LIVE · v1.1 STABLE
BUILD 2026.05.27
META-SKILL
AGPL-3.0 · 11 EVAL CASES · 0 ASSERTIONS

PROMPTS
FORGED
for LEGAL
ANALYSIS.

Turns any EU digital regulation document into a tailored expert prompt for structured legal analysis. Document characterisation, outcome + audience elicitation, calibrated prompt assembly, optional in-session execution, automatic freshness check, and a plain-English explainer — a meta-skill measured at +35.5pp over the no-skill baseline.

GDPR   ·   EU AI ACT   ·   DATA ACT   ·   DGA   ·   DSA   ·   DMA   ·   NIS2 · BSIG-NEU   ·   EPRIVACY   ·   CRA   ·   DORA   ·   EIDAS 2.0   ·   PLD DIR. 2024/2853   ·   EUR-LEX   ·   CURIA   ·   EDPB   ·   GDPR   ·   EU AI ACT   ·   DATA ACT   ·   DGA   ·   DSA   ·   DMA   ·   NIS2 · BSIG-NEU   ·   EPRIVACY   ·   CRA   ·   DORA   ·   EIDAS 2.0   ·   PLD DIR. 2024/2853   ·   EUR-LEX   ·   CURIA   ·   EDPB
§ 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
DOCUMENT CHARACTERISATION
One-screen profile: instrument type, binding force, issuer and legal basis, status, affected provisions, sectoral interfaces, temporal application. Flags 3–5 contestable interpretive moves with paragraph references on first read.
02
OUTCOME ELICITATION
At most three questions — outcome type, audience, stance. Ten deliverable types: consultation response, compliance memo, client memo, public commentary, talk prep, risk assessment, litigation input, comparative analysis, horizon-scan, skill input.
03
PROMPT ASSEMBLY
Builds a nine-block prompt: role, context, task, analytical framework, provisions to scrutinise, citation conventions, register constraints, output structure, self-check protocol. Saved to disk and printed inline.
04
FRESHNESS CHECK
Automatic for drafts, proposals, and consultation versions. Verifies against the issuer's authoritative source for a later or finalised text; surfaces the version choice before proceeding. Never silently substitutes documents.
05
LIVE RESEARCH
Opt-in verification on primary EU and national sources only — EUR-Lex, CURIA, EDPB, Commission domains, ENISA, BEREC, national DPAs and gazettes. Marks live-verified citations; flags gaps rather than guessing. Wikipedia, social media, and news excluded.
06
PLAIN-ENGLISH EXPLAINER
Produced alongside every executed analysis. Hard 150–300 word cap. Tells the practitioner what the document is, what the analysis found, and what remains uncertain — without duplicating the formal deliverable.
§ 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
Ingest
Read the document from PDF, pasted text, URL, or reference-by-name. For EUR-Lex prefer CELEX-linked HTML; for a named document, confirm title, date, CELEX or ECLI first.
02
Freshness check
Automatic for drafts, Commission proposals, and consultation versions. Search the issuer's source for a later version; if one exists, ask which to use before proceeding.
03
Characterise
One-screen profile: instrument type, binding force, issuer, legal basis, status, subject matter, sectoral interfaces, temporal application, plain-language summary. Confirm with the user.
04
Spot contestable moves
Flag 3–5 contestable interpretive moves with paragraph references — the provisions the analysis will scrutinise.
05
Elicit
At most three questions: outcome type (mandatory), audience, stance. Skip any already answered in context. Optional follow-ups on language, length, comparative dimension, register.
06
Generate prompt
Assemble the nine-block prompt from outcome_templates.md and analytical_canon.md. Save to ./[doc_slug]_prompt_[outcome_slug].md and output inline.
07
Execute (optional)
Run the prompt in-session to produce the deliverable. Apply the 12-point self-check before delivery. Save to ./[doc_slug]_analysis_[outcome_slug].md.
08
Plain-English explainer
Always produced alongside an executed analysis. 150–300 words, hard cap. User decides integration: executive summary, sidebar, annex, cover note, or standalone.
09
Refine (optional)
Handle re-audience, re-outcome, translate, tighten, expand, add comparison, or different stance — without restarting. Refinements append _v2, _v3; prior versions kept.
10
Route downstream
If the characterisation surfaces an operational task, suggest the relevant sibling skill (LIA, TIA, DPA, RoPA, AI Act suite, NIS2, breach, Data Act) — after the analysis is delivered, not before.
§ 03 · MODES

THREE PATHS.
One OUTCOME.

Match the workflow to the situation. The skill router picks automatically; you can override.

PROMPT ONLY
Stop at Step 4. Characterise the document and hand back a calibrated expert prompt for the user to run elsewhere.
EXECUTE
Run the generated prompt in-session, apply the 12-point self-check, and deliver the formal analysis plus the plain-English explainer.
REFINE
Iterate an existing deliverable: re-audience, re-outcome, translate, tighten, expand, or change stance — without restarting the workflow.
§ 04 · ANATOMY

WHAT'S
IN the SKILL.

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

legal-analysis-forge/
├── evals  # Test cases + assertions
│   └── evals.json  # 11 cases, 0 assertions
├── references  # Reference corpus
│   ├── analytical_canon.md
│   ├── document_taxonomy.md
│   └── outcome_templates.md
├── CHANGELOG.md  # Version history
├── README.md  # Deployment guide
└── SKILL.md  # Main skill instructions
§ 05 · 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 legal-analysis-forge/ folder
  3. Skill auto-triggers on relevant keywords

CLAUDE CODE / MCP

  1. Copy folder to your skills directory:
cp -r legal-analysis-forge/ \
   ~/.claude/skills/user/
§ 06 · OUTPUTS

WHAT YOU
get BACK.

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

Characterisation
Structured one-screen profile of the document — instrument type, binding force, issuer, legal basis, status, affected provisions, contestable moves, temporal application, plain-language summary.
Tailored prompt (.md)
Nine-block expert prompt assembled to outcome, audience, and stance. Saved as ./[doc_slug]_prompt_[outcome_slug].md and printed inline for direct copy.
Executed analysis (.md)
Optional in-session deliverable produced from the prompt — consultation response, memo, commentary, comparative analysis, or horizon-scan entry — after the self-check passes.
Plain-English explainer (.md)
150–300 word orientation for the practitioner: what the document is, what the analysis found, what remains uncertain. Integration into the deliverable is the user's choice.
Live-verified citations
Where opt-in research runs, citations marked [live-verified, accessed YYYY-MM-DD] against primary EU and national sources; unverifiable points flagged, not guessed.
§ 07 · EVALS

TESTED
before SHIPPED.

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

11
Test Cases
0
Assertions
100%
Coverage Required
00
Hi, I want to submit a stakeholder consultation response to the European Commiss...
0 ASSERTS
01
I need an internal compliance memo on a recent CJEU judgment under the GDPR — Ca...
0 ASSERTS
02
Ich möchte einen LinkedIn-Beitrag auf Deutsch zu den neuen EDPB-Leitlinien 02/20...
0 ASSERTS
03
I'm preparing a thought-leadership briefing for our partners on the overlap betw...
0 ASSERTS
04
Our client (a multinational SaaS provider with EU operations) has been contacted...
0 ASSERTS
05
Add a horizon-scan entry for our regulatory tracker
0 ASSERTS
06
I'm drafting litigation brief input for a national court challenge to a third-co...
0 ASSERTS
07
Ich halte einen Vortrag bei der Datenschutzkonferenz im Herbst 2026 zum Stichtag...
0 ASSERTS
§ 08 · REGULATORY BASIS

WHAT IT
cites.

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

GDPR (Reg. 2016/679)
Data protection — and the EDPB guidance and CJEU case law interpreting it.
EU AI Act (Reg. 2024/1689)
AI regulation, including harmonised standards under Art. 40 and codes under Art. 56.
NIS2 (Dir. 2022/2555)
Cybersecurity, including Member State implementations such as BSIG-neu.
Data Act (Reg. 2023/2854)
Data access and sharing across the EU digital stack.
DSA / DMA (Reg. 2022/2065 · 2022/1925)
Platform and gatekeeper regulation, including codes under Art. 45 DSA.
CRA (Reg. 2024/2847)
Cyber Resilience Act — horizontal cybersecurity for products with digital elements.
§ 09 · 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 prompts.
Every characterisation field and scrutiny point ties to a cited provision, paragraph, or authority — never an unsourced assertion.
VERIFIED
Reproducible deliverables.
Stable doc slugs and versioned refinements (_v2, _v3) track one document through every analysis; prior versions are never overwritten.
IMMUTABLE
†
Self-check before delivery.
The 12-point analytical-canon protocol and a hard 150–300 word explainer cap run before any executed analysis ships.
ENFORCED
**
EU-native.
Built on the EU digital regulation stack and primary EU and national sources — EUR-Lex, CURIA, EDPB, national DPAs and gazettes. Not retrofitted.
NATIVE
§ 10 · README.MD

legal-analysis-forge

A prompt forge for structured legal analysis of EU digital regulation documents. Given a Regulation, Directive, Commission Guidelines, EDPB Opinion, CJEU judgment, AG Opinion, national DPA decision, code of conduct, harmonised standard, or draft consultation document, the skill characterises the document, elicits the desired deliverable, and produces a tailored expert prompt. The prompt can be executed in-session to generate the deliverable, and is always accompanied by a plain-English explainer for the practitioner using the skill.

For the full six-step workflow, see SKILL.md. For version history, see CHANGELOG.md.

What this skill does

Given an EU digital regulation document and a deliverable type, the skill:

  • Characterises the document — instrument type, binding force, issuer and legal basis, status, subject matter, sectoral interfaces, contestable interpretive moves, temporal application, plain-language summary.
  • Generates a tailored expert prompt assembled from a role marker, document context, task description, analytical framework (general legal interpretation rules plus document-specific scrutiny points), citation conventions, register constraints, output structure, and a self-check protocol.
  • Optionally executes the prompt in-session to produce the deliverable directly, applying the self-check before delivery.
  • Always produces a plain-English explainer alongside the formal output; the user chooses whether to integrate it into the deliverable.

Deliverable types supported: stakeholder consultation response, internal compliance memo, external client memo, public commentary (LinkedIn / blog / newsletter), conference talk preparation, internal risk assessment, litigation brief input, comparative analysis, horizon-scan entry, skill input.

Scope

EU digital regulation only: GDPR, AI Act, Data Act, DGA, DSA, DMA, NIS2 (incl. BSIG-neu), ePrivacy, CRA, DORA, eIDAS 2.0, PLD (Dir. 2024/2853), AI Liability Directive, and adjacent secondary instruments (delegated acts, implementing acts, harmonised standards, codes of conduct).

Out of scope: competition, IP, tax, employment, sectoral law not touching the digital stack. For operational compliance (DPIA, DPA drafting, AI Act obligations mapping, RoPA, breach response, NIS2 scoping), the skill routes to the relevant compliance skill in the portfolio rather than absorbing the task.

Deployment

Place this skill in your Claude Code skill directory (~/.claude/skills/ or workspace-level). Claude routes to it automatically when a user provides an EU digital regulation document and asks for structured analysis, a consultation response, a memo, a briefing, or comparable deliverable — in English or German. The skill does not require any external tool configuration beyond WebFetch (for live research) and pdf-processing-anthropic (for PDF ingestion); both are standard Claude Code capabilities.

Outputs are written to the user's current working directory by default.

Disclaimer

This skill is not legal advice. It enforces interpretive discipline, citation hygiene, and an anti-hallucination protocol, but it does not guarantee accuracy. EU instruments evolve and only the CJEU can give authoritative interpretation. Use under qualified counsel.

License

AGPL-3.0 (see repo-root LICENSE file).