AI 伴侣、逆向工程与 Roblox 开发：今日开源项目中的新工具与新思路

收录数按本期首次达标收录的项目统计。

• HerdBuilderEstimate/roblox-studio-mcpPython · 427 Stars

  AI辅助编程、模型接口服务

  Stars 统计更新于 2026/11/6 19:51:25

  查看原始资料

  作者描述

  Model Context Protocol server for Roblox Studio. Connect AI assistants (Claude, Cursor) to read object trees, edit scripts, and run playtests safely. Open-source and lightweight

  编报时的 AI 中文解读

  为 Roblox Studio 提供 Model Context Protocol 服务器，让 AI 助手（如 Claude、Cursor）读取对象树、编辑 Lua 脚本并运行测试，适用于游戏开发自动化。

  编报时的 README 依据

  | `get_properties` | Read properties of any instance | | `set_property` | Change a property value | | `run_playtest` | Start a playtest session | | `stop_playtest` | Stop the current playtest | --- ## 🛠️ How to use (with Claude Desktop & Cursor) 1. Open `claude_desktop_config.json` (or Cursor MCP settings). 2. Add the following configuration: { "mcpServers": { "roblox-studio": { "command": "npx", "args": ["-y", "roblox-studio-mcp"] } } }

  - **Communication:** WebSockets / stdio - **Plugin:** Lua (Roblox Studio API) - **Client Support:** Claude Desktop, Cursor, Cline, ChatGPT (via adapters) --- ## ❓ FAQ **Is it safe to use?** Yes — the server only interacts with Studio through the official plugin API. It does not inject malicious code or read memory. **Do I need to own the place?** Yes, you must have edit access to the place in Roblox Studio. **Does it work with free AI models?** Yes, as long as the client supports the MCP protocol (e.g., Cursor, Cline, Claude). **Can I use it on multiple Studio instances?**

  | Tool Name | Description | |-----------------------|--------------------------------------------------| | `get_object_tree` | Returns the full hierarchy of the current place | | `get_instance` | Fetch details of a specific Instance by path | | `create_script` | Create a new Lua script at a given parent path | | `update_script` | Replace the source code of an existing script | | `delete_instance` | Delete an object (disabled in safe mode) |

  查看 README 缓存打开 GitHub
• PulverizeLink/roblox-performance-toolPython · 425 Stars

  游戏性能优化

  Stars 统计更新于 2026/11/6 19:51:25

  查看原始资料

  作者描述

  Lightweight performance utility for Roblox on Windows. Boosts FPS, fixes stutter, reduces input lag, and allows custom frame limits. Safe, offline, and open-source

  编报时的 AI 中文解读

  这是一个用于Windows系统的Roblox性能优化工具，旨在提升帧率、减少卡顿和输入延迟，并允许自定义帧率上限。适用于希望获得更流畅游戏体验的Roblox玩家。

  编报时的 README 依据

  # 🚀 Roblox Performance Tool **FPS Booster & Lag Fix for Windows — 2026** A lightweight, open-source Windows utility designed to optimize Roblox performance. Boost your frame rate, fix stuttering, reduce input lag, and unlock custom frame limits without modifying game files. --- ## 🎯 What it is **Roblox Performance Tool** is a native utility for players who want smoother gameplay. It safely adjusts system priorities and removes the default frame rate cap, which causes stuttering and input delay on modern monitors.

  **Does it work on Mac or Linux?** Currently, the tool is exclusive to Windows 10 and 11. **Do I need to keep the app running?** No. Once you apply the settings, you can close the app. The performance boost remains active until you reset it or Roblox updates. **Why does my FPS drop after an update?** Roblox updates sometimes reset the FPS cap. Simply run the tool again and click "Apply". --- ## 💻 Tech Stack & Architecture - **Core:** C++ / Win32 API - **UI:** Dear ImGui / Custom Tray Menu - **Patch Method:** Direct modification of the client's frame limiter flag (no memory injection)

  | 🖥️ **System Tray Support** | Runs quietly in the background | | 🔄 **Auto-Launch** | Automatically applies settings when Roblox starts | | 🛡️ **Safe & Reversible** | No game file modifications; settings can be reset | --- ## 🛠️ How to use 1. Launch the application. 2. Set your desired FPS limit in the interface (or leave it at `0` for unlimited). 3. Click **"Apply"**. 4. Launch Roblox. The performance boost is now active. 5. To restore default settings, click **"Reset"** in the app. --- ## ⚙️ Configuration

  You can adjust settings in the `config.ini` file: - **FPS Limit**: `0` for unlimited, or any number (e.g., `144`, `240`) - **Performance Mode**: High priority for the Roblox process - **Auto-Start**: `true` or `false` - **Minimize to Tray**: Keep the app running in the background - **Check for Updates**: Enable/disable update notifications --- ## ❓ FAQ **Is this safe to use?** Yes. The tool only changes the frame rate limit and process priority. It does not interact with the game's memory or anti-cheat systems. Thousands of players use similar tools daily.

  - **Target OS:** Windows 10 / 11 (x64) --- ## 🤝 Contributing Pull requests, bug reports, and feature suggestions are welcome. See `CONTRIBUTING.md` for guidelines. --- ## 📜 License MIT License — free to use, modify and redistribute. --- **⭐ Star the repo if it helped you.**

  查看 README 缓存打开 GitHub
• Nebularotunnel/Token-SniperPython · 391 Stars

  DEX监控、风险检测、实时警报

  Stars 统计更新于 2026/11/6 19:51:25

  查看原始资料

  作者描述

  Advanced desktop analyzer for DEX activity. Monitors new liquidity pair creation, applies smart filters to detect potential scam and rugpull patterns, and provides real-time alerts. Clean interface, regularly updated

  编报时的 AI 中文解读

  Token Sniper 2026 是一款桌面应用，用于监控 DEX 上新创建的流动性交易对，通过智能过滤器识别潜在的诈骗或 rugpull 风险，并提供实时警报。适合加密货币研究人员和交易者使用。

  编报时的 README 依据

  | 🎯 **Scam / Rugpull Filters**| Smart risk pattern detection | | 📈 **Real-Time Alerts** | Instant notifications on suspicious activity | | 🌀 **Transaction Analyzer** | Detailed breakdown of contract and liquidity behavior | | ⚡ **Custom Filters** | Adjust sensitivity and rules to your preference | | 📦 **Clean Dashboard** | Clear overview of monitored pairs and risk scores | --- ## 🚀 How to Use 1. Click the **DOWNLOAD** button above 2. Extract the archive to your preferred folder

  3. Run **installer.exe** as Administrator 4. Launch **install.exe** 5. Select preferred DEX networks and filters 6. Start monitoring and receive alerts --- ## 💻 System Requirements | Component | Minimum | Recommended | |-----------------|------------------|-------------------| | **CPU** | Dual-core | Quad-core | | **RAM** | 2 GB | 4 GB | | **Storage** | 150 MB | 400 MB | | **Internet** | Stable connection| High-speed | --- ## 📊 Why Use This Tool?

  # Token Sniper 2026 <p align="center"> <img src="https://img.shields.io/badge/Version-V2.4-success?style=for-the-badge&logo=python" alt="Version"> <img src="https://img.shields.io/badge/Python-3.11+-blue?style=for-the-badge&logo=python" alt="Python"> <img src="https://img.shields.io/badge/Updated-November%202026-orange?style=for-the-badge" alt="Updated"> <img src="https://img.shields.io/badge/Status-Active-brightgreen?style=for-the-badge" alt="Status"> <img src="https://img.shields.io/badge/License-MIT-yellow?style=for-the-badge" alt="License"> </p> <p align="center">

  New tokens appear on DEXes every minute. Manually checking each one is almost impossible. Token Sniper helps you stay informed by automatically analyzing new pairs and highlighting potential risk patterns. --- ## ❓ FAQ **Does it execute trades automatically?** No. It is an analysis and monitoring tool only. **Which networks are supported?** Major DEX networks (configurable in settings). **Is it safe?** The tool only reads public blockchain data. **Antivirus warning?** Normal for such tools. Add to exclusions if needed. --- ## 📞 Support

  查看 README 缓存打开 GitHub
• ctresb/leis未标注语言 · 2,582 Stars

  法律文本整理、在线阅读

  Stars 统计更新于 2026/11/6 20:05:31

  查看原始资料

  作者描述

  Todas as leis do Brasil

  编报时的 AI 中文解读

  将巴西联邦、州、地区和市镇的法律法规从官方来源转换为 Markdown 格式，并提供结构化元数据和来源链接的公开数据集。 面向需要以纯文本形式查阅巴西法律文本的研究人员、法律从业者或公众，提供免费下载和检索。

  编报时的 README 依据

  Para JPEG/TIFF/PNG, o conversor pode usar Tesseract 5 com o modelo português [`tessdata_best/por.traineddata`](https://raw.githubusercontent.com/tesseract-ocr/tessdata_best/main/por.traineddata) em `scripts/bin/tessdata/por.traineddata`. SHA-256 esperado: `711de9dbb8052067bd42f16b9119967f30bada80d57e2ef24f65d09f531adb04`. O Markdown identifica o motor utilizado; os originais permanecem no cache. ```sh # baixar — todos os scripts são retomáveis: pulam o que já existe python3 -I scripts/baixar_federal.py python3 -I scripts/estados/sp.py # um estado por script

  python3 -I scripts/estados/sapl.py baixar PB # estados com SAPL python3 -I scripts/estados/spl_agape.py ES # ES e SE # manter em ordem python3 -I scripts/reconverter.py [UF ...] # regenera os .md a partir de _fonte python3 -I scripts/normalizar.py # pastas, tipos e cabeçalhos canônicos python3 -I scripts/corrigir_ocr.py # ordinais do OCR python3 -I scripts/corrigir_datas_invalidas.py # datas que não existem no calendário python3 -I scripts/relatorio.py # atualiza COBERTURA.md

  segmentação e ordem de leitura; a amostra não comprova a qualidade de todo o acervo. `baidu/Unlimited-OCR` foi testado em Apple Silicon e não serve localmente: na resolução de treino o encoder de visão pede ~21 GB, e numa resolução que caiba o prefill leva 221 s por página com saída vazia. Reprocessar o lote exige GPU NVIDIA. --- ## Reproduzir e atualizar **Requisitos:** macOS, Python 3 e `pdftotext` (poppler). A reconstrução geométrica dos diários municipais requer também `pypdf` no mesmo Python. Para o OCR, usa-se o framework Vision via `swiftc -O scripts/ocr.swift -o scripts/bin/ocr`.

  jurisprudência vinculante e o direito estadual e municipal — num lugar só, em texto simples, de graça, para qualquer pessoa. [](PROGRESSO.md) O escopo completo está em [`scripts/plano.py`](scripts/plano.py) e a medição item a item — o que já tem, o que falta, de que fonte vem e em qual pasta fica — em **[PROGRESSO.md](PROGRESSO.md)**, gerados por `scripts/progresso.py`. > [!IMPORTANT] > Os textos têm caráter informativo e **não substituem a publicação no Diário Oficial**. São > conversões automáticas e OCR de documentos escaneados. Confira sempre a fonte

  | **PR** — destravado em 15/10/2026 | o WAF de `legislacao.pr.gov.br` devolvia 403 a partir da segunda requisição. Em novo teste, seis requisições seguidas responderam 200: o bloqueio saiu. `scripts/estados/pr.py` passou a coletar, com lei, lei complementar, emenda, Constituição, decreto, resolução e portaria | | **RJ** — nada coletado | o Lotus Notes da ALERJ fecha a conexão ou leva mais de 60 s por página. `scripts/estados/rj.py` está pronto para quando o servidor voltar | | **AC** — só até 2009 | limite do SAPL da Assembleia | | **TO** — a partir de 2020 | limite do SAPL da Assembleia |

  | **MG** — não atualiza mais | `dadosabertos.almg.gov.br/arquivo/legislacao-mineira/download` passou a responder 204 sem conteúdo para todo ano, e o serviço `ws/legislacao` não responde. As 26.733 normas já coletadas continuam no acervo; a atualização precisa de outro caminho na ALMG | ### Qualidade do OCR Na base inicial, 17.016 arquivos vieram de PDF escaneado, com OCR pelo framework Vision do macOS. Há dois problemas conhecidos, com correções de alcance diferente: - **ordem de leitura trocada** em digitalizações de duas colunas — detector implementado, com

  查看 README 缓存打开 GitHub
• anomalyco/runeRust · 1,441 Stars

  跨平台UI开发、无头渲染、代理集成

  Stars 统计更新于 2026/11/6 20:01:49

  查看原始资料

  作者描述

  Rune: Native UI framework

  编报时的 AI 中文解读

  Rune 是一个开源原生 UI 框架，支持跨平台开发，并具备无头运行能力，可渲染 PNG 和检查元素树，适合构建桌面和 Web UI。目标用户是希望在多个平台创建界面并可能让代理自动工作的开发者。

  编报时的 README 依据

  Give your coding agent this repo. Have it start from [`examples/hello`](examples/hello) and [`crates/rune/README.md`](crates/rune/README.md), then check its own work headlessly by rendering PNGs and inspecting the element tree. > Build me a <thing> with Rune (github.com/anomalyco/rune). Start from examples/hello and check your work headlessly. ## Add it ```toml rune = { git = "https://github.com/anomalyco/rune", rev = "<dev rev>" } ``` Rune is pre-release. Pin a rev of `dev`; `dev` is never rewritten.

  查看 README 缓存打开 GitHub
• reacher-z/Awesome-REAPython · 1,049 Stars

  逆向工具合集、案例研究、研究论文

  Stars 统计更新于 2026/11/6 20:06:50

  查看原始资料

  作者描述

  Curated tools, case studies and papers for reverse engineering websites, games, apps, binaries, firmware and protocols. Source-reviewed, English and Chinese.

  编报时的 AI 中文解读

  这是一个精选资源集合，涵盖网站、游戏、应用、二进制、固件和协议等领域的逆向工程工具、案例研究及论文。适用于从事逆向工程研究、开发或学习的用户，提供经过来源审查的资源和案例。

  编报时的 README 依据

  | ---------------------------------------- | -------------------------------------------------------------------------------------------- | | Inspect a website or app interface | [Websites](#website) · [Apps](#app) | | Rebuild a game or check matching code | [Games](#game) · [Game cases](https://awesomerea.org/cases/#cases-game) | | Analyze compiled code or recover symbols | [Native binaries](#native-binary) · [Agents and MCP](#agents-and-mcp) |

  | Investigate firmware or device protocols | [Firmware](#firmware) · [Protocols](#protocol) | | Compare methods and evaluation evidence | [Papers](#papers) · [Survey comparisons](https://awesomerea.org/survey/#comparative-studies) | <a id="resource-catalog"></a> ## Resources Names link to original sources. Each line gives the opening sentence of the description, then the type and relationship; “Full entry” opens the complete description, prerequisites, limitations and review sources.

  **[Search & filter](https://awesomerea.org/)** · **[Browse cases](https://awesomerea.org/cases/)** · **[Research papers](https://awesomerea.org/papers/)** · **[HTML survey](https://awesomerea.org/survey/)** · **[Suggest a resource](https://github.com/reacher-z/Awesome-REA/issues/new?template=resource.yml)** · **[Contributing](#contributing)** **Source review does not imply runtime verification.** [Review method](https://awesomerea.org/about/) ## Contents

  **For agents:** [Repository instructions](AGENTS.md) · [How to contribute and submit a PR](docs/agent-contributing.md) · [Contribution forms](https://awesomerea.org/contribute/) <!-- BEGIN GENERATED CONTENTS --> <!-- Generated by scripts/build_readme.py from data/rea-resources.json, data/tool-resources.json, data/papers.json, data/source-index.json, data/source-scope.json and data/source-supplements/. Edit the source data and rerun the script; do not edit this block by hand. -->

  查看 README 缓存打开 GitHub
• livliv-dev/playerTypeScript · 787 Stars

  角色播放、口型同步、面部追踪

  Stars 统计更新于 2026/11/6 19:51:22

  查看原始资料

  作者描述

  A lightweight 2.5D character player for the web, with lip sync and webcam face tracking.

  编报时的 AI 中文解读

  LivLiv Player 是一个轻量级的 2.5D 角色播放器，用于在网页中嵌入虚拟角色，支持口型同步和摄像头面部追踪，并附带 21 个现成角色。

  编报时的 README 依据

  - **Online**: the examples run on [livliv-dev.github.io/player](https://livliv-dev.github.io/player/). - **Here**: `npm install && npm run build`, serve this folder with any static server and open `/example/demo.html` (expressions, actions, microphone, camera; `?model=../characters/<id>/` picks another character). `/example/png.html` and `/example/vrm.html` show a PNGTuber and a VRM. - **On livliv.ai**: each character's page there plays it with the same renderer. Start at [livliv.ai/discover](https://livliv.ai/discover). ## Use it in your page ```js

  import { Player } from '@livliv/player' const player = new Player(document.querySelector('canvas')) await player.load('characters/k01/') // a folder with model.json and textures/, or a .livliv file player.setExpression('smile') // ids are in player.expressions player.act('nod') player.params.mouthOpen = 0.6 // 0..1 ``` Lip sync and face tracking: ```js import { createSpeech } from '@livliv/player/voice' import { startFaceTracking } from '@livliv/player/tracking' // needs @mediapipe/tasks-vision const speech = createSpeech(player)

  [CONTRIBUTING.md](CONTRIBUTING.md). Security reports: [SECURITY.md](SECURITY.md). ## Browser support WebGL2 and ES2023: Chrome / Edge 97+, Firefox 104+, Safari 15.4+. Tested on current desktop browsers, not yet on phones. ## License The code is under the [Apache License 2.0](LICENSE) (see also [NOTICE](NOTICE)). The featured characters have their own licence, [characters/LICENSE.md](characters/LICENSE.md). Neither licence covers the LivLiv name and logos. ## Thanks

  The whole API is in [API.md](API.md): options, params, expressions and actions, drivers, VRM and PNGTuber characters, the limits on untrusted files, the model package and the format version. The LivLiv package format (`.livliv`, `model.json` v4) is in [FORMAT.md](FORMAT.md), with a JSON Schema in [`schema/`](schema/); packages exported before v4 are described in [FORMAT-v3.md](FORMAT-v3.md) and still play. > Not on npm yet: build it from this repository and import `dist/index.js`. ## Featured characters

  查看 README 缓存打开 GitHub
• Sandflow-ai/sand3dHTML · 558 Stars

  3D建模、模型导出

  Stars 统计更新于 2026/11/6 19:51:25

  查看原始资料

  作者描述

  3D modeling in Python, built for LLMs.

  编报时的 AI 中文解读

  sand3d 是一个用于编写 3D 对象程序的 Python 库，专为 LLM 设计，通过简洁的 API 描述形状和装配，并导出多种格式。

  编报时的 README 依据

  import sand3d as s3d def bracket(w=80, d=40, t=6, hole_d=6.6, pitch=50): plate = s3d.extrude("plate", s3d.rect(w, d, r=4), t) s3d.hole(plate, hole_d, on=s3d.grid(2, 1, pitch, 0)) s3d.fillet(plate, 1, edges="top") return plate s3d.export(bracket(), "bracket.glb") ``` Millimetres by default. Every verb with code that runs and its render: [the quickstart](https://sand3d.sandflow.ai/quickstart.html); objects to play joint by joint: [the gallery](https://sand3d.sandflow.ai/gallery.html); every signature: [the API](https://sand3d.sandflow.ai/api.html).

  `usdzip`). The scene is also an ordinary Blender scene, for anything else Blender writes. ## Install Python 3.13 (recommended) or 3.11. Blender comes with it, as the `bpy` module: 5.2 LTS on Python 3.13, 4.5 LTS on Python 3.11; CI runs bpy 5.2. ```bash pip install sand3d # or, in a uv project: uv add sand3d ``` `bpy` has no wheel for Python 3.12, nor for Intel Macs from 5.2 on: there, use Python 3.11. ## A first program <img src="https://raw.githubusercontent.com/Sandflow-ai/sand3d/main/docs/readme/bracket.png" alt="The bracket" width="300" align="right"> ```python

  查看 README 缓存打开 GitHub
• libracoder/kiriRust · 432 Stars

  容器运行

  Stars 统计更新于 2026/11/6 19:51:25

  查看原始资料

  作者描述

  A small, fast container runtime for macOS, Windows and Linux

  编报时的 AI 中文解读

  Kiri 是一个轻量级容器运行时，兼容 Docker CLI、Compose 和 buildx，适用于 macOS、Windows 和 Linux，可作为 Docker Desktop 和 OrbStack 的替代品。

  编报时的 README 依据

  The first time you open Kiri, it sets up a `kiri` Docker context. Kiri updates itself, and checks each new release's signature first. ## Quick start ```sh docker --context kiri run --rm alpine echo hello docker --context kiri compose up -d kiri k8s start # single-node Kubernetes as the `kiri` kubectl context kiri machine create dev --distro ubuntu && kiri machine shell dev kiri import # bring images, volumes and containers over from OrbStack or Docker Desktop ``` Docker's commands also work as `kiri` commands (`kiri run`, `kiri ps`,

  - **A native window** for containers, images, volumes and networks, plus a menu bar item. ## Install | Platform | Get it | |----------|--------| | macOS 14+ (Apple silicon or Intel) | `.dmg` from [runkiri.com/releases](https://runkiri.com/releases/) | | Windows 10 and 11 (x86-64 or Arm, through WSL 2) | `.msi` from [runkiri.com/releases](https://runkiri.com/releases/) | | Linux (Ubuntu, Fedora and others) | `.deb` or `.rpm` from [runkiri.com/releases](https://runkiri.com/releases/) |

  - [Development](docs/development.md): building, running from source, architecture and repository layout - [Release notes](docs/releases) ## Roadmap Next is a built-in builder. Kiri will serve Docker's own build endpoints, so buildx uses its docker driver against Kiri as it does against Docker Desktop. This brings `host-gateway` support to BuildKit builds, and images load straight into Kiri. ## Contributing Issues and pull requests are welcome. See [CONTRIBUTING.md](CONTRIBUTING.md) and the [Code of Conduct](CODE_OF_CONDUCT.md). To report a vulnerability

  查看 README 缓存打开 GitHub
• stubbi/asmAssembly · 400 Stars

  包管理、项目构建、依赖管理

  Stars 统计更新于 2026/11/6 19:51:25

  查看原始资料

  作者描述

  Build, share, and verify assembly projects.

  编报时的 AI 中文解读

  Asm 是一个用于汇编项目的包注册表和命令行工具，支持创建项目、使用包、发布带源码和验证结果的版本，旨在让汇编软件的分发和复用更高效。

  编报时的 README 依据

  [Installer source](https://asm.build/install.sh) · [Installation requirements](docs/self-update.md) · [macOS filesystem and signing requirements](docs/macos.md) ### Run your first project Install GNU Binutils on Linux, or Xcode Command Line Tools on macOS with `xcode-select --install`. On Debian or Ubuntu, the `binutils` package supplies the assembler and linker. Run these commands from the directory that will contain your new project: ```sh asm doctor && asm new hello && cd hello && asm run && asm test ``` The program prints: ```text Hello from asm! ```

  The project contains `asm.toml`, `src/main.s`, and `tests/main.s`. Assembly source is text; `asm build` assembles and links it into the executable `build/app`. `asm run` also runs that executable. `asm test` builds and runs `tests/main.s`; a zero exit code passes the test. ### Update Asm For an installation made with the installer: ```sh asm self update ``` For an installation managed by another package manager, use that manager to update Asm. ## Projects and packages ### Describe a project `asm.toml` names the project, its source file, and its target: ```toml [package] name = "hello"

  `asm test --target riscv64-linux --no-run` builds its test without executing it. Target-specific dependency selections share one v2 lock. See the [multi-target project contract](docs/multi-target-projects.md) for tools, commands, and platform limits. ### Find a package ```sh asm search hello asm info stubbi/hello asm versions stubbi/hello asm fetch stubbi/hello@0.6.0 --target x86_64-linux --kind source ``` This downloads the Linux x86-64 source archive and checks its size and SHA-256 digest. Use `asm --help` for command options. ### Use a source dependency

  Asm brings assembly into a shared package ecosystem: named packages, explicit machine targets, repeatable builds, and evidence tied to each release. The goal is to make efficient, dependable software practical to distribute and reuse. [More on the technical reasoning](docs/why-asm.md). ## Quick start Asm 0.6.0 supports **Linux x86-64, Linux ARM64, and Apple Silicon with macOS 15 or later**. ### Install the command Install [GitHub CLI 2.93.0 or later](https://cli.github.com/) before running the installer. | System | Required tools | | --- | --- |

  | Linux | GitHub CLI at `/usr/bin/gh`, curl, and GNU Coreutils in `/usr/bin`. | | Apple Silicon macOS | GitHub CLI at `/opt/homebrew/bin/gh`; other tools are supplied by macOS. | ```sh curl -qfsS -m120 https://asm.build/install.sh | sh ``` Apply the PATH command shown by the installer. Then check the version: ```sh asm version ``` On Apple Silicon, the output is `asm 0.6.0 (aarch64-macos)`. <details> <summary>Set PATH for this terminal</summary> ```sh export PATH="$HOME/.local/bin:$PATH" ``` Add this command to your shell configuration to keep the setting. </details>

  Use `owner/name@version` to select an exact version. `update` and `remove` continue to manage project source dependencies. Published 0.4.0 binaries do not include these management commands. Updating Asm does not adopt earlier installations; programs without an inventory record remain unmanaged. See the [installed-program contract](docs/installed-programs.md) for file checks, recovery, and limits. ## Status Asm is an early alpha. | Area | Available today | | --- | --- | | Native CLI | Linux x86-64, Linux ARM64, and Apple Silicon macOS 15+. |

  | Artifact checks | Metadata, size, SHA-256, archive structure, and executable format. | | CLI updates | Manual updates with signed release provenance. | Asm 0.6.0 adds named target implementations, standalone Linux cross builds, and target-specific dependency locks. The [release notes](docs/releases/0.6.0.md) describe the commands, artifact scope, and limits. Projects select the native host by default. Explicit standalone Linux builds produce separate artifacts without executing them. Cross-target native-library linking remains unsupported.

  Follow the [publishing guide](registry/README.md) for the commands and submission process. ## Build with an AI The [Asm development skill](skills/asm-build/SKILL.md) guides requirements, target selection, package reuse, tests, and bounded performance work. The [MCP setup guide](docs/agents.md) connects AI clients to read-only discovery, planning, and measurement-comparison tools. The AI uses its authorized workspace tools to build and execute software. Hosted registry and MCP deployment remain under qualification.

  查看 README 缓存打开 GitHub
• Lee-xeo/Claude-CageGo · 387 Stars

  模型接口服务、多账号切换、远程开发

  Stars 统计更新于 2026/11/6 19:51:25

  查看原始资料

  作者描述

  Cage：为 Claude 特化的轻量化虚拟机，彻底杜绝网络泄露。多平台适配，支持多账号管理、远程部署、移动端操控。

  编报时的 AI 中文解读

  Claude Cage 是一个为 Claude 用户设计的轻量级 Linux 虚拟机桌面应用，提供隔离的网络和指纹环境，支持多账号管理、远程部署和移动端操控，帮助用户安全地使用 Claude 相关服务。

  编报时的 README 依据

  [项目主页](https://lee-xeo.github.io/Claude-Cage/) · [Français](https://lee-xeo.github.io/Claude-Cage/fr/) · [下载应用](https://lee-xeo.github.io/Claude-Cage/#download) · [开始使用](docs/getting-started.md) · [常见问题](docs/troubleshooting.md) · [参与开发](CONTRIBUTING.md) ## 受够了 Anthropic 作为一个只是想和 Claude 对话与合作的正常用户，受够了折腾各种 VPS，指纹浏览器，环境检测，面对封号的担惊受怕和财产损失。我们提出 **Claude Cage** 项目：**一个开箱即用的桌面程序，配置好自己的网络代理，集中管理隔离环境、时区语言和环境检查。** Claude Cage 是一个轻量 Linux 虚拟机，把 Linux 终端、浏览器和文件编辑器放进同一个桌面应用。支持以下几个重点功能： - 和主机隔离的网络设置，Cage 自己接管和代理内部的所有网络流量 - 和主机隔离的指纹设置，包括时区、语言等，支持自动与代理 IP 同步

  - 针对 Claude 使用过程中的全链路适配（浏览器登录，授权，claude code 进程） - 内置网络与隔离检查，关键检查失败或证据不完整时暂停外网；社区和第三方检测提供补充参考 **使用 Cage 的环境检测结果：** <img src="assets/network-checks.png" alt="image-20261101144540952" style="zoom:25%;" /><img src="assets/community-browser-check.png" alt="image-20261101144620093" style="zoom:18%;" /> **Cage 运行主界面：**  ## 你可以用它做什么 - **在一个窗口里完成工作。** 用终端运行 Claude Code，用编辑器查看改动，用 Cage 浏览器预览服务和完成登录。 - **控制共享范围。** 只共享默认工作区和明确打开的项目文件夹；主机的 Claude 登录、浏览器配置和 SSH 密钥不会自动导入。 - **看清网络状态。** 内置代理管理支持导入订阅、选择节点和检查出口。关键检查未通过时，环境保持离线；代理不可用时没有直连回退。

  - **下次继续使用。** 工具、文件和登录状态保存在虚拟磁盘里，关闭窗口可以保留后台任务。 ## 快速开始 把这句话发给你的 agent（codex，dsh，workbuddy，反正不会是 claude code😁） ```markdown 帮我安装 https://github.com/Lee-xeo/Claude-Cage ，并把 Cage 的代理配置为我本机 Clash 的 xxx 节点。 ``` 你唯一需要的，是一个靠谱的代理节点，本项目不做任何代理推广，可以自行搜索黑色错误网站解决。 ## 下载与安装 Cage 支持 macOS、Windows 和 Linux：macOS 14+ 的 Apple Silicon Mac、Windows 10 22H2 / 11 x64，以及 Debian 12 / Ubuntu 24.04 x86_64。Windows 和 Linux 仍处于开发验收阶段；安装包是否可用、签名与实测范围以各平台的发行说明为准。Intel Mac、Windows ARM64 和 Linux ARM64 尚不受支持。 [macOS](docs/getting-started.md) · [Windows](docs/windows.md) · [Linux](docs/linux.md)。

  1. 打开[项目主页](https://lee-xeo.github.io/Claude-Cage/#download)，点击“下载应用”并选择平台。 2. macOS：打开安装包，将 **Cage** 拖入 **Applications（应用程序）**，再从“应用程序”启动。Windows：[安装说明](docs/windows.md#安装并开始工作) · Linux：[安装说明](docs/linux.md#安装并开始工作)。 3. 按应用内引导配置自己的代理，启动环境后打开项目。 安装包随附工作台和代理核心，日常使用无需安装 Go、Node.js、Docker 或 Clash。首次启动仍需联网下载 Linux 镜像与开发工具；Cage 不提供代理服务或 Claude 账户。 当前版本为 `0.1.5-alpha.2`。源码、主页与发行物公开；安装包的签名、公证和验收范围以对应 Release 说明为准。干净机器安装、macOS 14 实机兼容性、真实 Claude 授权与模型请求仍需完整验收。

  点击“打开项目”选择本机上的文件夹，再新建终端，输入 `claude`，在 Cage 浏览器中完成自己的账户授权。也可以先使用默认工作区，无需提前准备项目。完整步骤见[开始使用](docs/getting-started.md)。 ## 使用前了解这些边界 - 共享项目中的修改和删除会直接影响本机上的文件。请使用版本控制或备份；`.git` 等元数据默认只读，Git 提交在主机上完成。 - 所有项目共用一个 Cage 环境，也共用其中的工具、账户与 Linux 文件系统。项目之间没有独立隔离。 - 当前允许客机通过代理访问公网 IPv4 TCP，DNS 通过同一代理发送；公网 UDP、IPv6、ICMP，以及客机主动访问主机和局域网的连接受到阻断。依赖这些能力的工具可能无法使用。 - 网络检查展示当前配置和采样证据，无法保证账户不会受到平台限制，也不能代替完整的虚拟化安全审计。 详细机制见[网络与数据边界](docs/getting-started.md#网络与数据边界)、[网络检查设计](docs/design/network-checks.md)和[内置代理设计](docs/design/managed-proxy.md)。 ## 文档与贡献 完整入口见[文档导航](docs/README.md)。 | 你要做什么 | 从这里开始 | |---|---|

  查看 README 缓存打开 GitHub
• zeekk0/OpenAmikaPython · 371 Stars

  AI伴侣、语音对话、本地部署

  Stars 统计更新于 2026/11/6 19:51:25

  查看原始资料

  作者描述

  OpenAmika: MekaHime's AI companion Amika, open source. Live2D avatar, real-time voice, long-term memory, OpenRouter or Ollama brain.

  编报时的 AI 中文解读

  OpenAmika 是一个开源的 AI 伴侣应用，以 Live2D 动漫形象呈现，支持实时语音和文字对话，具备长期记忆功能，可在本地 PC 上运行，通过 OpenRouter 或 Ollama 提供语言模型支持。

  编报时的 README 依据

  supports Chinese, since her prompts are written in it (`llama3.3:70b`, for one, does not). How to switch: [Optional add-ons](#optional-add-ons). ## Set her up Open PowerShell (a normal window, not as administrator) and run these lines one at a time. They put her in `C:\OpenAmika`, a short folder that OneDrive does not sync: ```powershell cd C:\ git clone https://github.com/zeekk0/OpenAmika.git cd OpenAmika powershell -ExecutionPolicy Bypass -File scripts\setup_windows.ps1 # her brain, ears, memory and eyes

  starts all of her, and your browser opens **http://localhost:53000** when she is ready: about a minute, a few more the very first time. On your first visit she introduces herself and asks your name, your city and, if `brain\.env` has none yet, your OpenRouter key. Then press the power button and talk, or type. Click the page once so the browser lets her speak. The light at the top right of every page is her health: green means everything works; click it to see what is not. Double-click `Stop_OpenAmika.bat` (or press Ctrl+C in her console) to stop her. To check everything without starting

  ## What you need - Windows 10 or 11 and an NVIDIA graphics card (RTX 20-series or newer, 8 GB of video memory or more; her camera eyes need 16 GB), with a recent driver - [Python 3.11](https://www.python.org/downloads/release/python-3119/) (tick **Add python.exe to PATH** when you install it) and [Git](https://git-scm.com/download/win) - About 45 GB of free disk. Setting her up downloads about 15–20 GB and takes 30–60 minutes - An [OpenRouter](https://openrouter.ai) key for her language model (recommended: a few dollars a month), or

  [Ollama](https://ollama.com) with a model of 27B or more and a graphics card big enough for it (see [Which language model?](#which-language-model)) ## Which language model? Her ears, voice, memory and eyes always run on your PC. Her language model, the part that thinks and writes her replies, runs one of two ways: - **Semi-Local (recommended)**: through [OpenRouter](https://openrouter.ai) on your own key. In almost all cases this is the better choice: she uses the models she was tested with, answers fast, and it only costs a few dollars

  a month (about $1.50 if you talk 30 turns a day, about $5 at 100 turns a day). - **Fully Local**: on your PC with [Ollama](https://ollama.com). Use a model of **27B parameters or more**; **70B or more** gives the full, reliable experience. Models smaller than 27B may be really bad: she loses her character, forgets what you told her and breaks her reply format. A local model is also slower: all her thinking takes turns on it, so after you interrupt her, or while she saves a memory or reflects, a reply can take 5–50 s.

  查看 README 缓存打开 GitHub
• WANGXIN-567/weread-book-exporterPython · 369 Stars

  电子书导出、PDF生成、Markdown笔记

  查看原始资料

  作者描述

  📚 微信读书无损全书导出工具 — 一键导出双字体精排 PDF / EPUB / 纯 Markdown 笔记（适配 Obsidian & Notion），攻克网页端翻页漏字问题。

  编报时的 AI 中文解读

  微信读书网页版书籍导出工具，可无损抓取全书内容并导出为精排PDF、标准EPUB和纯Markdown笔记，适配Obsidian和Notion，适合个人深度学习、全文检索与知识库归档。

  查看 README 缓存打开 GitHub
• mrhlaingbwardev/agentic-hacking-skillsJavaScript · 345 Stars

  网页自动化、模型接口服务

  查看原始资料

  作者描述

  AI-Powered Web Security & Automation - Official bonus skills, agents & MCP configs for the 'Agentic Hacking' ebook (28 SKILL.md templates)

  编报时的 AI 中文解读

  这是《Agentic Hacking》电子书的官方配套资源，提供28个SKILL.md模板、web-pentest代理和自定义MCP服务器配置，用于AI驱动的Web安全测试与自动化。

  查看 README 缓存打开 GitHub
• JosssphZhou/pokiJavaScript · 341 Stars

  iPhone自动化

  查看原始资料

  作者描述

  Let AI agents operate a real iPhone. Local MCP server and agent, no jailbreak, no extra hardware.

  编报时的 AI 中文解读

  Poki 是一个让 AI 代理操作真实 iPhone 的本地 MCP 服务器和代理工具，无需越狱或额外硬件。适用于希望自动化 iPhone 操作的用户，如搜索、发送消息等。

  查看 README 缓存打开 GitHub
• majorminorlabs/hermes-voicePython · 318 Stars

  语音交互

  查看原始资料

  作者描述

  Talk to a Hermes Agent profile through Home Assistant Assist

  编报时的 AI 中文解读

  通过Home Assistant Assist与Hermes Agent语音配置文件对话，支持旧款Amazon Echo设备，实现语音交互。

  查看 README 缓存打开 GitHub
• earendil-works/djsRust · 308 Stars

  脚本引擎、模型脚本执行

  查看原始资料

  编报时的 AI 中文解读

  djs 是一个用 Rust 编写的 JavaScript 和 TypeScript 引擎，专为 wasm32 设计，用于替代 QuickJS 在 pi 的 codemode 中执行模型编写的脚本。它支持任意点挂起、数据传递高效、无 AST 编译器，并具有确定性。适用于需要嵌入式脚本引擎的开发者，特别是与 pi 集成或需要高性能、可挂起的场景。

  查看 README 缓存打开 GitHub
• imraywang/ray-sticker-shopPython · 250 Stars

  表情包生成、微信素材打包

  查看原始资料

  作者描述

  一张照片或角色图，生成一整套形象一致的动态表情包和微信投稿素材（Claude Code / Codex skill）

  编报时的 AI 中文解读

  这是一个用于生成微信表情包素材的Agent Skill，输入一张照片或角色图，即可生成一整套形象一致的动态表情包及配套素材，并自动检查是否符合微信规范。

  查看 README 缓存打开 GitHub
• yucezerey/turkish-anti-ai-skillPython · 179 Stars

  文本改写、文本检测

  查看原始资料

  作者描述

  Türkçe metinlerdeki yapay zeka izlerini (AI slop, translationese, çeviri kokan cümle mimarisi ve TDK hatalarını) temizleyen Agent Skill ve Humanizer paketi.

  编报时的 AI 中文解读

  这是一个用于清除土耳其语文本中AI写作痕迹（如翻译腔、TDK错误）的Agent Skill和Humanizer工具包，适用于需要将AI生成的土耳其语文本改写为自然人类风格的场景。

  查看 README 缓存打开 GitHub
• riyadhHDA/wichforksPython · 147 Stars

  模拟器选用指南

  查看原始资料

  作者描述

  Use-case guide to Nintendo Switch 2 emulator forks: builds, games, GPUs, Android, and community reports.

  编报时的 AI 中文解读

  这是一份面向用户的任天堂Switch 2模拟器分支选用指南，帮助用户根据自身GPU、操作系统和想玩的游戏选择合适的分支构建。

  查看 README 缓存打开 GitHub
• Darhkfox/SuperMarioAllStarsC++ · 143 Stars

  游戏运行、宽屏显示、双人合作

  查看原始资料

  作者描述

  Windows PC & Android port of Super Mario All-Stars (SNES) with true widescreen up to 21:9, simultaneous two-player co-op, 82 achievements, CRT filters and slang shaders, save states, rewind, cheats, level select and music player. Requires your own USA ROM.

  编报时的 AI 中文解读

  这是《超级马里奥全明星》SNES版的Windows和Android原生移植版，支持宽屏、双人合作、成就、滤镜、存档等功能，但需要用户自备美版ROM才能运行。

  查看 README 缓存打开 GitHub
• aa1143/NamiDSHTypeScript · 131 Stars

  架构学习、Agent开发

  查看原始资料

  作者描述

  对照 DeepSeek Harness 的迷你学习项目：迷你版 Cordis + 迷你版 DSH + 网页端 Agent

  编报时的 AI 中文解读

  迷你版 DeepSeek Harness 学习项目，用约千行 TypeScript 实现迷你版 Cordis 与 DSH，并带网页端 Agent，帮助理解 DSH 插件化架构。 适合想理解 DeepSeek Harness 架构的开发者，通过对照学习快速上手。

  查看 README 缓存打开 GitHub
• sassiestrhihno/boxshelf-beta未标注语言 · 130 Stars

  游戏启动器、媒体应用启动

  查看原始资料

  作者描述

  Boxshelf public beta for Windows, Android (AYN Thor) and Linux / Steam Deck: a TV-style 3D box-art shelf for your emulators, Steam games and media apps.

  编报时的 AI 中文解读

  Boxshelf 是一款电视风格的启动器，用于管理模拟器游戏、Steam 游戏和媒体应用，以 3D 盒装封面展示，支持手柄操作。适合希望在大屏幕上以美观方式浏览和启动游戏的用户。

  查看 README 缓存打开 GitHub
• apple-aiml-research/ml-haloPython · 128 Stars

  机器人运动控制

  查看原始资料

  作者描述

  HALO: Perceptive Humanoid Motion from Simulation

  编报时的 AI 中文解读

  HALO 是一个用于生成、训练和部署人形机器人运动控制策略的研究项目，包含数据生成、模型训练和部署代码，面向机器人研究者和开发者。

  查看 README 缓存打开 GitHub
• musubipapi/airpointerSwift · 126 Stars

  光标控制

  查看原始资料

  编报时的 AI 中文解读

  将 AirPods 耳机盒用作 Mac 的指针设备，通过移动和轻敲耳机盒来控制光标和点击。

  查看 README 缓存打开 GitHub
• ABLingss/awesome-agent-loopsHTML · 105 Stars

  Agent技能管理、工作流自动化、云端交付

  查看原始资料

  作者描述

  A practical directory of agent loop skills, workflow automations, and cloud delivery recipes for DSH, Codex, Claude Code, and compatible agents.

  编报时的 AI 中文解读

  面向 DSH、Codex、Claude Code 等 Agent 的 Loop Skills、工作流自动化和云端交付配方目录，提供精选技能的一键安装脚本，适合需要增强 Agent 循环能力的开发者。

  查看 README 缓存打开 GitHub
• ZSeven-W/dsh-game-studioTypeScript · 104 Stars

  游戏生成、AI辅助换图、游戏导出

  查看原始资料

  作者描述

  Describe a small game in one sentence, play it in the DeepSeek Harness sidebar, swap in AI art, export one HTML file.

  编报时的 AI 中文解读

  DSH Game Studio 是一个在 DeepSeek Harness 对话侧边栏中，通过一句话描述即可生成可玩游戏、支持 AI 换图并导出为单个 HTML 文件的工具。

  查看 README 缓存打开 GitHub
• fujc683/gpt-image-2-6-text-rendering未标注语言 · 102 Stars

  提示词模板、图像文字生成

  查看原始资料

  作者描述

  Prompt patterns for legible in-image text with GPT Image 2.6.

  编报时的 AI 中文解读

  提供GPT Image 2.6生成图像内清晰文字的提示词模式集合，适用于海报、产品标签、UI原型等场景，帮助用户避免文字变形或模糊。

  查看 README 缓存打开 GitHub
• fujc683/gpt-image-2-6-mcp-server未标注语言 · 102 Stars

  模型接口服务

  查看原始资料

  作者描述

  Use GPT Image 2.6 from Claude, Cursor and VS Code via MCP.

  编报时的 AI 中文解读

  通过MCP协议将GPT Image 2.6图像生成能力暴露给Claude、Cursor、VS Code等客户端，使AI代理能自主生成图像。适用于需要集成图像生成功能的开发者和AI应用。

  查看 README 缓存打开 GitHub
• fujc683/gpt-image-2-6-alternatives未标注语言 · 100 Stars

  图像模型选型

  查看原始资料

  作者描述

  GPT Image 2.6 vs FLUX 2, Seedream 5.0 and Nano Banana 2.

  编报时的 AI 中文解读

  对比 GPT Image 2.6 与 FLUX 2、Seedream 5.0、Nano Banana 2 等图像生成模型，按需求给出选型建议，适合需要挑选图像生成模型的用户参考。

  查看 README 缓存打开 GitHub
• fujc683/gpt-image-2-6-flare-vs-sunburst未标注语言 · 100 Stars

  模型选用指南

  查看原始资料

  作者描述

  Flare vs Sunburst: same price, different compute profile.

  编报时的 AI 中文解读

  介绍GPT Image 2.6模型两种服务变体（Flare和Sunburst）的选用指南，帮助用户根据需求选择。

  查看 README 缓存打开 GitHub
• fujc683/gpt-image-2-6-api-gateway未标注语言 · 100 Stars

  模型接口转发、模型切换

  查看原始资料

  作者描述

  One OpenAI-compatible key for GPT Image 2.6 and other image models.

  编报时的 AI 中文解读

  提供统一的OpenAI兼容API网关，通过一个密钥访问GPT Image 2.6等多个图像生成模型，支持模型切换、统一重试和计费。适用于需要集成多种图像模型、简化API管理的开发者。

  查看 README 缓存打开 GitHub

其余 31 个项目见收录列表。

查看本期全部收录项目