# Changelog — ai-act-quick

All notable changes to this skill are documented here.

Format: `## [vX.Y] — YYYY-MM-DD`

---

## [v1.4] — 2026-07-07

Annex III Nr. 5 lettering correction (legal accuracy).

- `quick-decision-tree.md` sector shortcut table corrected per Commission Guidelines ¶279–¶336 (canonical: `../ai-act-high-risk/references/annex-iii-area-5-essential-services.md`): credit scoring → **5(b)** (was 5(a)), life/health insurance → **5(c)** (was 5(b)), fraud-detection carve-out → excluded from **5(b)** (was 5(a)), benefit eligibility → **5(a)** (was 5(c)).
- eval-0 already asserted the correct "Nr. 5(b)" for credit scoring — the reference file now agrees with the eval.

---

## [v1.3] — 2026-06-21

Art. 50 transparency effective-date correction (Digital Omnibus — provisionally agreed 7 May 2026 / EP plenary June 2026, pending OJ publication).

- `quick-decision-tree.md`: Art. 50 was shown as "OVERDUE since 2 Aug 2025" — corrected to **2 Aug 2026** (50(1)/(3)/(4) + new-system 50(2)), with the legacy generative-AI 50(2) marking grace to **2 Dec 2026**.
- `compliance-deadlines.md`: un-bundled Art. 50 from the 2 Dec 2027 Annex III high-risk phase — Art. 50 is not postponed by the Omnibus.
- Date now sourced from the canonical `../../ai-act-high-risk/references/ai-omnibus-timeline-postponements.md`. `Last verified` → June 2026.

---

## [v1.2] — 2026-06-11

Audience-clarity + active-confidence guidance from the LegalQuants QA review (PR #8). No change to the 2-batch intake, 6-gate sequence, or the preliminary output template.

- **"Who this is for" + escalation duty.** Names the operator (which may be a non-lawyer product owner / compliance manager / founder) and makes explicit that a non-lawyer must route the card to counsel rather than treat "Likely Minimal" as an all-clear.
- **Work shape stated.** Names the work as bounded-transactional triage that opens an accretive full assessment, anchoring the speed framing to a delegation posture.
- **Confidence bands now drive behavior.** The existing High / Medium / Low field is wired to three distinct behaviors — High = proceed, Medium = surface the assumption and ask, Low = lead with the uncertainty and hand back — so an ambiguous determination no longer reads as a uniform "Likely". Wired into the Critical Reminders.

**Status:** reviewed (carried from v1.1) — additive guidance, no change to classification logic.

---

## [v1.1] — 2026-05-19

AI Omnibus timeline + routing handoff.

- Effective dates updated to Omnibus-postponed values.
- Triage workflow now routes the high-risk branch to the new ai-act-high-risk skill for depth assessment.

## [v1.0] — 2026-05-12

First **reviewed** release. Eval pass via `/skill-creator` confirmed v0.9 content quality with no structural fixes required.

- 3 test cases run with-skill vs. no-skill baseline: clear high-risk triage (German bank deploying US credit-scoring AI), clear minimal-risk triage (room-booking suggestion engine), ambiguous limited-risk case (customer-service chatbot)
- Result: 100% (23/23) pass rate, tied with baseline
- All three triage decisions correct: high-risk → Annex III Nr. 5(b) + FRIA mandatory + EU rep required; minimal-risk → Art. 4 AI literacy only; limited-risk → Art. 50(1) chatbot transparency + Anthropic upstream GPAI provider
- Known subagent-synthesis quirk (eval-2): with-skill output stated "Art. 50(1) since Aug 2025" but Art. 50 (Chapter IV) actually applies from 2 Aug 2026 per Art. 113(c). Skill's `compliance-deadlines.md` reference has the correct date — the error originated in subagent synthesis, not skill content. Candidate v1.1 polish: add an explicit "common pitfall" note to make the easily-confused date more emphatic in the SKILL.md.
- See `../../ai-act-quick-workspace/iteration-1/` for full eval artifacts

## [v0.9] — 2026-05-08

Initial import from CLAUDE_SKILL_EU AI Act/ai-act-quick/ (worked on 2026-03-15). Status was **audited** as of 2026-03-15 (see `docs/ai-act-suite/2026-03-15-audit.md` — scored 4.4/5 across 12 dimensions).

- Fast 15-25 minute triage skill for preliminary AI Act classification and compliance assessment
- Triggers on "quick check", "preliminary assessment", "Schnellprüfung", "Ersteinschätzung"
- Reference files: quick-decision-tree.md, jurisdiction-flags.md, compliance-deadlines.md
- Companion to the full ai-act-classifier / ai-act-roles / ai-act-obligations / ai-act-report suite