docker/skills
A collection of Docker skills for AI coding agents to help them build, test, debug, and optimize containerized apps with consistent, reusable workflows.
Docker Skills for AI Coding Agents
What it is – A collection of Agent Skills that teach AI‑coding assistants (Claude Code, OpenAI Codex, GitHub Copilot, Gemini CLI, Cursor, etc.) how to work with Docker. Each skill lives in its own SKILL.md folder and follows the open‑source Agent Skills specification.
Why it matters – By installing these skills, an AI assistant gains concrete, Docker‑specific guardrails, best‑practice advice, and command templates. This lets the model suggest or execute Dockerfile changes, Compose files, sandbox lifecycle commands, or Docker‑Agent configuration with far fewer hallucinations or unsafe actions.
Key parts
- Skill catalog –
catalog.yamldefines product groups (Dockerfile & Build, Compose, Sandboxes, Docker Agent, cross‑product guardrails) and the individual skill IDs. - Skill directories – Under
skills/each folder (e.g.,docker-compose-patterns) contains aSKILL.mdwith description, inputs, outputs, and validation rules. - Installation – A
skillsCLI (npx skills add docker/skills …) or native marketplace plugins for each supported agent install the chosen skills into the agent’s skill directory. There are also manual copy and sandbox‑native options. - Guardrails – The
docker-destructive-guardrailsskill forces confirmation before any irreversible Docker operation. - Experimental features – Some sandbox‑related skills (
docker-sandboxes-env,docker-sandboxes-kits) are marked experimental and may change.
How to use
- Run the
skillsCLI or install via the agent’s marketplace. - Choose the skill(s) you need (e.g.,
docker-compose-patterns). - The agent will automatically load the skill, giving you context‑aware suggestions or safe command execution for that Docker task.
Development – The repo uses Taskfile.yml for CI, validation, and catalog regeneration. Contributions must keep the skill catalog and version numbers in sync.
License – Apache 2.0.
Related
- Project
- Project
- Project
- Project
- Project