Infrahub Skills is a set of AI skills for infrastructure engineers and platform teams that gives your AI coding assistant built-in Infrahub expertise.
You describe what you want in plain language and the Skills produce valid best-practice schemas, Generators, Transformations, checks, and more. Whether you’re prototyping a new use case, extending an existing implementation, or onboarding a new engineer, Skills help you get more out of Infrahub, faster.
Bring Infrahub expertise to how you already work
Infrahub is an AI-ready data management platform for network and infrastructure automation at scale. Its extensible, schema-first architecture lets you model any domain, version every change, and drive reliable automation across production environments. With that flexibility comes a set of conventions and best practices.
Teams typically pick up expertise in using a platform over time, through documentation, example projects, and conversations with more experienced colleagues. Gathering expertise this way works, and it’s how platforms and systems are learned. But the process of gathering expertise in this way can be time-consuming and sequential, and the expertise gained on one project doesn’t automatically carry over to the next engineer who joins the team.
AI accelerates the ability for engineers to learn and use a platform. With Infrahub Skills, you can now embed platform expertise directly into the AI coding assistants you already use so that anyone building with Infrahub gets the benefit of conventions and best practices they haven’t yet learned, applied automatically to their own work and explained along the way.
What it looks like to build with Infrahub Skills
Infrahub Skills are useful whether you’re just getting started, deep in a build, or running Infrahub in production.
Prototyping a new use case
Describe what you want to model in plain language and have an AI agent build it in Infrahub. No need to master the schema format first. What used to take weeks of ramp-up to build a proof of concepts can happen in hours (or minutes!).
Actively building
Infrahub Skills shorten the loop between ideation and having a working implementation. Infrahub conventions are applied by an AI agent at the point of generation, so output is built correctly the first time. This minimizes the rework loop. Six Skills cover the core build workflow:
- Schema Manager turns a plain-language description of your domain into a valid schema, with naming conventions, relationships, and generics applied automatically.
- Object Manager populates your schema with data files that load in the right order.
- Generator Manager builds Generators that create infrastructure objects from design definitions.
- Transform Manager creates Python or Jinja2 Transformations that turn Infrahub data into device configs, reports, and other formats.
- Check Manager writes Python validation checks that run in Proposed Change pipelines.
- Menu Manager shapes the web UI to match how your team thinks about your data.
Extending a production implementation
Infrahub Skills help you keep iterating. They help you add new features and build integrations with external systems like spreadsheets, CMDBs, NetBox, and Nautobot, which is historically one of the most time-consuming parts of any Infrahub project.
Additionally, when new engineers join the team, Skills allow Infrahub expertise to travel with the repo rather than being trapped in Slack history. Two Skills are built specifically for the ongoing management of a production implementation:
- Data Analyzer queries a running Infrahub instance in plain language to provide cross-node correlation, drift detection, blast-radius analysis, data quality audits. The Data Analyzer allows AI skills to do this using an MCP server, so you don’t have to handroll GraphQL queries to do so.
- Repo Auditor surfaces where a project has drifted from best practices over time and reports what needs to be adjusted.
Two ways to work with Infrahub Skills
Not every task needs the same treatment. Infrahub Skills support different modes, matched to task complexity:
- Direct mode: Using direct mode, you interact with your AI agent directly and it uses Skills on the back end. Use this mode for well-scoped, single-skill tasks. Describe what you need and the AI matches it to the right Skill and produces it. This mode is best used for adding an attribute, creating a check, or populating a batch of objects.
- Spec-driven development (SDD): SDD isn’t unique to Infrahub Skills. It’s becoming a widely used method of invocation to produce high quality outputs from LLMs. SDD uses a Socratic method to build specifications that the LLM then goes and deploys. Use this mode for complex, multi-part builds. The AI reasons through requirements with you first, validates the approach against Infrahub conventions, and lays out a task plan for review. Nothing is generated until you approve the plan.
Across both modes, the AI explains the reasoning behind every recommendation. The result is that you build real Infrahub expertise as you work.
How to use Infrahub Skills
Infrahub Skills are ingested as context to whichever AI coding assistant you use when they’re relevant to the task at hand. You simply need to install them to make them available either globally or into the project directory you’re working with. After they’re installed, you can describe a task in plain language and let the AI match it to the right Skill automatically, or invoke one directly if you know which one you want.
Infrahub Skills follow the open Agent Skills format, so they work with Claude Code, GitHub Copilot, Cursor, Windsurf, Amp, Cline, and Codex—any AI tool that reads custom context files from the project directory.
Getting started with Infrahub Skills
Head over to the Infrahub Skills documentation for installation options for your AI tool of choice, a walkthrough of each Skill, and more on how to get the most out of them. Have a look at our 45-minute technical walkthrough if you want to see Infrahub Skills in action, from a clean repo to putting Skills to work.
Community contributions are welcome! New rules, examples, and improvements can be submitted through a pull request on the GitHub repo.
