Agent skills based on the MLOps Coding Course
Overview
This project contains a collection of designed to accompany the MLOps Coding Course. These skills are tailored to help AI coding assistants (like Antigravity, Gemini CLI, GitHub Copilot, and Claude Code) understand and guide you through the best practices of MLOps, from prototyping to production. Agent Skills are specialized context or instructions that you can load into your AI assistant. Each skill is a folder containing a SKILL.md file (YAML frontmatter + Markdown instructions) that provides the "Know-How" for a specific task, ensuring the AI follows the course's methodology, standards, and patterns. Each skill derives from one chapter of the MLOps Coding Course: Skills follow the open SKILL.md standard, so most agents discover them from a well-known folder. The workspace-standard location is .agents/skills/; each tool then reads from its own path (see the table below). 1. to a location on your machine (e.g. ~/mlops-coding-skills). 1.
README
MLOps Coding Skills
Welcome to the MLOps Coding Skills repository! 🚀
This project contains a collection of Agent Skills designed to accompany the MLOps Coding Course. These skills are tailored to help AI coding assistants (like Antigravity, Gemini CLI, GitHub Copilot, and Claude Code) understand and guide you through the best practices of MLOps, from prototyping to production.
🧠 What are Agent Skills?
Agent Skills are specialized context or instructions that you can load into your AI assistant. Each skill is a folder containing a SKILL.md file (YAML frontmatter + Markdown instructions) that provides the “Know-How” for a specific task, ensuring the AI follows the course’s methodology, standards, and patterns.
📚 Skills in this repository
Each skill derives from one chapter of the MLOps Coding Course:
| Skill | Course chapter | Focus |
|---|---|---|
mlops-initialization |
1. Initializing | Set up a robust project structure with uv, git, and VS Code. |
mlops-prototyping |
2. Prototyping | Write clean, reproducible Jupyter notebooks and prevent data leakage. |
mlops-industrialization |
3. Productionizing | Turn prototypes into a distributable package (src layout, hybrid paradigm). |
mlops-validation |
4. Validating | Add typing (ty), linting (ruff), testing (pytest), logging, scans. |
mlops-automation |
5. Refining | Automate tasks, containers, CI/CD, and experiment tracking (MLflow). |
mlops-collaboration |
6. Sharing | Prepare the project for sharing: license, docs, templates, releases. |
mlops-observability |
7. Observability | Add reproducibility, lineage, monitoring, alerting, and explainability. |
🛠️ Installation & Usage
Skills follow the open SKILL.md standard, so most agents discover them from a well-known folder. The workspace-standard location is .agents/skills/; each tool then reads from its own path (see the table below).
Recommended setup (symlinks)
-
Clone this repository to a location on your machine (e.g.
~/mlops-coding-skills). -
Create the skills directory in your target project or workspace, then symlink the skills you need:
mkdir -p .agents/skills # Symlink every MLOps skill (or pick individual folders): ln -s ~/mlops-coding-skills/mlops-* .agents/skills/ -
Expose them to your agent using the per-tool path below.
Per-tool install paths
| Tool | Workspace path | Setup |
|---|---|---|
| Antigravity | .agents/skills/ |
Discovers the linked folders directly. |
| Gemini CLI | .agents/skills/ or .gemini/skills/ |
Discovers the linked folders; remote installation is an alternative. |
| Claude Code | .claude/skills/ |
Link the shared directory: mkdir -p .claude && ln -s ../.agents/skills .claude/skills. |
| GitHub Copilot | .agents/skills/, .github/skills/, or .claude/skills/ |
Discovers skills and loads relevant instructions on demand. |
For a Gemini-managed workspace installation instead of symlinks:
gemini skills install https://github.com/MLOps-Courses/mlops-coding-skills --scope workspace
Claude personal skills can live in ~/.claude/skills/; plugin skills live under the plugin’s skills/. In Claude Desktop, add the relevant files to Project Knowledge.
See the current Antigravity, Gemini CLI, Claude Code, and Copilot documentation for host-specific behavior.
🤝 Contributing
We welcome contributions! If you have a new MLOps pattern or improvement, please open a Pull Request.
Install the pinned toolchain and git hooks once with mise install && mise run install, then run the gate before pushing:
mise run all # format, then check — CI runs this exact task and nothing else
mise run check covers formatting (dprint), workflow linting (actionlint, zizmor), secret scanning (gitleaks), a filesystem scan (trivy), the frontmatter rule (name equals the folder name), and validation against the Agent Skills specification (gh skill publish --dry-run).
See CONTRIBUTING.md and AGENTS.md for the repository conventions.
Happy Coding! 🤖✨
Release tags matching v* are protected by the active release-tag ruleset: existing tags cannot be updated or deleted, and there are no bypass actors. Create a new version tag for corrections; this rule permits new tag creation and does not publish a release.
Recommended Tools
Try a different keyword or remove a filter.
Install
npx skillfish add mlops-courses/mlops-coding-skills