FiftyOne Skills

</p>

Expert workflows for computer vision powered by AI assistants

Documentation · MCP Server · FiftyOne Plugins · Discord

</div>

What are Skills?

Skills are packaged workflows that teach AI assistants to perform complex computer vision tasks autonomously. Combined with the FiftyOne MCP Server, you can find duplicates, run inference, and explore datasets using natural language.

"Find and remove duplicate images from my dataset"
"Import this COCO dataset and run object detection"
"Visualize my embeddings and identify outliers"

Skills bridge the gap between natural language and FiftyOne's 80+ operators, providing step-by-step guidance that AI assistants follow to complete complex workflows.

Available Skills

Skill	Description	MCP
📥 Dataset Import	Universal import for all media types, label formats, multimodal groups, and Hugging Face Hub	Yes
📤 Dataset Export	Export datasets to COCO, YOLO, VOC, CVAT, CSV, Hugging Face Hub, and more	Yes
🔍 Find Duplicates	Find and remove duplicate images using brain similarity	Yes
🤖 Dataset Inference	Run Zoo models for detection, classification, segmentation, embeddings	Yes
📈 Model Evaluation	Compute mAP, precision, recall, confusion matrices, analyze TP/FP/FN	Yes
📊 Embeddings Visualization	Visualize datasets in 2D, find clusters, identify outliers	Yes
🔌 Develop Plugin	Create custom FiftyOne plugins (operators and panels)	—
🎨 VOODO Design	Build UIs with VOODO React components and design tokens	—
📝 Code Style	Write Python code following FiftyOne's official conventions	—
📓 Create Notebook	Create Jupyter notebooks: getting-started guides, tutorials, recipes, ML pipelines	—
🏷️ Issue Triage	Triage GitHub issues: validate status, categorize, generate responses	—
🧹 Dataset Curation	End-to-end curation: quality checks, annotation audit, duplicates, class distribution, splits	Yes
🔧 Troubleshoot	Fix common issues: dataset persistence, App connection, MongoDB errors, codecs, performance	—
🛡️ Eval Plugin	Evaluate plugins for quality, security, and agent-readiness. Produces a structured report	—
🧩 Zoo Remote Model	Build remote model zoo integrations that work with `register_zoo_model_source` and `dataset.apply_model`	—
🔌 Generate Data Lens Connector	Generate a Data Lens connector from an external database schema (PostgreSQL, BigQuery, MySQL, etc.)	—
🎭 App Playwright	Drive the FiftyOne App via the Playwright MCP for operator/plugin verification, demos, and end-to-end UI automation	—

Quick Start

Step 1: Install Skills

Universal Installer (Recommended):

curl -sL skil.sh | sh -s -- voxel51/fiftyone-skills

Interactive prompts let you select skills, agents, and install scope (project or global).

Supported agents: Claude Code, Cursor, Codex, OpenCode, GitHub Copilot, Amp, Antigravity, Roo Code, Kilo Code, Goose

Claude Code:

# Register the skills marketplace
/plugin marketplace add voxel51/fiftyone-skills

# Install a skill
/plugin install fiftyone-find-duplicates@fiftyone-skills

Gemini CLI:

gemini extensions install https://github.com/voxel51/fiftyone-skills.git --consent

Step 2: Use It

"Write a FiftyOne plugin that displays model confidence"
"Write Python code following FiftyOne conventions"

Your AI assistant will automatically load the skill instructions and execute the workflow.

Step 3: Set Up MCP Server (Optional)

Skills marked with MCP in the table above require the FiftyOne MCP Server to interact with datasets and run operators.

pip install fiftyone-mcp-server

⚠️ Important: Make sure to use the same Python environment where you installed the MCP server when configuring your AI tool. If you installed it in a virtual environment or conda environment, you must activate that environment or specify the full path to the executable.

Then configure your AI tool:

<details> <summary><b>Claude Code</b> (Recommended)</summary>

claude mcp add fiftyone -- fiftyone-mcp

</details> <details> <summary><b>Claude Desktop</b></summary>

Edit ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "fiftyone": {
      "command": "fiftyone-mcp"
    }
  }
}

</details> <details> <summary><b>Cursor</b></summary>

Add to ~/.cursor/mcp.json:

{
  "mcpServers": {
    "fiftyone": {
      "command": "fiftyone-mcp"
    }
  }
}

</details> <details> <summary><b>VSCode</b></summary>

Add to .vscode/mcp.json:

{
  "servers": {
    "fiftyone": {
      "command": "fiftyone-mcp"
    }
  }
}

</details>

Once configured, you can use MCP-dependent skills:

"Find and remove duplicate images from my dataset"
"Import this COCO dataset and run object detection"

Skill Structure

Each skill follows the Agent Skills specification:

skills/
└── fiftyone-find-duplicates/
    └── SKILL.md                     # Instructions for AI

SKILL.md format:

---
name: skill-name
description: When to use this skill
---

# Overview
What this skill does

# Prerequisites
Required setup

# Key Directives
ALWAYS/NEVER rules for AI

# Workflow
Step-by-step instructions

# Troubleshooting
Common errors and solutions

Contributing

We welcome contributions! Whether you want to add a new skill, improve an existing one, or help with integrations and tooling — there's a place for you here.

See CONTRIBUTING.md for the full guide, including skill structure requirements, the quality bar, and how to test your skill before submitting.

Looking for ideas? Browse issues labeled help wanted or good first issue, or check the project milestones for planned work.

Feedback

Help us improve FiftyOne Skills!

Just ask your AI assistant:

"Help me submit feedback about [your issue]"

The agent will automatically gather session context, environment info, and can submit directly via gh CLI or generate content to paste at Submit Feedback

Resources

Resource	Description
FiftyOne Docs	Official documentation
FiftyOne MCP Server	MCP server for AI integration
FiftyOne Plugins	Official plugin collection
Agent Skills Spec	Skills format specification
PyPI Package	MCP server on PyPI
Discord Community	Get help and share ideas

🧡 Community

Join the FiftyOne community to get help, share your skills, and connect with other users:

Discord: FiftyOne Community
GitHub Issues: Report bugs or request features