描述

  Thirteen Claude skills that run an Instagram account: Reels off 26 scored hook formulas, a viral research skill, a caption linter, and a humanizer. Free, MIT.

  编报时的 AI 中文解读

  提供13个Claude技能，用于运营Instagram账号，包括Reels脚本生成、爆款研究、标题优化、内容人性化处理等，适合内容创作者使用。

  编报时的 README 依据

  And one is the humanizer, which is the reason the rest are usable. It strips the em dashes, the slop vocabulary and the invisible watermark characters out of a draft, then scores what is left against a five-check panel before you ever see it. **Nothing gets posted until you say yes.** These skills write. You post. ## Install Paste this into Claude: ``` https://github.com/Jakeschincariol/instagram-agent-skill Install this skill, then confirm /ig-reel works. ``` Or do it yourself, in Claude Code: ```bash git clone https://github.com/Jakeschincariol/instagram-agent-skill.git

  cp -r instagram-agent-skill/skills/ig-* ~/.claude/skills/ ``` Or as a plugin: ``` /plugin marketplace add Jakeschincariol/instagram-agent-skill /plugin install instagram-agent ``` Project-local instead of global: copy the same folders into your repo's `.claude/skills/`. No Claude Code at all? Paste any single `SKILL.md` at the top of a chat and it runs as a mode. You lose the five Python tools, which is most of the point of `/ig-reel` and `/ig-human`, but the rest works. Then spend ten minutes on `templates/voice.md`. Copy it to

  | `/ig-dm` | The keyword delivery, the first message, the collab pitch, and the two follow-ups. Two. | | `/ig-repurpose` | One video, podcast or newsletter into a week of reels and carousels that each stand alone. | | `/ig-audit` | Post-mortem on what you already posted. Ranks by outlier multiple and sends per reach, not views. | ## The five tools that actually run No dependencies, no network, nothing uploaded. They run on your machine, on your text. ### Hooks ```bash python3 hookscore.py hooks.txt # rank your options

  dollars" and "three marketing books" scored as having nothing concrete in them. Spoken hooks say their numbers out loud. If you re-run this on a bigger or cleaner corpus and get a different answer, I would rather know. The measurement script is not in the repo because it depends on `yt-dlp`, but the method is four lines and is written out in [`/ig-viral`](skills/ig-viral/SKILL.md). ## The fine print, which is the honest part **These skills do not post to Instagram.** There is a real Content Publishing API for Professional accounts, and it needs a Meta developer app, a linked

  The one exception is keyword auto-replies in DMs, which Instagram supports through its own tools and approved partners, and which only fire after somebody comments first. `/ig-dm` says where that line is. **`/ig-viral` reads, it does not scrape.** Ten accounts, a dozen reels each, at human speed, with you driving your own browser. It never asks for your password and never logs in as you. Automated collection at volume is the thing that gets accounts restricted, and a crawler is not what this is. **The five detection checks are local heuristics, not detector APIs.** They are

  modelled on the signals public detectors key on and they run entirely on your machine. They are not GPTZero, Originality, Copyleaks, Winston or Turnitin, they do not call those services, and they cannot promise those verdicts. Fixing what they measure tends to move those numbers, because they are measuring the same underlying things. That is the whole claim. Nobody can honestly sell you "undetectable", and anybody who does is selling you something. **The invisible-character pass is real and it is narrow.** It removes the

  查看 README 缓存打开 GitHub
• obsidianmd/peleTypeScript · 1,878 Stars

  图表渲染

  Stars 统计更新于 2026/10/19 09:55:33

  查看原始资料

  作者描述

  Renders Mermaid diagrams faster than you'd run from a volcano! 🌋

  编报时的 AI 中文解读

  Pele 是一个将 Mermaid 图表渲染为 SVG 的库，支持 Mermaid 12.1.0 的所有图表类型，适合需要在网页或应用中嵌入图表渲染功能的开发者。

  编报时的 README 依据

  # Pele Pele renders [Mermaid](https://mermaid.js.org) diagrams to SVG. It is small, synchronous, has no runtime dependencies, and uses CSS variables for colors and fonts. Pele supports every diagram type built into Mermaid 12.1.0. Documentation, examples, and a playground are at [pele.run](https://pele.run). ## Install ```sh npm install pele ``` Or load it [from a CDN](https://pele.run/api#using-a-cdn): ```html <script src="https://cdn.jsdelivr.net/npm/pele@1/dist/pele.global.js"></script> <script>

  for (const el of document.querySelectorAll('pre.mermaid')) pele.mount(el, el.textContent); </script> ``` ## Use ```ts import { render } from 'pele'; const { svg } = render(`flowchart LR A[Mermaid text] --> B[SVG]`); document.querySelector('#diagram').innerHTML = svg; ``` In a browser, `mount()` renders into an element, reads fonts from CSS, and adapts the layout to the container width: ```ts import { mount } from 'pele'; mount(document.querySelector('#diagram'), text); ``` See the [API](https://pele.run/api) for options and error handling.

  Pele targets Mermaid 12.1.0 syntax, with its own layout and visual style. Its parsers are tested against Mermaid's parser specs and with differential fuzzing. ## Develop ```sh npm install npm test npm run build npm run bench ``` The documentation site and playground are in `website/` and use Node 24 and pnpm 12. With nvm: ```sh cd website nvm use pnpm install pnpm dev ``` After installing the website dependencies, run `npm run test:website` from the repository root to check the playground. See [Contributing](CONTRIBUTING.md) for parser testing. ## License

  查看 README 缓存打开 GitHub
• SubmitGame/awesome-decomp-gamesPython · 1,765 Stars

  游戏资源导航、浏览器游玩

  Stars 统计更新于 2026/10/19 09:55:33

  查看原始资料

  作者描述

  Rated, auto-refreshed list of decompiled, recompiled & reverse-engineered games: decomps, static recomps, source ports, engine remakes, browser ports.

  编报时的 AI 中文解读

  这是一个自动刷新的游戏逆向工程资源列表，收录了反编译、静态重编译、源代码移植、引擎重制及浏览器移植等项目，方便玩家查找和筛选可玩的游戏版本。

  编报时的 README 依据

  41 games you can launch straight from a browser tab (WebAssembly/JS ports and recompilations). Links are re-checked daily in a real browser; ones that are taken down or broken are hidden. Some ask for files from your own copy of the game. Never install a launcher or extension a site pushes on you. <table>

  <!-- GENERATED FILE: edit data/ and run scripts/refresh, not this file --> # Awesome Decomp Games [](https://awesome.re) > A rated, auto-refreshed list of **decompiled, recompiled and reverse-engineered games**: matching decompilations, static recompilations, source ports, engine remakes, disassemblies and browser ports. **1457 projects** covering **1245 games** · updated 2026-10-11 · browse and filter on the **[website](https://submitgame.github.io/awesome-decomp-games/)** · data in [`data/games.json`](data/games.json) > [!IMPORTANT]

  > This list only links to source code and project pages. It does **not** host or link to ROMs, ISOs, game assets or other copyrighted material. Most projects need files from a copy of the game you own. See the [disclaimer](#disclaimer). ## Contents - [Play in your browser](#play-in-your-browser) (41) - [Top playable picks](#top-playable-picks) - [Static recompilations](#static-recompilations) (347) - [Source ports](#source-ports) (79) - [Engine remakes & reimplementations](#engine-remakes--reimplementations) (162) - [Browser ports](#browser-ports) (46)

  查看 README 缓存打开 GitHub
• chenglou/handwiredTypeScript · 1,732 Stars

  代码转神经网络、GPU并行化

  查看原始资料

  作者描述

  Turn your regular JavaScript code into a neural net.

  编报时的 AI 中文解读

  这是一个将普通JavaScript代码转换为神经网络表示的实验性工具，无需训练即可保持精确且可微分，并支持在JS或WebGPU上运行。适合对程序可微分化和GPU并行化感兴趣的开发者探索。

  查看 README 缓存打开 GitHub
• Folkbench/folkbenchRust · 1,690 Stars

  模型接口转发、用量统计

  查看原始资料

  作者描述

  AI API 中转站评测排行与切换工具：实测可用率、延迟与价格 · Open-source AI API relay benchmark, rankings & switcher for Claude Code / Codex

  编报时的 AI 中文解读

  Folkbench 是一个提供 AI 中转站评测排行和配套桌面客户端 Switch 的项目，用于帮助用户选择稳定的中转站，并一键切换 Claude Code、Codex 等工具的配置。

  查看 README 缓存打开 GitHub
• NVIDIA-Omniverse/usd-brepC++ · 1,667 Stars

  BRep建模、USD资产修复

  查看原始资料

  作者描述

  USD BRep is a standalone C++ and Python library for NURBS-based boundary representation (BRep) solid modeling

  编报时的 AI 中文解读

  USD BRep 是一个独立的 C++ 和 Python 库，用于基于 NURBS 的边界表示（BRep）实体建模，并提供修复 USD 资产中 BRep 几何的工具。

  查看 README 缓存打开 GitHub
• leemysw/yovoiceTypeScript · 1,661 Stars

  语音合成、语音克隆、音乐生成

  查看原始资料

  作者描述

  Open-source local AI audio & music studio for macOS & Windows. TTS, voice cloning, songs with lyrics, and multi-track scores — 100% offline, zero API fees.

  编报时的 AI 中文解读

  yovoice 是一款开源的本地语音创作工具，支持 macOS 和 Windows，可将文本转换为自然语音，提供语音克隆、情感控制、音乐生成和乐谱编辑等功能，完全离线运行，无需 API 费用。

  查看 README 缓存打开 GitHub
• darrenprx/lolPingTypeScript · 1,656 Stars

  桌面快捷操作、表情轮盘

  查看原始资料

  作者描述

  League of Legends ping wheel for your Windows desktop. Hold Alt, drag, and ping anywhere, with the original animations and sounds.

  编报时的 AI 中文解读

  lolPing 是一个桌面工具，让用户在 Windows 和 macOS 上通过按住 Alt 或 Ctrl 并拖拽，在任意应用和显示器上使用英雄联盟的 ping 和表情轮盘，并带有原始动画和音效。

  查看 README 缓存打开 GitHub
• shaongitbd/dooryTypeScript · 1,608 Stars

  图像编辑、视频剪辑、PDF处理

  查看原始资料

  作者描述

  Open source canva alternative

  编报时的 AI 中文解读

  Doory 是一个开源的在线设计编辑器，功能类似 Canva，可用于制作海报、社交媒体帖子、演示文稿、简历等，支持从模板或空白页开始编辑，并导出为多种格式。

  查看 README 缓存打开 GitHub
• opengeos/RustGISRust · 1,585 Stars

  地理数据分析、地图制图、地理数据共享

  查看原始资料

  作者描述

  A fast, free and open-source GIS in pure Rust.

  编报时的 AI 中文解读

  RustGIS 是一个用 Rust 编写的免费开源 GIS 应用，用于制图、分析和共享地理空间数据，支持桌面端和网页端。

  查看 README 缓存打开 GitHub
• krishhgg/InsomniaSwift · 1,566 Stars

  定时唤醒、合盖管理

  查看原始资料

  作者描述

  A macOS menu bar app that safely keeps your MacBook awake with the lid closed, for timed agent sessions.

  编报时的 AI 中文解读

  Insomnia 是一款 macOS 菜单栏应用，用于在合盖状态下安全地保持 MacBook 唤醒，并支持设定会话时长，适用于需要定时保持唤醒的代理会话场景。

  查看 README 缓存打开 GitHub
• perplexityai/SPACERust · 1,550 Stars

  沙箱代码执行

  查看原始资料

  作者描述

  Sandboxed Platform for Agentic Code Execution: pluggable-runtime sandboxes for running agent and user code in isolation.

  编报时的 AI 中文解读

  SPACE 是一个用于在隔离的 Firecracker 微虚拟机中运行代码的沙箱平台，应用程序通过 HTTP API 创建 Linux 环境、执行命令、收集输出并管理沙箱生命周期。适用于需要为代理、用户工作区或代码运行任务提供独立执行环境的开发者。

  查看 README 缓存打开 GitHub
• nano-muse/nanoMuseSwift · 1,478 Stars

  个人智能体、跨设备任务、模型服务接入

  查看原始资料

  作者描述

  nanoMuse: an open-source personal agent for every device you own. One agent with a name and a look of its own that does things, keeps working while the app is closed, remembers you, and asks before anything you cannot undo. On Android, iPhone and iPad, Windows / macOS / Linux and in the browser; your phone's screen and your computer as its hands.

  编报时的 AI 中文解读

  nanoMuse 是开源个人智能体，可在手机、桌面和浏览器上运行，能操作设备、跨设备派发任务、定时执行并在不可逆操作前询问。适合希望在自己设备上部署个人智能体的用户与开发者。 整套包含手机应用、桌面应用、Web 控制台和连接它们的 relay，均可自行部署；登录后可使用社区 relay 的免费模型额度，额度用尽后可改用自有密钥。

  查看 README 缓存打开 GitHub
• Supermedo/uncharted_pcC++ · 1,454 Stars

  游戏运行

  查看原始资料

  作者描述

  Uncharted: The Nathan Drake Collection running natively on Windows - by SuperMedo

  编报时的 AI 中文解读

  在Windows上原生运行《神秘海域：德雷克合集》三款游戏的启动器，需要用户自备PS4游戏解密文件。

  查看 README 缓存打开 GitHub
• Swordpdf/PS5SX2C++ · 1,408 Stars

  PS2游戏模拟、游戏存档管理、在线联机

  查看原始资料

  作者描述

  A port of PCSX2 for PS5

  编报时的 AI 中文解读

  PS5SX2 是 PCSX2 的 PS5 移植版，在越狱的 PS5 上原生运行 PS2 游戏，支持高分辨率、宽屏补丁、RetroAchievements 和在线联机，适合拥有越狱 PS5 并希望在大屏幕上重温 PS2 游戏的玩家。

  查看 README 缓存打开 GitHub
• youpi-pg/YouPI-PGJavaScript · 1,330 Stars

  支付网关、支付验证

  查看原始资料

  作者描述

  A self-hosted UPI & COD payment gateway with zero per-transaction platform fees, universal JavaScript checkout integration, dynamic QR payments, WhatsApp OTP authentication, and automated payment verification.

  编报时的 AI 中文解读

  YOUPI PG V2 是一个自托管的 UPI 支付网关，面向独立商户，提供零交易费、直接银行结算、WhatsApp OTP 验证、动态二维码支付和自动支付确认功能。 它通过 Android 应用、Gmail 和 Google Messages 观察银行通知，实现支付验证，并支持在线销售和线下柜台两种模式。

  查看 README 缓存打开 GitHub
• peiikai/LASRecompC · 1,302 Stars

  游戏运行、随机化游戏

  查看原始资料

  作者描述

  Static recompilation of Link's Awakening (Switch) to native macOS, Windows and Linux

  编报时的 AI 中文解读

  将《塞尔达传说：织梦岛》（Switch版）静态重编译为原生Windows、macOS和Linux应用，支持原版游戏和多种随机化模式，需要用户自备合法游戏文件。

  查看 README 缓存打开 GitHub
• dofastted/vm2apiJavaScript · 1,262 Stars

  模型服务接入、模型接口转发

  查看原始资料

  作者描述

  vm2api 使用虚拟机完全拟真方案的订阅转api

  编报时的 AI 中文解读

  vm2api 是一个将 Anthropic (Claude Pro/Team/Enterprise/Max) 和 OpenAI (ChatGPT/Codex) 订阅转化为标准 API 的虚拟机级生产网关，通过完全隔离的VM环境和官方真实客户端进程实现稳定、长效的订阅转API服务。适用于需要将AI订阅用作标准API的组织、开发者或高级用户。

  查看 README 缓存打开 GitHub
• trhhubsd/ARTEXGo · 1,255 Stars

  自主渗透测试、资产同步、漏洞复测

  查看原始资料

  作者描述

  ARTEX · AI autonomous penetration testing system (multi-agent LLM, Go + Next.js) · AI 自主渗透测试系统 · 百度 agent+ 攻防挑战赛冠军

  编报时的 AI 中文解读

  ARTEX 是一个基于 LLM 多智能体的自主渗透测试系统，提供仪表盘、任务管理、资产覆盖图、流量录制、人在环路对话等可视化界面，适用于在本地隔离环境中进行安全研究和技术验证。

  查看 README 缓存打开 GitHub
• pch/rawmakaseRust · 1,217 Stars

  RAW处理、照片调色、照片管理

  查看原始资料

  作者描述

  📸 Free, fast, Lightroom-compatible RAW photo editor for Linux, macOS, and Windows.

  编报时的 AI 中文解读

  RAWmakase 是一款跨平台的 RAW 照片编辑软件，提供非破坏性编辑、Lightroom 风格控制、目录管理和 Lightroom Classic 目录导入功能，适合需要处理 RAW 文件并希望迁移自 Lightroom 的用户。

  查看 README 缓存打开 GitHub
• devswha/herdr-web-uiTypeScript · 1,188 Stars

  远程代理管理、聊天辅助、推送通知

  查看原始资料

  作者描述

  Browser and phone client for herdr: chat and live terminal for every agent pane, remote PCs over SSH, web push alerts

  编报时的 AI 中文解读

  这是一个浏览器和手机客户端，用于连接 herdr 代理会话，支持聊天、实时终端、审批和推送通知，方便远程管理 AI 代理。

  查看 README 缓存打开 GitHub
• amontlabs/lcuJavaScript · 1,178 Stars

  桌面应用控制、浏览器控制、Harness集成

  查看原始资料

  作者描述

  Codex computer use, decoupled from the app, for usage inside any harness.

  编报时的 AI 中文解读

  LCU 是一个将 Codex 的计算机使用能力从官方 ChatGPT 桌面应用中解耦出来的工具，允许在任何 harness 中使用，无需 Codex 认证。它提供桌面应用控制、Chrome 标签页控制等功能，并支持多种 harness 适配器。

  查看 README 缓存打开 GitHub
• fyzugx/DEskC · 1,170 Stars

  桌面环境

  查看原始资料

  编报时的 AI 中文解读

  DEsk 是一个为 Linux 设计的 Wayland 桌面环境，采用拟物化风格，将桌面呈现为一张胡桃木书桌，窗口像纸张、程序在书架上、通知在邮箱里。适合追求独特视觉体验和拟物化交互的用户。

  查看 README 缓存打开 GitHub

其余 194 个项目见收录列表。

查看本期全部收录项目