hochrisiko-bestaetigt-end-to-end-roadmap

What it does

The skill organizes a Hochrisiko-KI validation process into a structured end-to-end roadmap. It defines a sequence of twelve steps spanning risk management, data governance, technical documentation, logging, transparency for operators, human oversight, accuracy and cybersecurity, quality management, conformity assessment, CE marking, EU database registration, and post-market surveillance. It results in a project plan with actors and deadlines and separates role, deadline, responsibility, evidence, and the desired output.

How it works

It instructs the agent to identify and document: the roles (who acts, what decision is due, what deadline applies, what output is needed); deadlines and urgency (marking the timelines from regulatory articles); the norms to verify (articulating applicable EU KI-VO articles and related references); responsible entities (provider, operator, importer, trader, authorized representative, and regulatory bodies); and required documents and evidence (conformity declarations, technical documentation, RMS, data governance, FRIA, EU database registrations). The twelve steps are enumerated with explicit outputs and corresponding detail-skills to consult for each step, such as RMS, data quality, technical documentation, logging, transparency, human oversight, accuracy, quality management, conformity assessment, CE marking, EU database registration, and post-market monitoring. It also provides optional operator-focused steps for deployment scenarios.

When to use it

Use this when an AI system is classified as Hochrisiko under Art. 6 and related annexes, and you need a regulator-aligned, stepwise plan to achieve CE marking and market release, including identifying required evidence, responsible roles, and timelines. The workflow emphasizes separating role, deadlines, responsibility, evidence, and output, and references the twelve-step sequence to follow.

What it can touch

This skill references actions and outputs such as: RMS, data governance documents, Anhang IV technical documentation, logging configurations, transparency declarations for operators, human supervision capabilities, accuracy and cybersecurity measures, quality management system, conformity assessment procedures, EU conformity declarations, CE marking, EU database registration entries, and post-market monitoring plans. It mentions required artifacts and the related detail-skills for each step, but does not perform actual regulatory actions itself.

Caveats

No legal advice is provided. The skill notes that the workflow is mechanical and relies on internal compliance and external conformity assessment, with day-to-day regulatory changes requiring live checks for updates. It instructs to verify regulatory references and avoid citing model knowledge. The description includes explicit references to articles and enforcement timelines and cautions that the concrete implementation requires domain-specific review by qualified personnel.