Decides whether a DPIA is required, then runs it end to end. EDPB nine-criteria threshold, multi-jurisdictional SA blacklists, two-track risk model, calibrated 5×5 scoring, Art. 36 consultation check, and EDPB template generation — one guided workflow under Art. 35.
Each capability is documented separately, tested separately, and called by the workflow at the right moment. Compose them or invoke individually.
Twelve structured steps. The human stays accountable; the skill carries the structure, the citations, and the document trail.
Match the workflow to the situation. The skill router picks automatically; you can override.
Single-folder skill. SKILL.md is the runtime spec; references hold the knowledge corpus; evals hold the proof.
dpia-sentinel/ ├── evals # Test cases + assertions │ └── evals.json # 8 cases, 72 assertions ├── references # Reference corpus │ ├── jurisdictions │ │ ├── be-apd.md │ │ ├── de-dsk.md │ │ ├── fr-cnil.md │ │ ├── ie-dpc.md │ │ ├── it-garante.md │ │ ├── nl-ap.md │ │ ├── pl-uodo.md │ │ └── whitelists.md │ ├── dpia-custom-population.md │ ├── dpia-custom-template-v1.docx │ ├── edpb-2026-custom-template-v1.docx │ ├── edpb-2026-explainer.md │ ├── edpb-2026-population.md │ ├── edpb-2026-template-v1.docx │ ├── edpb-2026-template.md │ ├── edpb-criteria.md │ ├── risk-catalog.md │ ├── scoring.md │ ├── sources.md │ └── templates.md ├── CHANGELOG.md # Version history ├── README.md # Deployment guide └── SKILL.md # Main skill instructions
Two deployment surfaces. The skill auto-triggers on relevant keywords once installed.
dpia-sentinel/ foldercp -r dpia-sentinel/ \ ~/.claude/skills/user/
Every output is documented, version-pinned, and traceable to its source citation.
Every release runs against a fixed test suite. Assertions check numeric consistency, citation accuracy, and decision-tree branches.
Every legal verdict resolves to one of these instruments. No invented articles, no synthetic recitals.
The trace is the product. Nothing happens off the record — no hidden tool calls, no silent retrieval, no opaque chain-of-thought.
GDPR Data Protection Impact Assessment Sentinel — a structured DPIA guidance skill for Claude that provides:
dpia-skill/
├── SKILL.md # Main skill instructions (deploy this)
├── CHANGELOG.md # Version history
└── references/
├── edpb-criteria.md # EDPB nine criteria + multi-jurisdictional framework
├── edpb-2026-template.md # EDPB 2026 DPIA template field-by-field spec
├── edpb-2026-template-v1.docx # Official EDPB template .docx (populatable)
├── edpb-2026-population.md # Table-by-table population guide for the template
├── edpb-2026-explainer.md # EDPB 2026 methodology reference
├── dpia-custom-template-v1.docx # Custom 12-section DPIA template .docx (populatable)
├── dpia-custom-population.md # Population guide for custom template
├── scoring.md # 5×5 risk scoring + modulating factors + two tracks
├── risk-catalog.md # Common DPIA risks by processing type (Track A+B)
├── templates.md # Document templates (5 formats)
├── sources.md # Regulatory source references
└── jurisdictions/
├── de-dsk.md # Germany — DSK blacklist
├── fr-cnil.md # France — CNIL blacklist
├── ie-dpc.md # Ireland — DPC blacklist
├── be-apd.md # Belgium — APD blacklist
├── nl-ap.md # Netherlands — AP blacklist
├── it-garante.md # Italy — Garante blacklist
├── pl-uodo.md # Poland — UODO blacklist
└── whitelists.md # FR, CZ, ES, AT whitelist exemptions
dpia-skill/ folder structuredpia-skill/ folder to your skills directory:
bash
cp -r dpia-skill/ /path/to/your/skills/user/dpia-skill/Just describe your processing activity:
"We're planning to deploy an AI system that scores job applicants based on their CVs and video interviews. The system will be used across Germany, France, and the Netherlands. Do we need a DPIA?"
The skill will activate and guide you through the assessment.
| Phase | Description |
|---|---|
| Threshold | Art. 35(3) triggers + nine-criteria analysis + national blacklist checks |
| Description | Systematic processing description per Art. 35(7)(a) |
| Asset Inventory | Risk-relevant assets grouped by type (EDPB 2026, Section 1.3) |
| Necessity | Effectiveness + least-intrusive test (upstream gate) |
| Proportionality | Benefits vs. impact balancing (upstream gate) |
| Inherent Risks | Track A (by-design) + Track B (operational), 5×5 matrix + modulating factors |
| Mitigation | Technical, organizational, and legal measures with implementation status |
| Residual Risk | Overall verdict: APPROVED / CONDITIONALLY APPROVED / CONSULT SA / REJECTED |
| Documentation | Audit-ready .docx generation (EDPB 2026 or custom format) |
| Template | Description |
|---|---|
| EDPB 2026 DPIA Report | Official harmonised format (Sections 0–6, recognized by all EU SAs) |
| Full DPIA Report (custom) | Custom 12-section assessment with threshold analysis + annexes |
| Threshold Justification Memo | 2-3 page document explaining why a DPIA is NOT required |
| Executive Summary | 1-2 page board/leadership summary |
| Art. 36 Consultation Package | Submission package for SA prior consultation |
| Document | Reference |
|---|---|
| GDPR Article 35 | DPIA obligation |
| GDPR Article 36 | Prior consultation |
| EDPB DPIA Template v1.0 (March 2026) | Harmonised EU-wide DPIA structure |
| EDPB Guidelines WP 248 rev.01 | DPIA methodology and nine criteria |
| EDPB Opinion 28/2024 | DPIA for AI processing |
| EDPB Guidelines 01/2025 | Pseudonymisation as risk reducer |
| National SA Art. 35(4) lists | Mandatory DPIA blacklists (7 jurisdictions) |
See CHANGELOG.md for full version history.
This skill provides structured guidance based on publicly available GDPR regulatory materials. It does not constitute legal advice. All DPIA decisions should involve your DPO (Art. 35(2)) and qualified legal counsel.
Created by Oliver Schmidt-Prietz — OneZero Legal