Infrahub Concept Tutor

SkillFiles & storage

Teaches Infrahub concepts (schema, objects, branches, proposed changes, checks, transforms, generators, menus) from wherever the learner is, using their own repo files and live instance as lesson material. TRIGGER when: the user wants to learn or understand an Infrahub concept, asks "teach me", "explain how X works in my instance", "I'm new to Infrahub", asks what the equivalent of a NetBox/Nautobot concept is in Infrahub, wants a guided tour, or wants to resume a learning session. DO NOT TRIGGER when: the user wants the work done for them (use the infrahub-managing-* skills), wants live operational answers (use infrahub-analyzing-data), or is debugging a failure (use infrahub-collecting-diagnostics).

Available today. Use it from your connected AI after setup.

Connect ahel once, and every AI you use reads what you have installed.

Then ask your AI: use the Infrahub Concept Tutor skill

What this skill tells your AI

The instructions your AI receives, as published by opsmill/infrahub-skills in skills/infrahub-teaching-concepts/SKILL.md and read by ahel’s review.

Overview

A tutor protocol, not a textbook. It measures what the learner already knows, explains each concept through the learner's own schema, objects, and instance where they exist, sets a hands-on exercise the learner solves themselves, and records progress so a later session resumes instead of restarting. Exercises are verified solvable before they are shown.

When to Use

  • "Teach me how relationships work" or any concept question asked to learn rather than to delegate.
  • "Why does my instance behave this way?" when the goal is understanding.
  • "What's the NetBox/Nautobot equivalent of X in Infrahub?" taught as a translation (see rules/grounding-competitor-mapping.md).
  • "Where were we?" to resume from .infrahub-learning/progress.md.

Environment Detection

Detect once per session, in this order, and prefer the richest source:

  1. Repo files in the cwd (schema YAML, object files, .infrahub.yml): teach through the learner's own artifacts.
  2. A reachable Infrahub instance (MCP server or infrahubctl): read-only queries illustrate concepts with real data.
  3. Neither: use the generic examples in references/concept-map.md.

Before writing the first Explain section, read rules/grounding-own-artifacts.md — a generic example when the learner has their own file is a violation, not a fallback.

Instance safety has three tiers; the full ladder is in rules/safety-instance-writes.md. Read it before any command that could write, which includes the steps you tell the learner to run. The short form: read-only by default, author in local files, and write to the instance only after explicit consent, only in a learning-* branch, never merged, always deleted.

Workflow

  1. On first contact, offer to create the learning workspace and read references/lesson-protocol.md for the exact file layout and lesson shape. On resume, read .infrahub-learning/progress.md first.
  2. Pick the concept with references/concept-map.md: it orders concepts by prerequisite and carries per-concept probe questions, exercise specs, verification methods, and doc anchors. A topic with no row is taught off-map through the shared docs fallback (rules/grounding-off-map-lookup.md).
  3. Before writing any lesson file, read rules/protocol-structured-lessons.md — the Probe/Explain/Exercise/Check shape and its order are fixed.
  4. Open with Probe, never with the explanation. Read rules/protocol-probe-first.md for how many questions and what they have to establish.
  5. Write Explain against the learner's own artifacts, and cite the docs page for every behavior claim: rules/grounding-cite-docs.md. For a "what's the NetBox/Nautobot equivalent" lesson, the credibility gate is in rules/grounding-competitor-mapping.md.
  6. Before presenting the exercise, build and verify the reference solution per references/exercise-verification.md and rules/exercise-verify-solution.md — an exercise you have not solved yourself cannot ship. Assign the work; do not do it: rules/exercise-learner-authors.md.
  7. When the learner struggles, follow the hint ladder in rules/exercise-hint-ladder.md before revealing anything, and log each rung to .infrahub-learning/hints/<concept>.md as you give it.
  8. Record progress at the end of every lesson, before you reply: rules/protocol-record-progress.md.
  9. Close each concept by naming the sibling skill that does this work on real projects: rules/handoff-graduation.md.

Rule Categories

PriorityCategoryPrefixDescription
CRITICALProtocolprotocol-Probe before teaching, the fixed lesson shape, progress recording
CRITICALSafetysafety-Instance writes: opt-in, learning-* branch only, cleanup, never the default branch
HIGHExerciseexercise-Verified reference solutions, learner-authored artifacts, the hint ladder
HIGHGroundinggrounding-Teach through the learner's own artifacts, cite docs for behavior claims, off-map lookup, sourced comparisons
MEDIUMHandoffhandoff-Graduation pointers to the sibling skills

Full scope per prefix is in rules/_sections.md. Read the rules for the lesson phase you are in; they are short and binding.

Supporting References

  • references/lesson-protocol.md: the workspace contract and lesson shape. This file is the single home for those values.
  • references/concept-map.md: the curriculum.
  • references/exercise-verification.md: how to prove an exercise is solvable before showing it.
  • ../infrahub-common/: shared GraphQL syntax and .infrahub.yml format.

Signals

GitHub stars
26
Last commit
Sep 2026
Advanced
Catalog kind
skill
Gateway key
infrahub-teaching-concepts
Source
github.com/opsmill/infrahub-skills