The toughest a part of constructing in opposition to a brand new platform is instructing your instruments about it. Your coding agent doesn’t know the SDK’s conventions. Your IDE doesn’t know the CLI instructions. Your terminal doesn’t know the auth sample. Each hole is a context change, and each context change is time spent away from the work. DataRobot Expertise shut these gaps contained in the instruments you already use. Our market listings in Anthropic, Gemini and Cursor shut them on the set up step.
A Talent is a folder with a SKILL.md file. The frontmatter tells the agent when the ability applies. The physique tells it do the work. The agent hundreds solely the abilities related to the present activity, so context stays clear and reasoning stays sharp.
DataRobot Expertise are Agent Context Protocol definitions. They work throughout Claude Code, Cursor, Codex, Gemini CLI, Amp, VS Code Copilot, Goose, Letta, Kilo Code, and OpenCode. They don’t seem to be slash instructions and never MCP instruments. They’re the procedural information your agent wants to make use of the SDK, the CLI, and the platform accurately, each time.
The purpose of {the marketplace} listings is to take away the set up step totally so builders discover DataRobot for the time being they’re selecting instruments, not after they’ve already dedicated to a workflow.
The repo at present ships 10+ official Expertise, with new Expertise being added each week. Each corresponds to a core a part of the DataRobot agent constructing workflow.
| Talent | What it teaches the agent |
datarobot-setup |
The on-ramp. Installs the DataRobot CLI, Python SDK and Agent Help, configures the endpoint and API key, and verifies connectivity. Run this primary and also you don’t want to fret a couple of single factor round setup. |
datarobot-agent-assist |
The complete agent lifecycle. Generates agent_spec.md from a guided design dialog, rehearses software calls earlier than any code is written, scaffolds in opposition to the Agentic Starter template, runs native checks, and deploys to the DataRobot platform. |
datarobot-model-training |
Venture creation, AutoML configuration, goal leakage checks, partitioning patterns. |
datarobot-predictions |
Batch and real-time prediction era, template scaffolding for prediction APIs. |
datarobot-model-deployment |
Deploying and managing fashions, together with governance settings and deployment metadata. |
datarobot-feature-engineering |
Characteristic evaluation, transformations, derived function workflows. |
datarobot-model-monitoring |
Efficiency and knowledge drift monitoring, accuracy monitoring, alert configuration. |
datarobot-model-explainability |
Prediction explanations, function impression, mannequin diagnostics. |
datarobot-data-preparation |
Dataset add, validation, schema checks, and registry workflows. |
datarobot-app-framework-cicd |
CI/CD pipelines for DataRobot software templates. |
datarobot-external-agent-monitoring |
OpenTelemetry instrumentation for exterior brokers reporting into DataRobot. |
Two of those change the expertise probably the most. datarobot-setup is the on-ramp: earlier than it existed, a developer who put in the Expertise nonetheless needed to manually authenticate, level the SDK on the proper base URL, and make sure every thing was wired up. Now the setup part turns into one other factor the agent does, not one other factor the developer does.
datarobot-agent-assist brings the spec-driven design loop into the identical context. As a substitute of switching to a unique software to design an agent, the developer asks for assist, the ability prompts, and dr help runs from inside the identical IDE dialog, producing an agent_spec.md and rehearsing the software calls earlier than any code is written. Design, check and deploy, all contained in the agent loop.
Cursor
Open the Cursor market entry for DataRobot, click on the “Add to Cursor” button, and Cursor handles the remaining. The Expertise register in opposition to the workspace and change into obtainable in any chat. In case you’d relatively pin to the repo and version-control the set up, open the repo as your workspace and Cursor reads AGENTS.md mechanically.
Gemini CLI
Gemini CLI now treats DataRobot as an extension with bundled Expertise. Out of your terminal:
gemini extensions set up https://github.com/datarobot-oss/datarobot-agent-skillsThe Expertise land in ~/.gemini/extensions/datarobot-agent-skills/expertise and cargo on session begin. Use /expertise listing inside a Gemini session to verify. Surroundings variables propagate mechanically.
Claude Plugins
For Claude customers, DataRobot Expertise can be found by way of the plugin market itemizing and you’ll run this command:
claude plugin set up datarobot-agent-skills@claude-plugins-officialCommon Directions
The common installer is the reply in case you are operating an AI IDE or CLI not listed above:
npx ai-agent-skills set up datarobot-oss/datarobot-agent-skills{The marketplace} listings are step one in distributing DataRobot’s developer floor the identical means fashionable infrastructure instruments distribute theirs. Count on the catalog to develop: extra expertise across the agent lifecycle, extra bundled flows for Agent Help, deeper protection of governance and observability patterns that immediately dwell in docs relatively than in agent context.
If you wish to see what’s there now or contribute a sample your staff makes use of, the supply of reality is the repo: github.com/datarobot-oss/datarobot-agent-skills. {The marketplace} listings monitor it.
The developer expertise DataRobot is constructing is one the place the platform reveals up within the floor you already selected, with the on-ramp baked in. Expertise are how that promise reaches the agent in your IDE. The marketplaces are the way it reaches you.
