FiftyOne Skills
Voxel51 skills for computer vision: import datasets, find duplicates, run inference and evaluate models in FiftyOne
- 48 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- Skills with MCP run FiftyOne operators that change datasets and remove duplicates
- The skill that connects to a Voxel51 deployment acts with your account permissions
- The universal installer runs through `curl | sh`, so read the script before running it
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/fiftyone-skillsDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Skills with MCP run FiftyOne operators that change datasets and remove duplicates
- The skill that connects to a Voxel51 deployment acts with your account permissions
- The universal installer runs through `curl | sh`, so read the script before running it
This is third-party code. Review the repository files before installing.
What it does
A set of 19 Voxel51 skills for computer vision datasets in FiftyOne. The agent imports and exports data in COCO, YOLO, VOC and CVAT formats, finds duplicate images, runs models from the Zoo, computes mAP, precision and recall, plots a two-dimensional embeddings map and spots outliers. Seven skills work through FiftyOne MCP Server and call its operators: import, export, duplicate search, inference, model evaluation, embeddings visualization and dataset curation. The rest need no MCP: FiftyOne plugin development, code style, Jupyter notebooks, issue triage, troubleshooting of common failures, plugin evaluation and connecting to a Voxel51 deployment. Each skill follows the Agent Skills specification: a SKILL.md file with strict directives and a step-by-step workflow.
Who it is for. Computer vision engineers who manage datasets and models in FiftyOne, and authors of plugins for it.
Good fit when
- A dataset is suspected of holding duplicates, and the agent finds similar images by embeddings and removes the extras
- A COCO or YOLO dataset has to be loaded into FiftyOne and run through a detector from the Zoo
- After training you need mAP, a confusion matrix and a list of false positives per class
- A FiftyOne plugin with an operator or a panel is being written, and the agent needs the structure and style rules
- The FiftyOne App fails to open or loses a dataset, and common causes need to be checked
Not a fit when
- Datasets are kept outside FiftyOne: the skills are tied to its operators and SDK
- FiftyOne MCP Server cannot be installed: seven of the nineteen skills do not work without it
- You need an annotation platform: CVAT appears in the skills only as an export format
Example request
Find and remove duplicate images from my dataset.Limitations
Skills marked MCP work only together with FiftyOne MCP Server, which is installed separately with `pip install fiftyone-mcp-server`. The skills target FiftyOne 1.10 and newer. The server must live in the same Python environment as FiftyOne, otherwise the agent will not find the `fiftyone-mcp` command. Brain operators for duplicate search need a running FiftyOne App, and the agent waits 5–10 seconds for it to start.
How to disable. Delete the skill folder from the agent's skills directory, or run `/plugin uninstall fiftyone-find-duplicates@fiftyone-skills` in Claude Code. The MCP server is removed with `claude mcp remove fiftyone`.
FAQ
How do I install FiftyOne Skills in Claude Code?
Through plugins: `/plugin marketplace add voxel51/fiftyone-skills`, then `/plugin install fiftyone-find-duplicates@fiftyone-skills` for the skill you need. The universal installer `curl -sL skil.sh | sh -s -- voxel51/fiftyone-skills` offers to pick skills, agents and the install scope.
Is FiftyOne MCP Server required?
Seven skills need it: import, export, duplicates, inference, model evaluation, embeddings and dataset curation reach FiftyOne through the server's operators. The skills for plugins, code style, notebooks and issue triage work without it.
How do I connect FiftyOne MCP Server to Claude Code?
First `pip install fiftyone-mcp-server`, then `claude mcp add fiftyone -- fiftyone-mcp`. If FiftyOne lives in a virtual environment, activate it or give the full path to `fiftyone-mcp`.
Which agents do the skills work with?
The installer supports Claude Code, Cursor, Codex, OpenCode, GitHub Copilot, Amp, Antigravity, Roo Code, Kilo Code and Goose. Gemini CLI has an extension: `gemini extensions install https://github.com/voxel51/fiftyone-skills.git --consent`.
Which skills work without the MCP server?
Plugin development, VOODO Design, code style, notebook creation, issue triage, troubleshooting, plugin evaluation, Zoo remote models, the Data Lens connector, the Voxel51 connection, App Playwright and SDK documentation search. The other seven reach FiftyOne through MCP.
Discussion
No questions or reviews yet.
Related
SkillsA skill that picks a scientific chart from nearly 100 templates for CSV or Excel data and outputs PNG, SVG, PDF and Python code.
Around 200 skills for Claude Code on bioinformatics, chemistry and biostatistics, from RNA-seq analysis to molecular docking
NVIDIA BioNeMo Agent Toolkit: skills for life sciences
NVIDIA BioNeMo Agent Toolkit
NVIDIA skills for biology and chemistry: the agent predicts protein structures, docks molecules and designs proteins
The Zvec skill guides an agent through the embedded vector database: collection schema, hybrid search, RAG and multimodal queries
DDC Skills for Construction
DDC Skills for AI Agents in Construction
238 skills for construction companies: BIM to Excel export, cost item search, estimates, site reports and document generation
166 skills for scientific work: bioinformatics, cheminformatics, clinical data, geospatial analysis and 100+ databases
README in short
The repository presents the skills as packaged workflows and lists all 19 in a table that marks which need MCP. Next come three ways to install, connecting the MCP server to Claude Code, Claude Desktop, Cursor and VS Code, the SKILL.md format and contribution rules.
README badge
Original SKILL.md
---
name: fiftyone-find-duplicates
description: Finds duplicate or near-duplicate images in FiftyOne datasets using brain similarity computation. Use when deduplicating datasets, finding similar images, or removing redundant samples.
---
# Find Duplicates in FiftyOne Datasets
## Key Directives
**ALWAYS follow these rules:**
### 1. Set context first
```python
set_context(dataset_name="my-dataset")
```
### 2. Launch FiftyOne App
Brain operators are delegated and require the app:
```python
launch_app()
```
Wait 5-10 seconds for initialization.
### 3. Discover operators dynamically
```python
# List all brain operators
list_operators(builtin_only=False)
# Get schema for specific operator
get_operator_schema(operator_uri="@voxel51/brain/compute_similarity")
```
### 4. Compute embeddings before finding duplicates
```python
execute_operator(
operator_uri="@voxel51/brain/compute_similarity",
params={"brain_key": "img_sim", "model": "mobilenet-v2-imagenet-torch"}
)
```
### 5. Close app when done
```python
close_app()
```