dora-skills

English | 中文

---

English

Introduction

Skills for building AI agents, workflows, and embodied intelligence applications with the dora-rs dataflow framework.

Dora is a high-performance dataflow framework that orchestrates AI models, sensors, and actuators through declarative YAML pipelines. It's ideal for scenarios requiring real-time data flow coordination — from multi-modal AI agents to robotic systems.

Current Focus: Embodied Intelligence — combining perception, language models, and physical control.

Install Skills

Method 1: Marketplace (Recommended)

Install from Claude Code Plugin Marketplace in two steps:

# Step 1: Add the marketplace
/plugin marketplace add dora-rs/dora-skills

# Step 2: Install the plugin
/plugin install dora-skills@dora-skills

Note: Step 1 only adds the marketplace (plugin source). Step 2 actually installs the dora-skills plugin with all features enabled.

Method 2: NPX

Install using npx:

npx skills add dora-rs/dora-skills

Note: NPX installs skills only. For full plugin features (hooks, agents), use Method 1 or Method 3.

Method 3: Full Plugin

This method enables all features including hooks for automatic skill triggering.

# Clone the repository
git clone https://github.com/dora-rs/dora-skills.git

# Launch with plugin directory
claude --plugin-dir /path/to/dora-skills

Method 4: Skills Only

This method only installs skills without hooks. You need to manually invoke skills.

# Clone and copy skills
git clone https://github.com/dora-rs/dora-skills.git
cp -r dora-skills/skills/* ~/.claude/skills/

Note: Without hooks, skills won't trigger automatically. You must manually call specific skills.

Feature Comparison

Feature	Marketplace	NPX	Full Plugin	Skills Only
All 20 Skills	✅	✅	✅	✅
Auto skill trigger	✅	✅	✅	❌
Hook-based routing	✅	✅	✅	❌
Background agents	✅	✅	✅	✅
Easy updates	✅	✅	❌	❌

Manual Usage

You can also browse the skills/ directory and use the SKILL.md files as reference documentation.

Project Structure

dora-skills/
├── skills/                     # Main skills directory
│   ├── dora-router/           # Main question router (load first)
│   ├── dataflow-config/       # YAML configuration
│   ├── node-api-rust/         # Rust DoraNode API
│   ├── node-api-python/       # Python dora.Node API
│   ├── operator-api/          # Operator development
│   ├── cli-commands/          # CLI usage
│   ├── integration-testing/   # Node testing
│   ├── domain-vision/         # Vision/ML pipelines
│   ├── domain-audio/          # Audio processing
│   ├── domain-robot/          # Robot control
│   ├── data-pipeline/         # Data recording/replay
│   └── hub-nodes/             # Pre-built node packages
├── commands/                   # Slash commands
├── agents/                     # Background agents
├── hooks/                      # Auto-trigger hooks
├── docs/                       # Development documentation
└── index/                      # Skill catalog

Auto-Triggering Mechanism

Skills are automatically triggered through two mechanisms:

1. Hooks (hooks/hooks.json): UserPromptSubmit hook matches keywords in user messages
2. Skill Frontmatter: Each SKILL.md contains a description field with trigger keywords

Example skill frontmatter:

---
name: hub-audio
description: "Use for audio processing nodes in dora.
Triggers on: dora-microphone, dora-vad, whisper, kokoro-tts, VAD, speech-to-text..."
---

When you ask about "dora-microphone" or "speech-to-text", the appropriate skill loads automatically.

Available Skills

Category	Skills	Description
Core	dataflow-config, cli-commands	Dataflow YAML and CLI usage
API	node-api-rust, node-api-python, operator-api	Node and operator development
Testing	integration-testing	Node testing with JSONL inputs
Vision	domain-vision	YOLO, SAM2, CoTracker, VLM
Audio	domain-audio	Whisper, Kokoro TTS, VAD
Robot	domain-robot	Arm control, chassis, kinematics
Data	data-pipeline	LeRobot recording and replay
Hub	hub-nodes	Pre-built node packages

Hub Node Skills (Detailed)

Skill	Description	Key Nodes
hub-camera	Camera capture	opencv-video-capture, dora-pyrealsense, dora-pyorbbecksdk
hub-audio	Audio processing	dora-microphone, dora-vad, dora-distil-whisper, dora-kokoro-tts
hub-detection	Detection/tracking	dora-yolo, dora-sam2, dora-cotracker
hub-llm	Language models	dora-qwen, dora-qwen2-5-vl, dora-internvl
hub-robot	Robot control	dora-piper, dora-reachy2, dora-ugv, dora-rdt-1b
hub-visualization	Visualization	dora-rerun (12 primitives), opencv-plot
hub-recording	Data recording	llama-factory-recorder, lerobot-dashboard
hub-translation	Translation	dora-opus, dora-argotranslate

Commands

• /new-dataflow - Create a new dataflow project
• /add-node - Add a node to existing dataflow
• /visualize - Generate dataflow visualization

Agents

• dataflow-builder - Generate dataflow YAML configurations
• node-debugger - Troubleshoot dataflow issues

Using with Dora

After loading skills, you can ask Claude to help you:

# Create a vision detection pipeline
"Help me build a dataflow with camera input and YOLO detection"

# Set up speech-to-speech
"Create an audio pipeline with microphone, Whisper, and TTS"

# Robot teleoperation
"I want to record demonstrations with a robot arm"

Dora Quick Start

1. Install dora CLI:

pip install dora-rs-cli
# or
cargo install dora-cli

2. Create and run a dataflow:

dora new my-project --lang python
dora build dataflow.yml --uv
dora run dataflow.yml

Resources

• Dora Documentation
• Dora GitHub
• Node Hub

---

中文

简介

使用 dora-rs 数据流框架构建 AI Agent、工作流和具身智能应用的技能集。

Dora 是一个高性能数据流框架，通过声明式 YAML 管道编排 AI 模型、传感器和执行器。适用于需要实时数据流协调的场景 —— 从多模态 AI Agent 到机器人系统。

当前重点: 具身智能 (Embodied Intelligence) —— 融合感知、语言模型与物理控制。

安装 Skills

方法 1: Marketplace (推荐)

通过 Claude Code 插件市场安装，分两步：

# 步骤 1: 添加 marketplace
/plugin marketplace add dora-rs/dora-skills

# 步骤 2: 安装插件
/plugin install dora-skills@dora-skills

说明: 步骤 1 仅添加 marketplace（插件源）。步骤 2 才是真正安装 dora-skills 插件。

方法 2: NPX

使用 npx 安装：

npx skills add dora-rs/dora-skills

说明: NPX 仅安装 skills。如需完整插件功能（hooks、agents），请使用方法 1 或方法 3。

方法 3: 完整插件

此方法启用所有功能，包括 hooks 自动触发。

# 克隆仓库
git clone https://github.com/dora-rs/dora-skills.git

# 使用插件目录启动
claude --plugin-dir /path/to/dora-skills

方法 4: 仅 Skills

此方法仅安装 skills，不含 hooks。需要手动调用 skills。

# 克隆并复制 skills
git clone https://github.com/dora-rs/dora-skills.git
cp -r dora-skills/skills/* ~/.claude/skills/

说明: 没有 hooks，skills 不会自动触发。必须手动调用特定 skill。

功能对比

功能	Marketplace	NPX	完整插件	仅 Skills
全部 20 个 Skills	✅	✅	✅	✅
自动触发 skill	✅	✅	✅	❌
Hook 路由	✅	✅	✅	❌
后台 agents	✅	✅	✅	✅
便捷更新	✅	✅	❌	❌

项目结构

dora-skills/
├── skills/                     # 主技能目录
│   ├── dora-router/           # 主路由器（首先加载）
│   ├── dataflow-config/       # YAML 配置
│   ├── node-api-rust/         # Rust DoraNode API
│   ├── node-api-python/       # Python dora.Node API
│   ├── operator-api/          # 操作符开发
│   ├── cli-commands/          # CLI 使用
│   ├── integration-testing/   # 节点测试
│   ├── domain-vision/         # 视觉/ML 管道
│   ├── domain-audio/          # 音频处理
│   ├── domain-robot/          # 机器人控制
│   ├── data-pipeline/         # 数据记录/回放
│   └── hub-nodes/             # 预构建节点包
├── commands/                   # 斜杠命令
├── agents/                     # 后台代理
├── hooks/                      # 自动触发钩子
├── docs/                       # 开发文档
└── index/                      # 技能目录

自动触发机制

Skills 通过两种机制自动触发：

1. Hooks (hooks/hooks.json): UserPromptSubmit 钩子匹配用户消息中的关键词
2. Skill Frontmatter: 每个 SKILL.md 包含带有触发关键词的 description 字段

Skill frontmatter 示例：

---
name: hub-audio
description: "Use for audio processing nodes in dora.
Triggers on: dora-microphone, dora-vad, whisper, kokoro-tts, VAD, 语音识别..."
---

当你询问 "dora-microphone" 或 "语音识别" 时，相应的 skill 会自动加载。

可用技能

类别	技能	说明
核心	dataflow-config, cli-commands	数据流 YAML 和 CLI 使用
API	node-api-rust, node-api-python, operator-api	节点和操作符开发
测试	integration-testing	使用 JSONL 输入测试节点
视觉	domain-vision	YOLO, SAM2, CoTracker, VLM
音频	domain-audio	Whisper, Kokoro TTS, VAD
机器人	domain-robot	机械臂控制、底盘、运动学
数据	data-pipeline	LeRobot 记录和回放
Hub	hub-nodes	预构建节点包

Hub 节点技能 (详细)

技能	说明	主要节点
hub-camera	摄像头采集	opencv-video-capture, dora-pyrealsense, dora-pyorbbecksdk
hub-audio	音频处理	dora-microphone, dora-vad, dora-distil-whisper, dora-kokoro-tts
hub-detection	检测/跟踪	dora-yolo, dora-sam2, dora-cotracker
hub-llm	语言模型	dora-qwen, dora-qwen2-5-vl, dora-internvl
hub-robot	机器人控制	dora-piper, dora-reachy2, dora-ugv, dora-rdt-1b
hub-visualization	可视化	dora-rerun (12种图元), opencv-plot
hub-recording	数据记录	llama-factory-recorder, lerobot-dashboard
hub-translation	翻译	dora-opus, dora-argotranslate

命令

• /new-dataflow - 创建新的数据流项目
• /add-node - 向现有数据流添加节点
• /visualize - 生成数据流可视化图

Agents

• dataflow-builder - 生成 dataflow YAML 配置
• node-debugger - 调试数据流问题

配合 Dora 使用

加载 skills 后，你可以让 Claude 帮助你：

# 创建视觉检测管道
"帮我构建一个带摄像头输入和 YOLO 检测的数据流"

# 搭建语音对话
"创建一个带麦克风、Whisper 和 TTS 的音频管道"

# 机器人遥操作
"我想用机械臂录制演示数据"

Dora 快速开始

1. 安装 dora CLI：

pip install dora-rs-cli
# 或者
cargo install dora-cli

2. 创建并运行数据流：

dora new my-project --lang python
dora build dataflow.yml --uv
dora run dataflow.yml

资源

• Dora 文档
• Dora GitHub
• 节点仓库

---

License

Apache-2.0