LIVE · v1.4 STABLE
BUILD 2026.05.27
15 MIN TRIAGE
AGPL-3.0 · 3 EVAL CASES · 0 ASSERTIONS

AI ACT
TRIAGE
in FIFTEEN
MINUTES.

Fast preliminary EU AI Act classification before you commit to full analysis. Adaptive 2-batch intake, 6-step gate sequence, confidence-tagged verdict, tier-specific deadlines, jurisdiction flags, and automatic routing to the depth skills — in a single triage pass.

REG. (EU) 2024/1689   ·   AI ACT ART. 2   ·   AI ACT ART. 3(1)   ·   AI ACT ART. 5   ·   ANNEX I   ·   ANNEX III   ·   AI ACT ART. 6(3)   ·   AI ACT ART. 50   ·   AI ACT ART. 51   ·   AI ACT ART. 53   ·   GPAI   ·   AI OMNIBUS 2026   ·   REG. (EU) 2024/1689   ·   AI ACT ART. 2   ·   AI ACT ART. 3(1)   ·   AI ACT ART. 5   ·   ANNEX I   ·   ANNEX III   ·   AI ACT ART. 6(3)   ·   AI ACT ART. 50   ·   AI ACT ART. 51   ·   AI ACT ART. 53   ·   GPAI   ·   AI OMNIBUS 2026
§ 01 · CAPABILITIES

SIX PRIMITIVES.
One SKILL.

Each capability is documented separately, tested separately, and called by the workflow at the right moment. Compose them or invoke individually.

01
ADAPTIVE INTAKE
Conversational 2-batch flow capped at two turns — one if you answer in detail. Silently extracts the 8 required fields, asks follow-ups only for genuine gaps.
02
6-STEP GATE SEQUENCE
Scope (Art. 2) → AI-system test (Art. 3(1)) → prohibited screen (Art. 5) → high-risk (Annex I + III) → GPAI → transparency (Art. 50). Run internally, output as one verdict.
03
CONFIDENCE TAGGING
Every classification carries a HIGH / MEDIUM / LOW confidence indicator. Low confidence is an explicit instruction to escalate before deciding.
04
DEADLINE + EXPOSURE
Tier-specific compliance dates on the AI Omnibus 2026 timeline (Annex III: 2 Dec 2027; Annex I: 2 Aug 2028) plus preliminary Art. 99 penalty band.
05
JURISDICTION FLAGS
Member-State-specific signals — works-council co-determination (BetrVG §87), CSE consultation, sector regulators — surfaced from the deployment country.
06
SUITE HANDOFF
Routes the high-risk branch to ai-act-high-risk and emits a paste-ready context block for the classifier / roles / obligations / report skills.
§ 02 · WORKFLOW

FROM INPUT
to ARTEFACT.

Ten structured steps. The human stays accountable; the skill carries the structure, the citations, and the document trail.

01
Disclaimer & web check
Non-blocking preliminary disclaimer. Search for current AI Act enforcement and Commission-guidelines status on activation.
02
Batch 1 intake
Three essential questions in plain language: what the system does, where it is deployed, your organization's relationship to it.
03
Coverage analysis
Silently check the answer against the 8 required fields. Be generous extracting — sector and affected persons are often inferable from the description.
04
Batch 2 follow-up
Only if gaps remain. One conversational message covering just the missing or partial fields. Never re-ask what was already answered.
05
Normalization
Map all gathered context into the structured 8-field format so the gate sequence can reference fields consistently.
06
Gate sequence
Run the 6 gates internally: scope, AI-system test, prohibited screen, Annex I + III high-risk, GPAI, Art. 50 transparency.
07
Confidence call
Assign HIGH / MEDIUM / LOW to the verdict. Err toward the higher risk tier when uncertain.
08
Preliminary output
Single consolidated assessment: classification, role, top obligations, deadline, jurisdiction flags, financial exposure, warnings.
09
Suite routing
Route the high-risk branch to ai-act-high-risk; recommend classifier / roles / obligations / report with per-skill priority.
10
Template offer
Optionally produce PRELIMINARY-marked drafts: classification record, compliance register entry, management briefing.
§ 03 · ANATOMY

WHAT'S
IN the SKILL.

Single-folder skill. SKILL.md is the runtime spec; references hold the knowledge corpus; evals hold the proof. 

ai-act-quick/
├── evals  # Test cases + assertions
│   └── evals.json  # 3 cases, 0 assertions
├── references  # Reference corpus
│   ├── compliance-deadlines.md
│   ├── jurisdiction-flags.md
│   └── quick-decision-tree.md
├── CHANGELOG.md  # Version history
├── README.md  # Deployment guide
└── SKILL.md  # Main skill instructions
§ 04 · DEPLOYMENT

INSTALL
and INVOKE.

Two deployment surfaces. The skill auto-triggers on relevant keywords once installed.

CLAUDE.AI USER SKILLS

  1. Settings → Profile → Custom Skills
  2. Upload the entire ai-act-quick/ folder
  3. Skill auto-triggers on relevant keywords

CLAUDE CODE / MCP

  1. Copy folder to your skills directory:
cp -r ai-act-quick/ \
   ~/.claude/skills/user/
§ 05 · OUTPUTS

WHAT YOU
get BACK.

Every output is documented, version-pinned, and traceable to its source citation.

Preliminary assessment
Single consolidated verdict: AI-system test, scope, risk tier, classification basis, and a HIGH / MEDIUM / LOW confidence rating.
Role assessment
Likely role — provider, deployer, quasi-provider, importer, distributor — with quasi-provider trigger flags (e.g. finetuning may trigger Art. 25).
Top obligations table
Up to five high-risk / GPAI obligations with article, urgency, and effort estimate, plus Art. 4 AI competence for all tiers.
Compliance timeline
Applicable deadline on the AI Omnibus 2026 schedule with days remaining and urgency band.
Assessment context block
Paste-ready handoff summary that feeds directly into the next skill in the suite.
Preliminary templates
Optional PRELIMINARY-marked drafts: classification record (Prüfprotokoll), compliance register entry, management briefing (Entscheidungsvorlage).
§ 06 · EVALS

TESTED
before SHIPPED.

Every release runs against a fixed test suite. Assertions check numeric consistency, citation accuracy, and decision-tree branches.

03
Test Cases
0
Assertions
100%
Coverage Required
00
Quick AI Act check, please — we're considering deploying an off-the-shelf AI too...
0 ASSERTS
01
Quick AI Act triage please
0 ASSERTS
02
Quick AI Act check — we're rolling out a customer-service chatbot on our B2C e-c...
0 ASSERTS
§ 07 · REGULATORY BASIS

WHAT IT
cites.

Every legal verdict resolves to one of these instruments. No invented articles, no synthetic recitals.

EU AI Act (Reg. (EU) 2024/1689)
Governing regulation for the full triage.
Art. 2 · Art. 3(1) · Art. 5
Scope, AI-system definition, and prohibited-practice screen.
Annex I + Annex III · Art. 6(3)
High-risk category mapping and the Art. 6(3) exception check.
Art. 50 · Art. 51 / 53 / 55
Transparency triggers and GPAI thresholds and obligations.
AI Omnibus 2026 timeline
Postponed effective dates — Annex III: 2 Dec 2027; Annex I: 2 Aug 2028.
§ 08 · TRUST

EVERY STEP,
auditable.

The trace is the product. Nothing happens off the record — no hidden tool calls, no silent retrieval, no opaque chain-of-thought.

§
Source-anchored output.
Every gate, tier, and deadline traces to a cited article of Regulation (EU) 2024/1689 — not vibes.
VERIFIED
Reproducible triage.
Pin a build; recreate the preliminary verdict later for the audit trail or the management briefing.
IMMUTABLE
†
Confidence before commitment.
Every verdict is marked PRELIMINARY and confidence-tagged; LOW ratings force escalation to the depth skills.
ENFORCED
**
EU-native.
Built around Annex III, Art. 50, and Member-State flags like BetrVG §87 and CSE consultation. Not retrofitted.
NATIVE
§ 09 · README.MD

EU AI Act Quick Assessment — Deployment Guide

📄 View the interactive skill page →

See CHANGELOG.md for version history.

Overview

EU AI Act Quick Assessment — fast 15–25 minute triage for preliminary classification:

  • Adaptive 2-batch intake — minimal questions, system-description-informed
  • 6-step gate sequence — scope, AI-system test, prohibited, Annex I, Annex III, GPAI
  • Preliminary classification output with confidence level
  • Compliance deadlines for the identified tier
  • Jurisdiction flags highlighting Member-State-specific considerations
  • Template offer — option to escalate to full assessment (classifier / roles / obligations / report)
  • Clear scope boundary — designed for triage, not final determination

File Structure

ai-act-quick/
├── SKILL.md                              # Main skill instructions (deploy this)
├── CHANGELOG.md                          # Version history
├── evals/
│   └── evals.json                        # Test cases
└── references/
    ├── quick-decision-tree.md            # 6-step gate sequence
    ├── compliance-deadlines.md           # Tier-by-tier deadline lookup
    └── jurisdiction-flags.md             # Member State-specific flags

Deployment

Claude.ai (User Skills)

  1. Go to Settings → Profile → Custom Skills (or equivalent)
  2. Upload the entire ai-act-quick/ folder structure
  3. The skill will auto-trigger on "quick AI Act check", "preliminary assessment", "Schnellprüfung", or "Ersteinschätzung"

Claude Code / Custom MCP Setup

  1. Copy the ai-act-quick/ folder to your skills directory: bash cp -r ai-act-quick/ /path/to/your/skills/user/ai-act-quick/
  2. Ensure the skill is registered in your configuration

Usage

Quick Start

Dump what you know about the system:

"Quick AI Act check please — we're a SaaS in Berlin selling a meeting-summary tool that transcribes calls and produces action items using GPT-4. Customers are EU businesses. Does the AI Act apply, and what tier?"

The skill will run a 15–25 minute triage and return a preliminary verdict with confidence.

Trigger Phrases

  • "Quick AI Act assessment" / "Preliminary check" / "Does the AI Act apply?"
  • "Schnellprüfung" / "Ersteinschätzung" / "AI Act triage"
  • "Run a quick classification"

Workflow

Phase Description
Phase 1: Quick Context Adaptive 2-batch intake (system description + role/jurisdiction)
Phase 2: Rapid Classification 6-step gate sequence — scope, AI-system, prohibited, Annex I, Annex III, GPAI
Phase 3: Preliminary Output Tier verdict + confidence + jurisdiction flags + deadlines
Phase 4: Template Offer Optional handoff to full assessment skills (classifier / roles / obligations / report)

Capabilities Summary

Feature Description
Adaptive Intake Description-informed minimal-question flow (2 batches)
6-Step Gate Scope → AI-system → prohibited → Annex I → Annex III → GPAI
Confidence Tagging Each verdict carries a HIGH / MEDIUM / LOW confidence indicator
Deadline Lookup Tier-specific compliance dates
Jurisdiction Flags Member-State-specific signals (BSI / CNIL / Garante / etc.)
Suite Handoff Smooth escalation to full skills (classifier, roles, obligations, report)
Scope Discipline Outputs explicitly marked PRELIMINARY — not a substitute for full assessment

Regulatory Basis

Document Reference
EU AI Act Regulation (EU) 2024/1689
Art. 5 / Annex I / Annex III Risk-tier classification anchors
Art. 51 / 53 / 55 GPAI thresholds and obligations
Compliance deadlines Title XIII + Commission implementation timeline

License & Disclaimer

This is a preliminary AI Act assessment based on Regulation (EU) 2024/1689, designed for rapid triage. It is not legal advice and does not replace a full assessment — validate results using the detailed skills (ai-act-classifier, ai-act-roles, ai-act-obligations, ai-act-report) and qualified legal counsel.

Licensed under AGPL-3.0 — see LICENSE at the repo root.

Created by Oliver Schmidt-Prietz — OneZero Legal