EU AI Act Reviewer Skill
A portable, read-only Skill that reviews supplied product evidence, flags EU AI Act provisions that may matter, and prepares focused questions without issuing a legal or compliance verdict.
What it helps with
Teams reviewing an AI product need to understand which parts of the EU AI Act may deserve attention, but the answer depends on the product’s intended purpose, the organisation’s role, who may be affected, where it is used, and when it will be deployed. A quick answer based on a product name or marketing page can create more confidence than the evidence supports.
What was built
The Reviewer Skill works from evidence supplied by the user and remains read-only by default. It establishes the review basis, checks the current official law, and separates what can be seen in source code, called code, tests, deployment, and visible behaviour. Its findings distinguish observed facts from assumptions, attach the relevant official provisions and dates, identify missing evidence, and turn the remaining uncertainty into focused operational or legal questions.
What made it different
The most important part is what the Skill refuses to do. It does not issue legal advice, legal clearance, a final risk classification, a compliance verdict, or a fine prediction. Instead, it makes the evidence and limits of the review visible, giving a human reviewer or legal adviser a clearer starting point without disguising issue-spotting as certainty.
A compliance review should narrow uncertainty and expose its evidence, not offer false clearance.
What it demonstrates
- Read-only review of supplied product evidence
- Current official sources, application dates, and evidence gaps
- Focused next questions for operational and legal review