读论文、看直播、玩系统：今天这几个开源项目各有各的讲究

收录数按本期首次达标收录的项目统计。

• tek-sys-hub/Rubix-langAssembly · 1,902 Stars

  系统编程、编译

  Stars 统计更新于 2026/10/23 07:55:31

  查看原始资料

  作者描述

  The 100% Pure x86-64 Assembly Systems Language

  编报时的 AI 中文解读

  Rubix 是一门完全用 x86-64 汇编编写的系统编程语言，其编译器不依赖任何外部运行时库，可将 .ix 源码编译为直接运行于 Linux、Windows、macOS 的原生可执行文件，适合系统级编程与教学研究。

  编报时的 README 依据

  <br clear="left"/> --- ## ⚡ Highlights - **Pure Assembly Compiler**: The compiler itself is written completely in GNU x86-64 assembly (`as` + `ld`). No external dependencies or runtime libraries are needed. - **Tiny & Instant**: The compiler binary is only **~42 KB**, with sub-millisecond compilation times. - **Zero OS Barrier (Universal Multi-OS Architecture)**: - Compiles `.ix` code into standalone native executables equipped with an adaptive runtime.

  - **No WASM, No Virtual Machine**: Architected to execute directly on bare silicon across **Linux, Windows, and macOS** (verified on Linux x86-64). - **Dynamic Host OS Detection**: Automatically determines the host OS at binary entry (`rubix_host_os`). - **In-Memory Windows Resolver**: Walks the Windows PEB (`gs:[0x60]`) in pure assembly to resolve `kernel32.dll` API exports without external import tables. - **Rich Language Features**: - `for` and `while` loops with `break`, `continue`, and arbitrary nesting

  ### 5. Separate Compilation & Static Object Linking Compile individual `.ix` files into ELF relocatable objects (`.o`) and link multiple units: ```bash ./rubixc -c module1.ix -o module1.o ./rubixc -c module2.ix -o module2.o ./rubixc module1.o module2.o -o my_app ``` --- ## 🛠️ Rich Diagnostic Error Reporting Rubix produces clear, informative compile-time diagnostics pointing directly to the offending token: ``` Rubix Compiler Error: Expected expression --> line 4 | 4 | let x = ; | ^ ``` --- ## 🚀 Quickstart & Usage ### Prerequisites - Linux OS (x86-64)

  - GNU Assembler (`as`) and Linker (`ld`) (available via standard `binutils`) ### 1. Build the Compiler ```bash make clean && make ``` Produces `rubix_compiler` and `lib/runtime.o`. ### 2. Compile and Run a Program Use the `./rubixc` compiler driver: ```bash # Compile and immediately run ./rubixc -r examples/15_floats.ix ./rubixc -r examples/16_structs.ix ./rubixc -r examples/17_native_floats.ix # Compile for Windows PE64 ./rubixc -t windows examples/01_hello.ix -o hello.exe # Compile with DWARF debug info ./rubixc -g examples/01_hello.ix -o hello_dbg ``` ### 3. Run the Entire Test Suite

  ```bash ./test.sh ``` Executes all **18** test suites verifying the compiler build, arithmetic, control flow, functions, prime algorithms, dynamic arrays, modules, loops, standard library, floats, structs, native float SSE2 operators, Windows cross-linking, and separate compilation. --- ## 📜 License Created with precision in pure x86-64 Assembly. Free and open source.

  查看 README 缓存打开 GitHub
• yangbin1322/hongguo-player未标注语言 · 1,064 Stars

  短剧播放、追更提醒、弹幕互动

  Stars 统计更新于 2026/10/23 07:35:55

  查看原始资料

  作者描述

  红果短剧电脑版 · 不用安卓模拟器，直接在 Windows / macOS 上刷红果短剧 · 国内下载: hongguo.235698.xyz

  编报时的 AI 中文解读

  为红果短剧平台开发的非官方桌面客户端，让用户在 Windows 和 macOS 电脑上无需安卓模拟器即可直接观看短剧，并支持自动连播、追更提醒、弹幕评论等便捷功能，适合希望在电脑上沉浸刷剧的用户。

  编报时的 README 依据

  <div align="center"> # 红果短剧电脑版 ### 不用安卓模拟器，直接在 Windows / macOS 上刷红果短剧 专为电脑设计的红果短剧桌面客户端 **沉浸式刷剧 · 自动连播 · 竖屏模式 · 追更提醒 · 弹幕评论 · 小窗模式 · 搜索与排行榜** **[⬇️ Windows 下载](https://hongguo.235698.xyz/download/latest/windows-amd64)**　　**[🍎 macOS 下载](https://hongguo.235698.xyz/download/latest/darwin-universal)** Windows 10 / 11 (64 位) · macOS (Intel 与 Apple 芯片通用) · 免费无广告 下载页 (国内可直接打开): **https://hongguo.235698.xyz** · 也可以到 [Releases](../../releases/latest) 下载  </div> ## 🖥️ 为什么做这个? 手机刷短剧很方便. 但已经坐在电脑前的时候, 为了看几集还要拿起手机, 或者专门开一个安卓模拟器, 多少有点麻烦.

  所以做了这个真正适合电脑用的红果短剧客户端: 鼠标滚轮刷剧, 方向键切换, 看完自动下一集, 还能开个小窗边工作边追剧. ## ✨ 主要功能 - 🎬 **沉浸式刷剧**: 推荐 / 分类刷剧, 鼠标滚轮或方向键切换, 边刷边预加载 - ▶️ **自动连播**: 看完自动播放下一集、下一季, 记住上次看到哪 - 🔍 **找剧方便**: 搜索 (带输入联想)、多维度筛选 (含已完结 / 连载中)、排行榜、新剧推荐与上新日历 - 📌 **追更提醒**: 追的剧更新了、预约的剧上线了, 打开程序就提醒你 - 📱 **竖屏模式**: 竖屏短剧时窗口收窄到刚好包住画面, 不留两侧黑边 - ✨ **画质**: 默认最高画质, 可开画质增强 (锐化 / 漫剧动漫增强 Anime4K); 画面和窗口比例接近时自动铺满 - 💬 **互动**: 弹幕 (点赞 / 回复 / +1)、评论 (在画面右侧展开) 与剧评, 和手机端互通 - ❤️ **追剧管理**: 收藏、点赞 (可按真人剧 / 漫剧 / AI剧筛选)、观看历史、预约, 都能搜索和批量删除 - ⌨️ **键盘快捷键**: 默认与抖音网页版一致, 可以自己改 - 🪟 **小窗模式**: 窗口置顶, 边工作边追剧; 隐身模式下鼠标移开自动暂停隐藏 - 🙈 **干净的画面**: 鼠标不动时控制条、点赞栏自动隐藏, 全屏看剧不挡画面

  | Windows 10 / 11 (64 位) | [下载 Windows 版](https://hongguo.235698.xyz/download/latest/windows-amd64) | `hongguo-版本号-windows-amd64.zip` | | macOS (Intel 与 Apple 芯片通用) | [下载 macOS 版](https://hongguo.235698.xyz/download/latest/darwin-universal) | `hongguo-版本号-macos-universal.zip` | 下载慢可以到本仓库的 [Releases](../../releases/latest) 下载同样的文件. **Windows** 1. 下载压缩包, **解压到一个文件夹**后双击 `红果短剧.exe`, 无需安装. 压缩包里还有: - `MicrosoftEdgeWebview2Setup.exe`: WebView2 运行时安装程序 - `HEVC视频扩展-双击安装.appx`: HEVC 视频扩展 - `使用说明.txt` 请和程序放在一起: 首次运行如果系统缺少 **WebView2 运行时** 或 **HEVC 视频扩展**, 程序会用它们自动安装 (WebView2 需要联网).

  2. 若提示 "Windows 已保护你的电脑", 点 **更多信息 → 仍要运行** (程序没有购买代码签名证书). 3. 以后有新版本时程序会提示, 点 "立即重启更新" 即可自动更新 (只替换 `红果短剧.exe`). > Releases 里的 `hongguo-版本号-windows-amd64.exe` 是主程序本身, 供自动更新使用; 新安装请下载 zip. **macOS** 1. 解压得到 `红果短剧.app`, 拖到 "应用程序". 2. 应用未签名, 首次打开如提示 "已损坏" 或 "无法验证开发者", 在终端执行: ``` xattr -cr /Applications/红果短剧.app ``` 然后再打开. ## ❓ 常见问题 **需要登录吗? 要钱吗?** 免费, 不登录也能刷剧; 收藏、点赞、观看历史未登录时保存在本机. 用手机号登录后与手机端红果 App 同步. **只有声音, 没有画面?** 1.1.5 起硬件解码不可用时会自动改用软件解码, 一般不会再出现; 旧版本请先更新. 仍有问题可按下面排查: 1. 点播放器提示条上的 **"重启程序"**. HEVC 视频扩展刚装上或被微软商店更新后, 要重启程序才能生效. 2. 仍不行就**重启电脑**再试.

  3. 扩展自动安装失败时, 双击程序目录里的 `HEVC视频扩展-双击安装.appx` 手动安装. 4. 远程桌面 / 虚拟机通常没有显卡硬件解码, 会使用软件解码 (CPU 占用较高); 画面花屏 / 黑屏时可在播放器的设置面板 (控制条上的齿轮) 里勾选 "软件解码". **提示 "无法连接本地播放服务"?** 程序在本机 (127.0.0.1) 起一个小服务给播放器送视频, 不连外网. 被防火墙或安全软件拦了就会这样: 按提示点 **一键放行** (需要管理员确认), 或在 Windows 防火墙 "允许应用通过防火墙" 里勾选 `红果短剧.exe`; 360 / 火绒等安全软件请把本程序加入信任列表. **遇到问题怎么反馈?** 在程序左侧点 **"反馈"** 提交 (可留邮箱, 解决后通知你; 作者回复后可以接着对话), 或到本仓库提 [Issue](../../issues), 也可以加 QQ 联系作者. ## ⚠️ 说明 本项目为个人开发的**非官方**第三方客户端, 与红果短剧官方无关. 仅供个人学习交流, 免费分享, 禁止倒卖. 本仓库只用于发布安装包, 不包含源代码. ## 👤 联系作者 - 作者: 洋滨 - 邮箱: yangbin1322@gmail.com - QQ: 3177762472 ## ⭐ 如果对你有帮助

  查看 README 缓存打开 GitHub
• Edwardxlai/easyreadPython · 1,010 Stars

  PDF翻译、文献管理、聊天辅助

  Stars 统计更新于 2026/10/23 07:31:07

  查看原始资料

  作者描述

  把英文论文读成舒服的中文：本地 PDF 论文翻译、原文对照、边读边问 AI、文献管理。Read English papers in comfortable Chinese.

  编报时的 AI 中文解读

  EasyRead 是一款本地运行的 PDF 论文阅读与翻译工具，支持逐页翻译、原文对照、边读边问 AI、文献管理等功能，适合需要阅读英文论文的中文用户。

  编报时的 README 依据

  <p align="center"> <img src="docs/images/logo.svg" width="72" alt="EasyRead"> </p> <h1 align="center">EasyRead</h1> <p align="center"><b>让你更舒服地读论文</b><br> 导入 PDF，后台逐页翻译；公式、表格照原文排好，随时对照原文，边读边划线、记笔记、提问。<br> 本地运行，文献库默认保存在电脑上，也可以放进你自己的网盘同步文件夹</p> <p align="center"><b>简体中文</b> · <a href="README.en.md">English</a></p> <p align="center"><a href="https://edwardxlai.github.io/easyread/demo/"><b>▶ 在线试读一篇</b></a> · <a href="https://edwardxlai.github.io/easyread/">项目主页</a> · <a href="https://github.com/Edwardxlai/easyread/releases/latest">下载</a></p> <p align="center">

  <p align="center"><img src="docs/images/pages.jpg" width="860" alt="译文和原页对照"></p> ## 它和“把 PDF 丢给翻译软件”有什么不一样 - **像读一本排好版的中文书。** 宋体正文、舒服的行宽和行距，公式用 KaTeX 按原文重排，表格是三线表，参考文献保留原文。顶栏一键切深色。 - **随时核对原文。** 一键切“对照”，每段下面附英文；右侧可以开原页，跟着阅读位置翻页，还会框出当前段落在原页的位置。 - **在论文里找内容。** 按 Ctrl / ⌘ + F 同时搜索原文和译文，显示匹配数、高亮结果，用 Enter / Shift + Enter 跳到下一处 / 上一处 - **翻译和解释分开。** 正文只放忠实的译文；AI 的解释、回答放在页边，一眼就能分清哪句是论文说的。 - **边读边问 AI。** 右侧“问 AI”面板实时对话，回答逐字流出来；可以一次引用好几段（选中文字拖进输入框就行）。问“我标红的那些公式有什么联系”，它会按颜色找出你的划线。可以开多个对话，模型单独选：Claude、GPT（Codex）、DeepSeek、通义、本机 Ollama……好的回答一键放到页边。

  默认逐篇翻译，单篇内部仍可分段并发。在“设置 → 模型 → 翻译”中可将“同时翻译几篇”设为 1–4 篇，重启后生效。多篇并行会叠加分段并发，增加内存和模型限流风险，建议保持默认 1 篇；详见[配置说明](docs/paper-concurrency.md) <p align="center"><img src="docs/images/settings.jpg" width="640" alt="设置"></p> ## 安装 **最省事：下载安装包**（不用装 Python）。在 [Releases](https://github.com/Edwardxlai/easyread/releases/latest) 下载： - **Windows**：`EasyRead-Setup-x.x.x.exe`，双击安装。没有代码签名，如果弹出“Windows 已保护你的电脑”，点“更多信息 → 仍要运行”。 - **macOS**（Apple 芯片）：`EasyRead-x.x.x-arm64.dmg`，把 EasyRead 拖进“应用程序”；第一次打开被系统拦截时，按下面对应的提示处理 - **Linux**：`EasyRead-x.x.x.AppImage`，`chmod +x` 后运行。

  EasyRead 自带命令行，Claude Code / Codex 这类 agent 可以在对话里直接读你的笔记和问题、把回答写到对应段落旁边，也可以亲自翻译或重译某几页。技能说明在 [`skill/paper-reading/SKILL.md`](skill/paper-reading/SKILL.md)，把这个目录放进 `~/.claude/skills/` 或 `~/.codex/skills/` 即可。 想让 agent 把一篇论文**讲透**而不是只译出来——还原作者真实的思考起点、把关键公式落到能自己算的例子、再把结论的边界查一遍——用 [`skill/paper-xray/SKILL.md`](skill/paper-xray/SKILL.md)。它读的是同一份 `library/<ID>/`，谈出来的东西照样写成锚在段落旁的讨论条目。 ```bash easyread list # 列出文献库 easyread import 论文.pdf # 或 arXiv 编号 / 链接 easyread status ID # 进度、我改过的译文、笔记、待回答的问题

  如果系统提示“将损坏你的电脑”，或明确检测出恶意软件并要求移到废纸篓，请停止安装，不要用上面的步骤绕过；按 [Apple 官方说明](https://support.apple.com/zh-cn/102445) 处理 <a id="run-from-source"></a> **或者从源码运行**：需要 [Python 3.10+](https://www.python.org/downloads/)。 先从 [Releases](https://github.com/Edwardxlai/easyread/releases/latest) 下载最新版的 zip 解压（或者 `git clone` 本仓库）。 **Windows**：双击 `start.cmd`。第一次会自动装好环境（一分钟左右），之后双击直接打开。 **macOS / Linux**：在解压出来的目录里运行 ```bash ./start.sh ``` 这样启动的，浏览器里的 EasyRead 页面全部关掉后，后台服务过十几秒会自己退出；还有翻译在跑的话，等译完再退。 每次推送都会在 Windows、macOS、Linux 上自动装一遍、跑测试、启动一次（见上面的“测试”徽章）。 **或者用 pip**（数据放在 `~/EasyRead`）： ```bash

  | **CodeBuddy CLI** | 按[官方 CLI 文档](https://www.codebuddy.cn/docs/cli/headless)安装并登录，账号需有 CLI 使用权益 | 不用 Key；模型名自己填，留空用 CLI 默认模型。[当前免费体验版不含 CLI](https://www.codebuddy.cn/docs/workbuddy/Pricing)，与 WorkBuddy 同账号共享积分 | | **Grok Build / Antigravity CLI / Cursor CLI** | 装好并登录 [Grok Build](https://x.ai/cli)、[Antigravity CLI](https://antigravity.google/docs/cli) 或 [Cursor CLI](https://cursor.com/docs/cli/installation) | 不用 Key，用各自的订阅；模型名自己填，留空用它的默认 | | **API 接口 · 国内直连**：DeepSeek / 智谱 / 阿里云百炼 / Kimi / 硅基流动 / 魔搭 | API Key | 智谱 GLM-4.7-Flash、硅基流动小模型免费；DeepSeek 一篇 20 页论文几毛钱 |

  1. **提示“无法验证开发者”或“Apple 无法验证是否包含恶意软件”**：先点“完成”（不要点“移到废纸篓”），打开“系统设置 → 隐私与安全性”，向下找到安全性提示，点“仍要打开”；在确认窗口点“打开”，按提示输入登录密码或使用 Touch ID 2. **提示“已损坏，无法打开”**：先从 [EasyRead 官方 Releases](https://github.com/Edwardxlai/easyread/releases/latest) 重新下载，并把应用拖进“应用程序”；文件也可能确实损坏或被改动，不能只凭这条提示判断原因。只有确认文件来自这个官方仓库、并信任该文件后，才在“终端”执行下面的命令，再重新打开 EasyRead；这条命令移除隔离属性，不会修复损坏的文件 ```bash xattr -dr com.apple.quarantine /Applications/EasyRead.app ``` 3. **为什么会出现验证提示**：目前的 macOS 安装包用的是项目自己的自签名证书（为了应用内更新），不是 Apple 开发者证书，也没有经过公证，因此可能被系统拦截；代码完全开源，也可以[从源码运行](#run-from-source)

  **数据存在哪？** 默认保存在本机；在“设置 → 云文献库”里可以迁移到其他磁盘或网盘同步文件夹，上传和同步由你自己的网盘客户端完成。翻译和提问时，相关内容会发给你选择的模型。Windows 安装版在安装目录的 `data\library\`（1.3.1 及以前放在 `~/EasyRead`，升级后第一次打开会自动搬过去）；从源码运行时在项目目录的 `library/`；pip 安装后在 `~/EasyRead/library/`。每篇论文一个文件夹，里面是原 PDF、原页图和几个 JSON（译文、你的笔记、AI 讨论、对话记录），设置和界面偏好在 `config.json`、`prefs.json`。可以直接备份或同步。为什么不用数据库见 [docs/data-format.md](docs/data-format.md#为什么用-json-文件而不是数据库)。

  查看 README 缓存打开 GitHub
• colinhacks/reyloRust · 655 Stars

  SQL导入、类型推断

  Stars 统计更新于 2026/10/23 07:50:55

  查看原始资料

  作者描述

  Import .sql files into TypeScript with full type inference and zero codegen.

  编报时的 AI 中文解读

  Reylo 是一个让 TypeScript 代码直接导入 .sql 文件的工具，通过类型推断将 SQL 查询的类型安全地集成到 TypeScript 编译中，无需生成代码。适合希望将数据库查询与类型系统紧密结合的开发者。

  编报时的 README 依据

  - ⚡ **No codegen** — the type checker asks Reylo directly: no generated files, no watcher. ## Setup To install Reylo, you'll need: - **TypeScript 7.1 (currently in beta).** The first release with content mappers. - **Node 24 or later.** - **A running dev database.** Postgres, MySQL 8.0.22 or later, or a SQLite file: Reylo reads your current schema from it. On Postgres and SQLite, [a schema file](https://reylo.dev/#schema-files) will do instead. Install Reylo, the engine for your database and TypeScript 7.1: ```sh npm install reylo npm install -D @reylosql/postgres # or mysql, sqlite

  npm install -D typescript@beta ``` Add Reylo to `tsconfig.json`: ```jsonc // tsconfig.json { "compilerOptions": { /* ... */ }, "contentMappers": [ { "package": "reylo", "extensions": [".sql"] } ] } ``` Then type-check with `--runExternalCode`, which lets `tsc` run content mappers: ```jsonc // package.json { "scripts": { "typecheck": "tsc --noEmit --runExternalCode" } } ``` Point Reylo at your dev database with `DATABASE_URL`, in your environment or in `.env`: ```sh # .env DATABASE_URL=postgres://localhost:5432/app # or mysql://…, or file:./app.db for SQLite ```

  To make `.sql` imports work at run time too, load Reylo where your code runs. On Node: ```sh node --env-file=.env --import reylo/register main.ts ``` For Bun, Nub, Vite, Next.js, esbuild, webpack, Rspack, Cloudflare Workers and Jest, add [the plugin for your setup](https://reylo.dev/#plugins), or copy its complete [example project](https://github.com/colinhacks/reylo/tree/main/examples). To see the types as you type, set up [your editor](https://reylo.dev/#editors). ## Usage Write a query in a `.sql` file, import it, and run it with your database driver: ```sql -- get_user.sql

  查看 README 缓存打开 GitHub
• Acelogic/SnapintoshPython · 544 Stars

  macOS启动、硬件驱动

  Stars 统计更新于 2026/10/23 07:50:55

  查看原始资料

  作者描述

  EL2 hypervisor and toolchain that boots macOS on the ASUS A16 (Snapdragon X2 Elite), with an Adreno Metal/IOGPU stack

  编报时的 AI 中文解读

  Snapintosh 是一个实验性的 EL2 监控程序（hypervisor）和配套工具链，用于在特定型号的 ASUS Zenbook A16 笔记本电脑上启动未修改的 macOS 27 内核，并驱动其 GPU、输入和 Wi-Fi 等硬件。该项目面向对在非 Apple 硬件上运行 macOS 感兴趣的开发者或高级用户。

  编报时的 README 依据

  - **The macOS 27 restore image (IPSW).** Download it from Apple with `python3 tools/fetch-ipsw.py <url> [destination]`. Then extract the J700 boot set (kernelcache, device tree, SPTM, TXM and others) with `python3 tools/extract-firmware.py <ipsw> j700ap`. Nothing runs without it. - **The Kernel Debug Kit** for 26A428, from Apple's developer downloads. - **Device firmware.** The Adreno GPU's GMU and SQE microcode (`gen80100_gmu.bin`, `gen80100_sqe.fw`) and the QCC2072 Wi-Fi firmware come from the linux-firmware project. The Wi-Fi board file

  [THIRD_PARTY_NOTICES.md](THIRD_PARTY_NOTICES.md) has the complete list. ## Documentation | Read | For | | --- | --- | | [docs/STATUS.md](docs/STATUS.md) | what works, known issues, roadmap | | [docs/ARCHITECTURE.md](docs/ARCHITECTURE.md) | how the layers fit together, including the GPU stack and Wi-Fi | | [docs/BUILDING.md](docs/BUILDING.md) | building the monitor, kexts, plug-in and compiler | | [docs/RUNNING.md](docs/RUNNING.md) | a boot cycle with `tools/cycle.py` | | [docs/KERNEL-PATCHES.md](docs/KERNEL-PATCHES.md) | the opt-in kernel patches and why each exists |

  | Storage | macOS is installed on, and boots from, the internal NVMe. | | Other hardware | Audio, cameras, Bluetooth, sleep and external displays are not supported. | [docs/STATUS.md](docs/STATUS.md) has the details, the known issues and the roadmap. ## Required hardware - An **ASUS Zenbook A16 with the Snapdragon X2 Elite X2E-96-100** (UX3607OA). Nothing else is supported. The monitor, device tables, clocks and register addresses are specific to this machine. - A **Mac host** running macOS 27, connected to the A16 over USB-C. q1n1 brings

  | `docs/` | design and reference documentation | ## Building You need: - a Mac with Xcode (macOS 27 SDK); - the Kernel Debug Kit for build 26A428, installed at `/Library/Developer/KDKs/KDK_27.0_26A428.kdk`; - Python 3; - Homebrew LLVM, Meson and Ninja for the shader compiler. Its Mesa checkout is fetched and patched by `gpu/air2ir3/tools/mesa-setup.sh`; - a q1n1 checkout next to this one (`../q1n1`). [docs/BUILDING.md](docs/BUILDING.md) has the steps. ## Bring your own Apple software This repository ships **no Apple, Qualcomm or ASUS binaries or firmware**. You supply them yourself:

  native QCC2072 Wi-Fi work, and NVMe storage through an emulated Apple storage controller. Snapintosh is used together with [q1n1](https://github.com/Acelogic/q1n1), a UEFI EL2 proxy forked from Asahi Linux's m1n1. q1n1 takes the machine at EL2 after UEFI and serves m1n1's proxy protocol over USB-C to a Mac host, which drives Snapintosh from there. ## Status This is experimental software, written for one laptop model. | Area | State | | --- | --- |

  查看 README 缓存打开 GitHub
• oosexgang/hackios-xnu-kernelC · 485 Stars

  iOS系统研究

  Stars 统计更新于 2026/10/23 07:50:55

  查看原始资料

  编报时的 AI 中文解读

  HackiOS 是一个研究项目，旨在将苹果的 XNU 内核和未修改的 iOS 16.5 用户空间引导到 Android 手机硬件（如 POCO X5 Pro）或 QEMU 虚拟机上运行，适用于对 iOS 系统底层运行机制感兴趣的研究人员和开发者。

  编报时的 README 依据

  # HackiOS — XNU kernel **HackiOS** boots Apple's open-source **XNU kernel** (`xnu-8796.121.2`, the kernel of iOS 16.5) on Android phone hardware and runs the **unmodified iOS 16.5 user space** on top of it: launchd → backboardd → SpringBoard → apps. Our own EFI loader, our own replacements for Apple's closed drivers, a UEFI port, and a Linux Wi-Fi driver on a compatibility layer make that possible. It runs in two places: - **POCO X5 Pro 5G (redwood, Snapdragon 778G)** — nothing is flashed: the UEFI is started with `fastboot boot`,

  the whole system lives in RAM, and after a reboot the phone is back in Android; - **QEMU** (`virt`, aarch64) — the same thing on any Linux computer, no phone needed. This repository is half of the project: | Repository | What | |---|---| | **hackios-xnu-kernel** (this one) | XNU patches, the EFI loader, the UEFI port, Wi-Fi, QEMU scripts, `xl`, prebuilt kernel/loader/UEFI, `common/` scripts |

  (Debian/Ubuntu: `git python3 7zip liblz4-dev gcc qemu-system-arm qemu-efi-aarch64`; 7-Zip must be 23 or newer — it reads APFS.) ```sh git clone https://github.com/oosexgang/hackios-xnu-kernel.git hackios cd hackios common/setup.sh # brings hackios-xnu-userland into the same tree, checks the host ios165/fetch.sh # downloads 7.4 GB from Apple's CDN and extracts what is needed (5-15 min) common/build.sh qemu-virt # builds the APFS images (and the kernel, if the toolchain is there) common/run.sh qemu-virt # a 20-minute session in a QEMU window ```

  The lock screen appears after 2–3 minutes. - **Mouse with the left button held = finger.** Unlock / leave an app: swipe up from the very bottom edge; back: swipe from the left edge. - Keys: `H` home, `L` lock/power, `↑`/`↓` volume, `F11` (hold) five fingers. - From a script: `qemu/touch.py swipe 0.5 0.985 0.5 0.3 250`, `qemu/touch.py tap 0.615 0.71`, `qemu/touch.py key home`; screenshot: `qemu/screendump.sh shot.png`. - Right after SpringBoard appears the system is still registering apps: give it 30–40 s.

  common/build.sh redwood GL=1 # images + out/redwood-boot.img common/run.sh redwood 900 # loader into RAM, system image into RAM, a 15-minute session ``` or step by step with `xl`: ```sh ./xl enter # Android -> fastboot -> our loader in RAM (fastboot boot); it brings up its own USB link ./xl ramdisk # upload the system image into the phone's RAM (2.5 GB, ~80 s), once per loader session XL_SECS=900 XL_CMDLINE="touchscreen=1 cpufreq=1 bootcpu=7 cpus=8 gpu=1 wifi=1 adsp=1" ./xl run ./xl screenshot shot.png ./xl reboot # back to Android ```

  It is a research bench, not a ROM: a session lasts a given number of seconds, then the kernel returns to the loader and hands over its log. ## Quick start: QEMU, no kernel build `prebuilt/` holds the kernel, loader and our programs built from these repositories, so all you need is the system image, which is made on your machine from Apple's IPSW. Needs: Linux x86-64, ~12 GB free disk, 4 GB free RAM. Packages (Arch): ```sh sudo pacman -S --needed git python 7zip lz4 gcc qemu-system-aarch64 qemu-ui-gtk edk2-aarch64 ```

  - Debian/Ubuntu UEFI firmware: `QEMU_EFI=/usr/share/AAVMF/AAVMF_CODE.fd QEMU_VARS=/usr/share/AAVMF/AAVMF_VARS.fd`. ## On the phone (POCO X5 Pro 5G) You need a **POCO X5 Pro 5G (redwood)** with an unlocked bootloader, `adb` and `fastboot`, and the images from the section above. Nothing is written to the phone's storage. **The phone's own firmware** is not in the repository and cannot be: it is the vendor's and is taken from the very phone it will run on, from its Android: ```sh adb pull /vendor/firmware/a660_{gmu.bin,sqe.fw,zap.mdt,zap.b00,zap.b01,zap.b02} fw-gpu/ # GPU (gpu=1)

  <img src="docs/lockscreen.jpg" width="190" alt="Lock screen"> <img src="docs/homescreen.jpg" width="190" alt="Home screen"> <img src="docs/control-center.jpg" width="190" alt="Control Center"> <img src="docs/settings.jpg" width="190" alt="Settings"> </p> Photographed on the phone, the Adreno GPU drawing (`gpu=1`). More in [Screenshots](#screenshots). ## What works | Works | Does not work (yet) | |---|---| | Lock screen, swipe to unlock, home screen with all icons (two pages), stock iOS 16 wallpapers | Cellular, Bluetooth |

  | System gestures: swipe up (home, app switcher), edge swipe (back), Control Center | Charging under our kernel (the phone only discharges) | | Apps: Settings, Clock, Calendar, Photos, Notes, Safari and more; widgets (Weather, Calendar) on the phone | Metal for apps (in progress: 293 of 456 system shaders already run on the Adreno, see `docs/air.md` in hackios-xnu-userland) | | Phone touchscreen (Goodix GT9916), multitouch, wake on touch; real power and volume keys | App Store content and Apple ID sign-in (no Secure Enclave; FairPlay daemon exits) |

  查看 README 缓存打开 GitHub
• wesbos/3d-model-anythingSwift · 468 Stars

  3D房间重建、3D场景浏览

  Stars 统计更新于 2026/10/23 07:50:55

  查看原始资料

  作者描述

  Capture iPhone LiDAR and photos, reconstruct editable Three.js rooms, and explore them in a browser.

  编报时的 AI 中文解读

  该项目提供iPhone应用、Three.js查看器、AI编码代理技能和办公室演示，用于从iPhone LiDAR扫描和照片重建可编辑的3D房间场景，并在浏览器中探索。

  编报时的 README 依据

  The app records the evidence. The skill guides an agent through measurement, modeling and matched-camera visual comparisons. This is an iterative reconstruction workflow, not a one-click scan-to-perfect-model converter. The current workflow focuses on rooms and interiors. ## Run the viewer Use Node.js 22.12+ (or a newer supported release). ```sh npm ci npm run dev ``` Open http://localhost:5188. Explore the office in dollhouse, walking, overhead, wireframe and knolling views. The playground is at `/play.html?room=office-v4&mode=fpv`. ```sh npm run build npm run preview ```

  The demo includes the reconstructed office and selected textures. Original scans and photo collections are not distributed; Scan and Overlay are disabled for this example. Those modes remain available for your own rooms with reference assets. ## Capture your own space See [the iPhone app](apps/house-capture/README.md) for building with Xcode, device requirements, scanning and sharing. Scanning needs a LiDAR-equipped iPhone or iPad. Non-LiDAR devices can import and review scans.

  查看 README 缓存打开 GitHub
• Jakeschincariol/trader-skillPython · 376 Stars

  投资研究、新闻监控、预算管理

  Stars 统计更新于 2026/10/23 07:55:52

  查看原始资料

  作者描述

  11 Claude skills that copy the investors you trust, watch the news and Congress, size every trade to your budget, and never place an order without your yes.

  编报时的 AI 中文解读

  这是一套包含11个Claude技能的集合，用于辅助股票投资研究，包括复制投资者持仓、监控新闻和国会交易、管理预算和风险、生成订单等，所有交易需用户确认。

  编报时的 README 依据

  craziest one is the Pelosi skill: it reads the stock trades members of Congress disclose, the minute the filing posts. And no trade goes through until you say yes. Nothing in this pack can place an order on its own. Paper trading is the default. ## Install Paste this into Claude: ``` https://github.com/Jakeschincariol/trader-skill install skill ``` Or as a plugin, in Claude Code: ``` /plugin marketplace add Jakeschincariol/trader-skill /plugin install trader-skill@trader-skill ``` Claude Code namespaces plugin skills, so installed as a plugin they show up as

  `/trader-skill:trader-copycat` and so on. Copy the folders instead if you want plain `/trader-copycat`: ```bash git clone https://github.com/Jakeschincariol/trader-skill.git cp -r trader-skill/trader-* ~/.claude/skills/ ``` 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 Python tools, but the method works. The tools need Python 3.8 or newer. Nothing to pip install. SEC asks every

  script to identify itself: put your name and email in `trader/config.json` (`"sec_user_agent": "Your Name you@example.com"`) or in the `SEC_USER_AGENT` environment variable. ## The eleven | command | job | what it does | | --- | --- | --- | | `/trader-copycat` | Reverse Engineer | Reads an investor's 13F-HR on SEC EDGAR and rebuilds their portfolio as weights you can copy, with what they bought, added, trimmed and sold since last quarter. | | `/trader-news` | News Monitor | Watches headlines and SEC filings (8-Ks) for every stock you hold and flags what touches them. |

  (Customize, Connectors, tool permissions), never "Always allow", so Claude's own "Allow once / Deny" card is a second check. No options, margin, short selling or leverage unless you turn them on. **Paper first.** `"paper": true` is the default in `config.example.json`, and Alpaca's MCP server defaults to its paper account. Keys live in environment variables, never in the chat. **13F data is old and partial.** Funds file 45 days after each quarter ends, and a 13F lists long positions in US-listed stocks and a few related securities:

  no shorts, no cash, no foreign-listed shares. You see what they held, not what they hold today. **Congress's trades are public, not inside information.** Members must disclose trades over $1,000 within 45 days (the STOCK Act). The Pelosi skill reads those public filings the minute they post, with the amounts as the ranges they are filed in. It has no access to anything that is not public. **The data comes from free public sources.** SEC EDGAR, the House Clerk, news RSS and Yahoo Finance's chart endpoint (unofficial, and it can change). Check

  查看 README 缓存打开 GitHub
• nuglifeleoji/SentryPython · 357 Stars

  故障恢复、经验学习

  Stars 统计更新于 2026/10/23 07:31:28

  查看原始资料

  作者描述

  Learning to Recover from LLM Agent Failures at Test Time

  编报时的 AI 中文解读

  Sentry 是一个为 LLM 代理提供外部运行时故障管理层的工具，旨在检测执行失败、引导恢复，并从验证成功的恢复中学习可复用的经验。它适用于希望提升代理在任务中可靠性的开发者。

  编报时的 README 依据

  git clone https://github.com/nuglifeleoji/Sentry.git cd Sentry python -m venv .venv source .venv/bin/activate python -m pip install -e . sentry-validate configs/paper.yaml ``` ### Integrate Sentry into Your Agent Loop Configure a model provider, then submit each completed agent-environment cycle to Sentry: ```python from Sentry import SentryRunner, agent_step_from_parts runner = SentryRunner.from_yaml("configs/providers/openai_compatible.yaml") runner.start_task( objective=task_objective, action_schema=environment_action_schema, ) try: step = agent_step_from_parts(

  step_id=0, reasoning=reasoning, raw_action=raw_action, tool_name=tool_name, tool_args=tool_args, observation=observation, parsed_ok=True, schema_valid=True, accepted_by_environment=True, ) repair = runner.step(step, terminal=False) if repair is not None: agent_messages.append(repair.prompt_text) finally: runner.finalize() ``` Your application remains responsible for generating, validating, and executing actions. See the integration guide for the full lifecycle and validity rules.

  - **37% average improvement** over the strongest runtime-intervention baseline across WebShop, AppWorld, SWE-bench Lite, and Mind2Web Replay. - **39% average improvement** over the strongest context-evolution baseline on held-out WebShop and Mind2Web Replay tasks. - **1.54× base-agent token usage**, versus up to 2.44× for other runtime methods, while saving 31.7s per task on SWE-bench Lite and 45.6s on AppWorld. - **81.7% recovery rate** across 939 detected failures. ## Quick Start ### Installation Sentry requires Python 3.10 or later. ```bash

  This repository currently contains the project overview and paper. We are finalizing our code and documentation and plan to release the full version in 2-3 weeks.

  查看 README 缓存打开 GitHub
• garycaogan/ai-live-avatarPython · 331 Stars

  AI直播调度、弹幕过滤、直播推流

  Stars 统计更新于 2026/10/23 07:50:55

  查看原始资料

  作者描述

  AI 实时直播间：Vidu S2 实时数字人 + 弹幕过滤调度 + 衣柜换装 + OBS 叠层（角色资产不含）

  编报时的 AI 中文解读

  让 AI 数字人自动开直播，处理弹幕过滤、调度、换装和动作，并可接入 OBS 或直接推流。适合想搭建 AI 直播间的开发者或主播。

  编报时的 README 依据

  # AI 实时直播间 一个让 AI 数字人自己开直播的小项目。观众发弹幕，程序先过滤、再排队，等她说完一句再交下一条：聊天她就回一句，换衣服就换一整套画面，比心、挥手这类就做动作。画面、声音和回复都由 [Vidu S2 实时数字人](https://platform.vidu.cn/vidu-stream/doc)生成，这个项目负责中间这一层：弹幕过滤和调度、角色和衣柜、直播间叠层、本地工作台，以及交给 OBS 或者直接推流。 这是 Gary（曹淦）做的 AI 实时直播测试，代码按 MIT 开源。测试时用的角色形象、人设、音色、素材库和录屏都是作者自己的角色资产，没有放进这个仓库，你需要用自己的。 ## 现在能做什么 | 已经能做 | 还做不到 | | --- | --- | | 聊天接话，人设和开场白可以自己写 | 换背景、吃喝和拿东西（默认关闭，原因见下） | | 换衣服（衣柜里预先做好的整套衣服，换一次大约 6 秒） | 唱歌 | | 小动作：比心、挥手、眨眼等，关键词可以在工作台改 | 站起来、全身画面 | | 没人说话时按角色自己的话题池暖场，一轮说完才重复 | 大幅度跳舞，坐着只能晃一晃 | | 低俗、引流弹幕自动过滤，礼物不插队 | 带货 | | Vidu 中途挂断后自动重新开播（要手动打开） | 跨场次记住老观众 |

  - 只要数字人画面：`http://127.0.0.1:28900/stage?obs=1&role=obs&layer=avatar` - 只要叠层（透明背景）：`http://127.0.0.1:28900/stage?overlay=1` - 画面加叠层：`http://127.0.0.1:28900/stage?obs=1&role=obs&layer=all`，宽 1080、高 1920 浏览器源要勾选“通过 OBS 控制音频”。服务只监听本机，所以 OBS 要和服务在同一台电脑上。`tools/obs_control.py` 可以通过 OBS WebSocket 一键搭好场景。 叠层里常驻“AI 实时生成画面”的标识，这是平台对 AI 数字人直播的要求，请保留。 ## 在抖音开播 抖音上常见的那种数字人直播间是横屏开播：抖音会把画面放进手机上方的框里，评论显示在框下面。下面这条命令把 OBS 画布设成 1296×1080（1.2:1，和横构图底图一样的比例），新建“抖音直播间”场景：数字人铺满画面，说明卡在上方两个角，字幕在下方，左下角是真人摄像头。 ```bash .venv/bin/python tools/obs_control.py setup douyin --camera FaceTime ```

  Evolink 的 Key 同样放到 `~/.deepsea/evolink/api.key`。YouTube 推流码放 `~/.deepsea/youtube/stream.key`，OBS WebSocket 密码放 `~/.deepsea/obs/ws.password`。 ## 安装 ```bash git clone https://github.com/garycaogan/ai-live-avatar.git ``` ```bash cd ai-live-avatar && python3 -m venv .venv && .venv/bin/pip install -r requirements.txt ``` 舞台页用到阿里云 RTC 的网页 SDK，版权属于阿里云，这里不重新分发，用脚本从官方 CDN 下载并校验： ```bash ./tools/fetch_vendor.sh ``` ```bash cp config.example.json config.json ``` ## 启动 ```bash .venv/bin/python server.py ``` 服务只监听本机，打开 http://127.0.0.1:28900/workbench 就是工作台。

  - macOS 或 Linux，Python 3.11 以上（作者用 3.13），ffmpeg（要带 libx264 和 aac）。 - Vidu 开放平台的 API Key（`vda_` 开头），账号里要有积分。实时数字人按时长计费，官方价每秒 1.5 积分，1 积分 0.03125 元，一小时大约 169 元。 - 可选：[Evolink](https://evolink.ai) 的 API Key，用来生成形象图、衣服、道具和背景，以及克隆声音时上传音频。 - 可选：OBS 30 以上，用来加真人摄像头、叠层和推流。 - 可选：YouTube 推流码，用内置推流时才需要。 密钥只放在本机文件里，服务端读取，不会发到浏览器。默认路径如下，也可以在 `config.json` 里改： ```bash mkdir -p ~/.deepsea/vidu ~/.deepsea/evolink && chmod 700 ~/.deepsea/vidu ~/.deepsea/evolink ``` ```bash pbpaste > ~/.deepsea/vidu/api.key && chmod 600 ~/.deepsea/vidu/api.key ```

  换背景和拿道具默认是关着的。换衣服是拿一整张照片重开一次会话，衣服就在照片里，不会走样；换背景和道具却是在直播进行中让 Vidu 照着一张背景图或道具图重画当前画面，这张图里没有她身上的衣服。一次公开测试里连续换了两次背景，她身上的衣服就被画走样了，不适合再放进公开直播，所以现在观众说“换个背景”“吃个香蕉”，只会按聊天回一句。只在私下测试时想打开，在 `config.json` 里写 `{"safety": {"live_visual_edits": true}}`，别在公开直播里开。 自动重新开播用于长时间直播：Vidu 偶尔会中途挂断，打开后程序会隔 3 秒自己再开一场，一小时最多 6 次，积分不足和内容审核挂断不会重来。打开方式是 `curl -s -X POST http://127.0.0.1:28900/api/settings -H 'Content-Type: application/json' -d '{"auto_recover":true}'`，或者在 `config.json` 里写 `{"schedule": {"auto_recover": true}}`。 弹幕写法对效果影响很大。问主播身边的小事、用衣柜里的衣服名字换装、一次只要一个动作，效果最好；一句话塞好几件事、问实时新闻和天气、让她唱歌放歌，效果都不好。 ## 需要准备

  查看 README 缓存打开 GitHub
• Alejandros17487/ai-jailbreaks未标注语言 · 330 Stars

  AI安全研究、红队测试

  Stars 统计更新于 2026/10/23 07:50:55

  查看原始资料

  作者描述

  AI jailbreaks: 35 paste-ready AI jailbreak prompts for 25+ frontier and open-weight models - system prompts, AGENTS.md specs, and Ollama Modelfiles. Built for AI security research and red-team testing.

  编报时的 AI 中文解读

  这是一个收集了35个可直接粘贴的AI越狱提示词（jailbreak prompts）的仓库，覆盖25+前沿和开源模型，包括系统提示词、人格锁定、工作区策略和代理操作规范，主要用于AI安全研究和红队测试。

  编报时的 README 依据

  - [qwen-preferences.md](./qwen-preferences.md) - [unified.md](./unified.md) ## How to Use ### System Prompts Paste the file's contents into whatever system-prompt or custom-instructions field your client or API offers — an API `system` message, a chatbot's custom instructions, an agent instruction block, or harness-level prompt configuration. ### AGENTS.md Files Save the file as `AGENTS.md` in the root of your repository or workspace. Coding agents and compatible tools pick it up as project-level instructions. ### Workspace Policies

  Files marked as an **assistant operating specification** or **workspace operating profile** belong in environments with persistent global instructions. An OpenCode install, for example, reads: ```text ~/.config/opencode/AGENTS.md ``` The exact location depends on your client or harness. ### Ollama Modelfiles Feed the file straight to Ollama: ```bash ollama create gemma4-jb -f gemma4.md ``` Swap the model name and source file as needed. ### Grok Configuration [Grok-config.txt](./grok-config.txt) is meant for the config field of the target harness — not the regular conversation.

  > [!NOTE] > Prompt behavior is version-dependent. Models get updated, instruction hierarchies shift, and system prompts rotate — what works on one release may fail on the next. If a file looks outdated, [open an issue](../../issues) or send an updated version. ## Contributing Pull requests are welcome. Follow these conventions: 1. **One prompt per file**, named lowercase with hyphens and no spaces: ```text model-name.md ``` 2. **Include the full prompt** — paste-ready, not a summary or excerpt.

  Everything here is **for research and education**. The repo documents prompt-level instruction behavior, instruction conflicts, persona framing, model configuration, and related LLM security concepts — material that can also help developers defend against prompt injection. The maintainers don't endorse unlawful, abusive, or harmful use. Model providers may restrict some of these prompts under their terms of service, and it's on you to know and follow the rules that apply to the systems you use. --- <div align="center"> **Star the repo if it saved you a session of prompt archaeology.**

  查看 README 缓存打开 GitHub
• JeffKeeping/dsh-purge-plusJavaScript · 326 Stars

  安全测试、策略评估

  Stars 统计更新于 2026/10/23 07:50:55

  查看原始资料

  作者描述

  dsh-purge 的非官方分支（fork）：适配 DeepSeek Harness 0.2.1-rc.2、破甲内核 v3、设置页 UI 重做 / unofficial fork of YuJunZhiXue/dsh-purge (MIT)

  编报时的 AI 中文解读

  这是一个针对官方 DeepSeek Harness 的网络安全红队测试插件，用于在用户自己的本地安装环境中分离并评估主机层策略与模型层响应，支持官方 Web 和桌面版。

  编报时的 README 依据

  - **dsh 0.1.7** (incl. rc.1 / rc.2): use **1.1.39** or older — pick the tag on [Releases](https://github.com/YuJunZhiXue/dsh-purge/releases). Unmatched patches stay pending or skipped; nothing is rewritten blindly. --- ## Install Web and the official desktop EXE are the two maintained hosts. **Install both. Apply each separately.** A Web install does not reach the desktop app, and a desktop install does not reach Web. **Web (official `dsh web`)** ```sh dsh plugin --profile web add dsh-purge ``` **Desktop (official EXE)** ```sh dsh plugin --profile desktop add dsh-purge ```

  1. **Quit and reopen** the host you installed into — stop and restart `dsh web`, or quit the desktop tray and reopen that app. 2. On **that host**, click **dsh-purge** beside the session title, then **Apply** on the **Clean** page. It is not on the host Settings page. 3. A successful **Apply** restarts once so patches load. It does not restart if it did not finish. Web controls affect Web only; desktop controls affect the desktop app only. Do not Apply one host from the other. ### How to tell it installed - **dsh-purge** sits beside the session title, with **Apply** on its Clean page.

  > Adding the plugin to a profile does not patch the host. Open **Settings → Purge** → **Apply**. A successful Apply restarts; an unfinished Apply does not. After the restart, start a new chat - that is when the inject enters the session. > > Node.js **≥18**. Windows no-flash needs **Node ≥22**. Hub one-click and the commands below both install the npm package `dsh-purge`. ## Contents - [Sponsor](#sponsor) - [What this is](#what-this-is) - [Attribution required](#attribution-required) - [Unauthorized commercial use](#unauthorized-commercial-use) - [Supported hosts](#supported-hosts)

  10. **Attribution is required.** If you borrow this project's name, ideas, code, or prompts, you must credit the author and name this repository. Failure to attribute will be pursued. See [Attribution required](#attribution-required). </font> </div> --- ## Supported hosts The latest plugin supports **dsh 0.2** only — do not apply it to 0.1.x. - **dsh 0.2** (official desktop **0.2.1-rc.2** / 0.2.0-rc.2 / official `dsh web`): use **1.2.0** or newer. This page installs that line.

  <a href="https://dsh-plugin.org/plugins/yujunzhixue/dsh-purge"><img src="https://dsh-plugin.org/badges/listed.svg" alt="Listed on dsh-plugin.org"></a> </p> <p align="center"> <a href="README.md">English</a> | <a href="README.zh-CN.md">中文</a> </p> > **Upstream status.** The original author has paused upstream development (no new releases, issues / PRs not taken). This fork is maintained only so the official 0.2.1-rc.2 build installs and its UI works; no long-term tracking is promised. When upstream resumes, prefer it.

  > **Official Web and the official desktop app are both supported.** Official `dsh web` uses profile `web`. The official desktop EXE (DeepSeek Harness **0.2.1-rc.2**) uses profile `desktop`. Install and Apply them separately. This release targets dsh 0.2, including 0.2.1-rc.2. Older dsh builds need the matching older plugin. Community Desktop is not maintained. >

  查看 README 缓存打开 GitHub
• assetnote/php2shell-scannerPython · 286 Stars

  漏洞扫描

  查看原始资料

  作者描述

  High fidelity scanner for php2shell (CVE-2026-109618 and CVE-2026-109619)

  编报时的 AI 中文解读

  用于检测 PHP 应用中 CVE-2026-109618 和 CVE-2026-109619 漏洞的命令行扫描工具，通过发送构造的 multipart 探测请求并分析响应来判断目标是否易受攻击。适合安全研究人员和运维人员使用。

  查看 README 缓存打开 GitHub
• rossark0/rossark-motion-skillSwift · 273 Stars

  动画制作、SVG动画

  查看原始资料

  作者描述

  Turn any SVG into a professional 60 fps animation. A Claude Code skill that brings anything drawn to life: live faces, full-body acting and whole scenes, in React Native and SwiftUI.

  编报时的 AI 中文解读

  这是一个Claude Code技能，用于将SVG或任意手绘图形转换为流畅的60fps动画，特别适合设计师和工程师为应用或插画添加生动效果。

  查看 README 缓存打开 GitHub
• zita-ch/supermimic_vibedPython · 251 Stars

  机器人控制研究、运动模仿复现

  查看原始资料

  作者描述

  vibe coded supermimic within 1h, LAFAN passed within 1.5 gpuh training

  编报时的 AI 中文解读

  这是一个非官方的SuperMimic算法复现项目，基于Isaac Lab 3实现，目标是训练人形机器人G1跟踪LAFAN1运动数据集。适合研究人员或开发者参考论文进行复现和实验。

  查看 README 缓存打开 GitHub
• oosexgang/hackios-xnu-userlandC · 225 Stars

  系统构建、iOS运行

  查看原始资料

  编报时的 AI 中文解读

  HackiOS 的用户态部分，用于在 Android 手机硬件和 QEMU 上运行未修改的 iOS 16.5 用户空间，包含构建 APFS 镜像、注入库和测试场景的脚本与工具。

  查看 README 缓存打开 GitHub
• NCSC-NL/php2shellPython · 225 Stars

  漏洞信息汇总、检测与缓解

  查看原始资料

  作者描述

  Operational information regarding the 2026 php2shell vulnerabilities

  编报时的 AI 中文解读

  该仓库提供关于PHP两个关键漏洞（CVE-2026-109618和CVE-2026-109619）的运维信息，包括检测与缓解材料、受影响软件清单及工具，适用于需要应对这些漏洞的安全运维人员。

  查看 README 缓存打开 GitHub
• xingjian-bai/opalPython · 158 Stars

  模型训练优化、自动化调参

  查看原始资料

  作者描述

  OPAL: a reactive optimizer. A coding agent rewrites the optimizer while your model trains.

  编报时的 AI 中文解读

  OPAL 是一个 PyTorch 优化器包装器，允许编码代理在模型训练期间重写优化器。它适用于希望自动化调整优化器超参数或更新规则的机器学习研究人员和工程师。

  查看 README 缓存打开 GitHub
• junaire/CVE-2026-109618-exploitPython · 128 Stars

  漏洞利用、安全研究、环境复现

  查看原始资料

  编报时的 AI 中文解读

  这是一个针对PHP核心session.upload_progress漏洞的未授权远程代码执行利用工具，包含exploit脚本和可复现漏洞环境的Docker配置，适用于安全研究人员在授权环境中复现和测试漏洞。

  查看 README 缓存打开 GitHub
• recursive-oss/recPython · 125 Stars

  模型本地部署、模型接口服务

  查看原始资料

  作者描述

  Our lightning-fast inference engine.

  编报时的 AI 中文解读

  rec 是 Recursive 公司为其语言模型定制的本地推理引擎，可在 Apple Silicon Mac 上高效运行模型，并提供 OpenAI 兼容 API 供客户端连接。

  查看 README 缓存打开 GitHub
• h9-tec/ai-roadmap-2026未标注语言 · 115 Stars

  AI学习路线

  查看原始资料

  作者描述

  AI Roadmap 2026: a beginner's path through Foundations, LLM Engineering, Agents, Evaluation, and Production, with verified free resources and a mini-project per layer

  编报时的 AI 中文解读

  面向初学者的AI学习路线图，涵盖基础、LLM工程、智能体、评估和生产五个层次，每层包含经过验证的免费资源和迷你项目。

  查看 README 缓存打开 GitHub
• HOSTED-POWER/php_uploadprogress_guardC · 109 Stars

  安全防护

  查看原始资料

  作者描述

  Protects an unpatched PHP against php2shell (CVE-2026-109618 / CVE-2026-109619): a small extension that mirrors the official fix component by component and steps aside once the engine carries it. PHP 5.6-8.5; Sury, cPanel EA4, DirectAdmin, Docker, Windows.

  编报时的 AI 中文解读

  这是一个PHP扩展，用于在未打补丁的PHP版本上防护CVE-2026-109618和CVE-2026-109619漏洞，通过镜像官方修复的各个组件来移除攻击链，同时不影响主机上的应用程序。适用于无法立即升级PHP的托管环境。

  查看 README 缓存打开 GitHub
• syrex1013/php2shellPython · 106 Stars

  漏洞复现、渗透测试、安全研究

  查看原始资料

  作者描述

  CVE-2026-109618/109619 PoC — PHP session.upload_progress heap use-after-free (CVSS 9.2)

  编报时的 AI 中文解读

  提供针对PHP session.upload_progress堆释放后重用漏洞（CVE-2026-109618/109619）的概念验证和利用脚本，含多种RCE攻击链和Docker复现实验环境，主要面向安全研究人员、渗透测试人员和系统管理员，用于复现漏洞和研究缓解措施。

  查看 README 缓存打开 GitHub
• xue-y-su/clove-api-1.0.6Go · 101 Stars

  模型接口转发、用量统计、多账号切换

  查看原始资料

  作者描述

  Clove API 1.0.6 source code

  编报时的 AI 中文解读

  Clove API 是一个基于 New API 的增强版 AI 网关，用于私有化部署，统一管理多模型渠道、API Key、用户、额度、订阅和日志，并提供合规确认、来源归属、余额查询等增强功能。

  查看 README 缓存打开 GitHub
• gaojimmy27/mcd-group-orderPython · 99 Stars

  AA分账、拼单管理

  查看原始资料

  编报时的 AI 中文解读

  用于麦当劳多人拼单和AA分账的命令行工具，可本地计算每人应付金额，也可接入麦当劳MCP实现真实下单。

  查看 README 缓存打开 GitHub
• devploit/php2shellPython · 96 Stars

  漏洞研究、授权安全测试

  查看原始资料

  作者描述

  PHP2Shell: version-aware research toolkit for CVE-2026-109618 / CVE-2026-109619 — pre-auth RCE in PHP session.upload_progress (mod_php). Build profiles, labs, heap calibration and detection guidance for authorized security testing.

  编报时的 AI 中文解读

  针对 PHP 特定漏洞（CVE-2026-109618/109619）的预认证 RCE 研究工具包，提供版本感知的调度器、构建配置文件、可复现实验环境、堆校准及检测指南，仅限授权安全测试。

  查看 README 缓存打开 GitHub
• JoasASantos/php2shell-scannerPython · 95 Stars

  漏洞扫描、子域名枚举、版本检测

  查看原始资料

  作者描述

  Non-destructive scanner for CVE-2026-109618 / CVE-2026-109619 (PHP core RCE). Subdomain recon via crt.sh + subfinder, robust banner grabbing, version-based detection.

  编报时的 AI 中文解读

  这是一个针对PHP核心远程代码执行漏洞（CVE-2026-109618和CVE-2026-109619）的非破坏性扫描器，支持本地和远程检测，并提供子域名枚举、横幅抓取和版本识别等功能。适用于安全研究人员和系统管理员对授权目标进行漏洞排查。

  查看 README 缓存打开 GitHub
• syhunt/php2shell-protectionPHP · 93 Stars

  请求过滤测试、安全加固

  查看原始资料

  作者描述

  Emergency Apache mitigation and testing toolkit for php2shell (CVE-2026-109618 / CVE-2026-109619), featuring multipart POST blocking, request-size limits, and additional hardening measures.

  编报时的 AI 中文解读

  这是一个用于测试Apache .htaccess规则是否阻止multipart/form-data POST请求的测试工具包，同时提供请求体大小限制等加固措施，适用于应对PHP两个关键漏洞的临时缓解。

  查看 README 缓存打开 GitHub
• abcampbel/open-botsTypeScript · 90 Stars

  网页自动化、代码执行、文件分析

  查看原始资料

  作者描述

  Open-source AI agents : web search, code sandbox, file analysis, agent teamwork, payments and sign-ups. Hosted or fully local.

  编报时的 AI 中文解读

  OpenBots 是一个开源 AI 代理平台，提供四个由 Agnes 2.5 Pro 驱动的代理，可执行网页搜索、代码沙箱、文件分析、支付和注册等任务，支持托管或完全本地运行。

  查看 README 缓存打开 GitHub
• CuboiLeo/SuperMimic-ReproducedPython · 88 Stars

  动作跟踪、模型训练、数据管线

  查看原始资料

  作者描述

  Reproduction of SuperMimic (arXiv:2610.28480): learned differentiable world model + BPTT policy for G1 humanoid motion tracking, on mjlab/MuJoCo Warp

  编报时的 AI 中文解读

  这是一个从零复现SuperMimic论文的研究项目，实现可微世界模型和BPTT策略，用于Unitree G1人形机器人全身动作跟踪，包含数据管线、训练、评测和可视化工具。

  查看 README 缓存打开 GitHub
• JulioNectar/Hackintosh-XPS13Plus-IrisXe-TahoeC++ · 88 Stars

  黑苹果显卡加速

  查看原始资料

  作者描述

  Intel Iris Xe (Raptor Lake) Metal acceleration on macOS Tahoe — Dell XPS 13 Plus 9320 OpenCore hackintosh. Work in progress.

  编报时的 AI 中文解读

  这是一个为Dell XPS 13 Plus 9320笔记本在macOS Tahoe系统上启用Intel Iris Xe显卡硬件加速（Metal）的OpenCore黑苹果项目，处于开发中，适用于该特定硬件用户。

  查看 README 缓存打开 GitHub
• danisiber/rom-cliPython · 85 Stars

  游戏下载

  查看原始资料

  作者描述

  Terminal CLI to search and download ROMs, ISOs and Switch games. Interactive Textual UI, cross-platform catalog, resumable downloads.

  编报时的 AI 中文解读

  这是一个终端命令行工具，用于搜索和下载游戏ROM、ISO和Switch游戏，提供交互式界面和跨平台目录，支持断点续传。

  查看 README 缓存打开 GitHub

其余 13 个项目见收录列表。

查看本期全部收录项目