Guide · Agent skill
How to use the EU AI Act Readiness Check agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The ai-act-readiness skill is a structured review that an AI agent runs against an AI system to check it against the EU AI Act. It is aimed at compliance, product, privacy and legal teams, and at developers who want a first-pass map of which obligations might apply. After this guide you will know what the skill does, how to install it, how to ask for it in plain language, and what it needs to run.
EU AI Act Readiness Checkalirezarezvani/claude-skills · MIT Download the skill Quick start
- 1Download and unzip the skill. Download the ZIP from our page and unzip it into ~/.claude/skills/ for personal use or .claude/skills/ inside a project. In the Claude apps, you can upload it under Settings → Capabilities → Skills instead.
- 2Check the helper scripts exist. Confirm that the three Python helper scripts and cross_framework_mapper.py are present at the paths the skill references. They are not included in SKILL.md alone.
- 3Prepare your input files. Write a JSON description of the AI system and a JSON description of the roles your company plays. Be specific about what the system does and who it affects.
- 4Run the readiness check. Ask the agent in plain language or use /compliance-os:ai-act-readiness followed by the system name.
- 5Review the verdict. Read the colour-coded verdict, the Article citations and the top three actions. Stop immediately if the verdict is PROHIBITED.
- 6Send the legal list to counsel. Take the Legal Review Required section to a lawyer, especially for GPAI threshold, Article 5 boundary and Article 25 substantial-modification questions.
What the skill makes the agent do
The skill runs a six-question review of one AI system. Each question is tied to a specific article of the EU AI Act, so every verdict can be traced back to the text of the law. The six questions cover whether the system is a prohibited practice (Article 5), whether it is high-risk (Article 6 and Annex III), which conformity route applies (Article 43), what role your company plays (Article 25), whether transparency duties apply (Article 50), and whether it is a general-purpose or systemic-risk model (Articles 51-55).
- Ends with a colour-coded readiness verdict: READY-FOR-EU, GAPS-IDENTIFIED, NOT-READY or PROHIBITED.
- Lists the top three actions with an owner and an article-linked deadline.
- Lists article-level points to send to outside counsel.
- The skill describes itself as a structured checklist, not a substitute for legal advice.
Install the skill from the ZIP
Download the ZIP file for the skill from our page, then unzip it. The skill folder must end up in one of two places. For all your projects, use ~/.claude/skills/. For one project only, use the .claude/skills/ folder inside that project.
If you use the Claude apps instead of Claude Code, you can upload the skill under Settings → Capabilities → Skills.
- Personal install: unzip into ~/.claude/skills/
- Project install: unzip into .claude/skills/ inside your project folder
- Claude apps: Settings → Capabilities → Skills, then upload
Set up the helper scripts it needs
The skill's SKILL.md file calls three Python helper scripts: ai_system_risk_classifier.py, conformity_assessment_planner.py and ai_act_obligation_tracker.py. It also refers to cross_framework_mapper.py for cross-framework reuse. These scripts are not included in the SKILL.md file you download. The skill's references assume they exist inside the installed plugin, under the vendor path and the compliance-os scripts folder. Before relying on the skill, check that these files are present where the paths point. If they are missing, the skill's script steps will fail.
You also need Python installed, and a JSON description of the system and the roles your organisation plays.
- Python, to run the helper scripts
- A JSON file describing the AI system (for example systems.json or system.json)
- A JSON file describing roles (for example roles.json)
Trigger the skill with a request
You can call the skill with the slash command, naming the system you want to review. You can also ask in plain language, such as asking the agent to check whether a chatbot needs an EU AI Act review before launch. The skill is meant for three situations: AI-system intake, before EU deployment, and the annual compliance refresh.
Describe the system accurately. The verdict depends entirely on the inputs you give. A vague description produces a vague or unreliable verdict.
- Slash command: /compliance-os:ai-act-readiness <system>
- Plain request: 'Run an EU AI Act readiness check on our resume screening tool before we sell to German customers.'
- Plain request: 'Do an annual AI Act refresh for our three customer-facing AI systems.'
/compliance-os:ai-act-readiness <system>Read the output
The report starts with the decision being made and then gives a risk classification with its article citation and rationale. If the system is high-risk, a conformity assessment section follows. Next comes an obligation matrix, a transparency check for each Article 50 duty, and cross-framework reuse notes for ISO 42001, ISO 27001 and GDPR DPIAs. The final sections give the verdict, the top three actions and the legal review list.
Treat a PROHIBITED verdict as a hard stop. The skill says the system cannot be placed on the EU market under Article 5. Treat the legal review list as the questions to take to a lawyer, not as answers.
- Risk Classification: tier, citation, rationale, GPAI and systemic-risk flags
- Conformity Assessment: module, notified body requirement, Annex IV pack status
- Obligation Matrix: total obligations and deadline phases
- Transparency: Article 50(1) to 50(4) yes, no or not applicable
- Legal Review Required: ambiguities to send to outside counsel
Safety notes: scripts and network
The skill runs scripts on your machine. According to its description, it runs scripts and does not use the network. Read the helper scripts before running them on sensitive system descriptions, because the skill executes them from the plugin folder. Keep your system and role JSON files local unless your policy allows sharing them.
The skill does not replace legal advice. Its own text says novel cases need outside counsel, including disputes over the GPAI threshold, Article 5 boundary cases and Article 25 substantial-modification questions.
- Runs scripts: yes
- Uses network: no
- Do not treat the output as final legal sign-off
What people use it for
- Intake review of a new AI feature before it is offered to EU customers.
- Checking whether a customer-facing chatbot needs Article 50 interaction disclosure.
- Annual compliance refresh for a privacy or risk team that tracks several AI systems.
- Developer first-pass map of which EU AI Act obligations might apply to a model before a material change.
- Deciding whether a change that alters your company's role could make you a provider under Article 25.
Frequently asked questions
Is the ai-act-readiness skill free to use?
The source says the skill is released under the MIT licence. The source does not list prices, so check the licence text on the page where you downloaded it for the exact terms.
What do I need to install the EU AI Act readiness skill?
You need the skill folder placed in ~/.claude/skills/ or a project's .claude/skills/ folder, or uploaded in the Claude apps under Settings → Capabilities → Skills. You also need Python and JSON files describing your system and roles, plus the helper scripts the skill references.
Why does the skill say it cannot find scripts?
The helper scripts such as ai_system_risk_classifier.py are not included in the SKILL.md file. They must exist inside the installed plugin at the paths the skill references. Check those paths before running the skill.
Does the skill give legal advice?
No. The skill describes itself as a structured checklist and not a substitute for legal advice. It says novel cases, including GPAI threshold disputes and Article 5 boundary cases, need outside counsel.
Does the EU AI Act readiness skill send my data to the internet?
According to the skill's description, it does not use the network. It does run scripts on your machine, so review the scripts and keep sensitive system descriptions local.
What is the difference between this skill and a general compliance checklist?
This skill focuses on the EU AI Act alone and ties each verdict to a specific article. The source also lists related skills such as compliance-readiness for a multi-framework view and aims-audit for an ISO 42001 deep-dive.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use this skill if you are a compliance, product, privacy or developer team that needs a structured, article-cited first pass on EU AI Act obligations for a specific system. Pick something else, or bring in a lawyer from the start, if your organisation has no EU market exposure or your case involves a novel legal question, since the skill is not a substitute for legal sign-off.
Download the skill