I recently automated an entire Terraform workflow using an AI agent, and in this post I’ll walk through how agents and skills represent a fundamentally different way of working with AI than traditional chatbots. By the end you’ll understand the distinction — and how to install and use open-source agent skills yourself.

Chatbot vs agentic AI

A traditional chatbot is a one-shot interaction: you paste context in, get an answer, and paste the same context again next time. An agent retains file context, can fetch data from the internet, and maintains its working knowledge across a session through instruction files like CLAUDE.md.

Agents vs skills: the key distinction

  • An agent embodies professional expertise and purpose — who is doing the work.
  • A skill is a specific, detailed instruction on how to perform a particular task.

Skills are what make this reusable: instead of re-explaining your Terraform style guide in every session, you encode it once and the agent loads it when relevant, keeping the context focused.

Where to get skills: the HashiCorp agent-skills repo

HashiCorp publishes an agent-skills repository with skills for Terraform and Packer. You can install them two ways:

  1. Manually copy the skill folders into your project, or
  2. Use the npx installer.

Installation generates an .agents/ folder plus a skills-lock.json file so the whole team gets consistent skill versions.

Three demonstrations

I put the skills through three real tasks:

  1. Scaffolding — generated a full Terraform project using the style-guide skill, following HashiCorp’s own conventions out of the box.
  2. Testing — created unit and integration tests with the terraform-test skill.
  3. Refactoring — turned the project into a reusable module with the refactor-module skill, including the moved blocks needed to migrate existing state safely.

Wrap-up

Skills encode expert knowledge. With them, an agent can scaffold projects, write tests, and refactor infrastructure in a fraction of the time — and to a consistent standard. If you write Terraform for a living, this is worth an afternoon of experimentation.