Built by Klara Hermesz, Co-founder and Chief Learning Architect at AI Enablement Academy.
An AI competency assessment that runs as a conversation, not a quiz. Ask it where you honestly stand with AI, and it places you on the AI Competency Framework, reflects that placement back in plain language, and helps you pick one thing worth learning next. Team leads use it the same way to get a real read on team AI adoption and figure out where upskilling should focus, before it turns into a formal program.
Anyone who's ever wondered if they're behind on AI and wants an honest answer instead of a guess. Individually, it's for the person who uses AI daily but has never had a clear picture of how that adds up. For a manager, team lead, or L&D person, it sketches where a team actually sits across AI maturity levels, so the next training decision is based on evidence instead of a hunch.
The AI Competency Framework was built and published openly by AI Enablement Academy after running it inside real training cohorts. Two axes describe AI maturity without flattening it into one score:
- Enablement Layer (across): the impact radius of what you do with AI. Augmentation (your own productivity) → Automation (your team's processes) → Agentic (whole systems).
- Capability Level (down): how deep your expertise goes. Foundations → Essentials → Fluency → Native.
Twelve cells, each a real, nameable capability. You don't have one global level. You sit in different cells across the three columns, and that's the point. It's a map, not a ladder.
The repo root is already shaped as the skill, so clone it straight into your skills folder and it loads as a plugin next session:
git clone https://github.com/klarahermesz/ai-competency-compass.git ~/.claude/skills/ai-competency-compassUse .claude/skills/ inside a single project instead if you want it scoped there.
These need a zip you upload once. It then stays available in every conversation.
- Clone this repo (or download it as a zip and unpack it).
- Build the skill zip yourself:
bash build-skill-zip.sh. It readsSKILL.md,references/,assets/, and the license files, and writesai-competency-compass.zipin the layout Claude's uploader expects. - Go to Customize → Skills (claude.ai/customize/skills). In the desktop app and in Cowork, Customize sits in the left sidebar.
- Click Add, then select the zip you just built.
- Start a conversation and describe your situation, or type
/and pick the compass from the list.
If the upload or the skill itself doesn't work, check Settings → Capabilities (claude.ai/settings/capabilities) and turn on Code execution and file creation. Skills need it. On Team and Enterprise plans, an owner enables it under Organization settings → Skills instead. Skills are available on every plan, Free through Enterprise (Anthropic's docs).
You don't have to name the skill or explain the framework. Just say what's on your mind:
where am I honestly on the whole AI thing?
help me figure out my team's AI skill gaps
I feel behind on AI. What should I learn next?
From there, it asks about work you've actually done rather than how you'd rate yourself. Once it has enough evidence, it names where you sit on each axis, shows you the map, and closes on a single next step you can start this week, not a list of five things you'll do someday.
build-skill-zip.sh # builds the upload-ready zip from the files below
SKILL.md # the coaching flow
references/
├── framework.md # the 12 cells, verbatim
├── diagnostic-questions.md # evidence-based placement probes
├── progression-map.md # what "next" looks like from each cell
├── team-lens.md # team-distribution diagnosis
├── learning-pathways.md # self-serve resources (+ disclosure below)
└── grid-map.md # cell-to-course map, read only when you ask about programs
assets/
├── framework.png
└── result-card.html # branded shareable result-card template
The repo root is the skill folder, which is why cloning it straight into ~/.claude/skills/ works. Running build-skill-zip.sh produces the same files plus LICENSE, wrapped in an ai-competency-compass/ folder for the uploader, so you always get a zip built fresh from the current source instead of one that's quietly drifted out of date.
- Evidence over self-rating. People misjudge their own level in both directions, so the skill asks about artifacts and habits, never "what level are you?"
- Map, not ladder. Placement per column. Foundations is a respectable place to be.
- One next step. Every conversation ends with a single action you can start this week. A list of five improvements is a list of zero.
- Honest limits. It diagnoses a person's capability, not organizational readiness (governance, data, and sponsorship are out of scope), and team snapshots are labeled as one-perspective sketches.
This skill is published by AI Enablement Academy, which runs cohort training programs mapped to the framework. It's built to be fully useful without them: it leads with self-serve learning paths and mentions the Academy at most once, at the end, as one option among others. A cell-by-cell course map lives in references/grid-map.md, but the skill only opens it when you explicitly ask about training programs. It never volunteers course recommendations. Links to the Academy carry UTM tags so we can see whether the skill is useful, and no other tracking of any kind. If you find it pitching harder than that, that's a bug. Please open an issue.
Apache License, Version 2.0 applies to material authored by AI Enablement Academy; see LICENSE and NOTICE for attribution. Bundled third-party components carry their own licenses; see THIRD-PARTY-LICENSES.md (html2canvas, MIT).
Issues and PRs are welcome, especially better diagnostic probes, placement edge cases you've hit in real conversations, and translations. The framework itself evolves in public. If you think a cell is wrong, argue with evidence and it may well change.