agentleFS
waldo-van-der-codewaldo-van-der-code/deutsches-recht-mit-claude/skills/gdpr-dsgvo/SKILL.md

legal-audit

Skill45 starsChanged 27 days agoMITView on GitHub

This skill should be used when the user asks to "check privacy compliance", "audit GDPR", "is this legal", "do we need a DPA", "check our privacy policy", "DSGVO check", "is our site compliant", "update the privacy policy", "cookie consent legal", "Datenschutz", "Impressum", "TTDSG", "data protection", "review our terms", "legal compliance", "privacy audit", or asks any question about GDPR, DSGVO, TTDSG, BDSG, data processor agreements, cookie law, or German data protection law.

1 other project uses a version of this.

Adopted by

Most recently taken up by

  • Other versions1 project uses a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 1

    1. idimsh/tdds-business-skills★ 0same name

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • In this repository3 other files it gives its agents.

The file

We split this skill into the rules it gives an agent

8 rules. About 0 are common, 5 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Check MCP availability: attempt semantische_rechtssuche with query DSGVO Artikel 6.

    line 27
  2. Very specific:

    Run these grep patterns (load references/audit-patterns.md for the full list):

    line 40
  3. Distinctive:

    No retention periods for booking/invoice data → 🟡 Mittel — Art.

    line 68
  4. Very specific:

    Use WebFetch on the target URL.

    line 111
  5. Distinctive:

    Every draft must include:

    line 186
  6. Distinctive:

    Run all four modes sequentially:

    line 196
  7. Distinctive:

    Citations: every finding cites a DSGVO article, TDDDG section, or BDSG section — no unanchored claims

    line 208
  8. Distinctive:

    Law source note: if MCP used, state which queries ran.

    line 209

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

/legal-audit — German DSGVO Legal Compliance

German-jurisdiction legal compliance for websites and apps. Covers DSGVO, TDDDG, BDSG, DDG, and relevant German case law. Grounds all findings in actual statute via the rechtsinformationen-bund-de MCP when available (see README for setup).

Default jurisdiction: Germany. Supervisory authority defaults to the authority of your Bundesland — see references/law-reference.md for a complete list. Also covers Austria (DSG) and Switzerland (nDSG) where noted.


Step 0 — Route

Parse the invocation args or user intent:

PatternMode
audit <path> or a file/dir path given→ Code Audit
scan <url> or arg starts with http→ URL Scan
flow <path>→ Data Flow Review
draft <type>→ Document Draft
No args or natural-language legal question→ Full Suite (Audit + Scan + Flow → Draft offer)

Check MCP availability: attempt semantische_rechtssuche with query DSGVO Artikel 6. If it responds, use it for every law lookup. If unavailable, fall back to built-in knowledge — note which source was used in each finding.

Display:

legal-audit | Mode: [X] | Target: [Y] | Law source: [MCP / built-in knowledge]


Mode A — Code Audit

Scan the codebase for data collection points and classify each finding by severity.

A1 — Discover

Run these grep patterns (load references/audit-patterns.md for the full list):

# Third-party fetches
grep -r "fetch(\|axios\.\|import.*resend\|import.*nodemailer\|gtag(\|posthog\|supabase\|stripe\|loadStripe" --include="*.ts" --include="*.tsx" --include="*.js" .

# Storage writes
grep -r "document\.cookie\|localStorage\.setItem\|sessionStorage\.setItem\|setCookie\|set-cookie" --include="*.ts" --include="*.tsx" .

# PII field names
grep -rEi "email|phone|iban|firstName|lastName|birthDate|passport|ipAddress" --include="*.ts" --include="*.tsx" .

# Consent gating
grep -r "consent\|cookie_consent\|analytics_storage\|posthog\.init\|opt_out" --include="*.ts" --include="*.tsx" .

A2 — Classify

For each finding, assign severity (🔴 Hoch / 🟡 Mittel / 🟢 Niedrig) and cite the relevant article.

Load references/law-reference.md for the full article-to-violation mapping table. Key rules:

  • Third-party script fires before consent check → 🔴 Hoch — TDDDG § 25(1), DSGVO Art. 6(1)(a)
  • Google Consent Mode "advanced" (gtag.js loads before consent and sends cookieless pings, gcs=G100 in /g/collect) → 🔴 Hoch — IP/device data reach Google before any choice; privacy texts usually claim "only after consent". Fix: inject gtag.js only after Accept ("basic" mode); set ad_user_data/ad_personalization too
  • Third-party iframes (Google Maps, YouTube, Google Drive video previews, social embeds) loading on page view → 🔴 Hoch — click-to-load gate; check every page, not just the one the privacy text mentions
  • Operator alert channels (Telegram/Slack/Discord bots, error trackers) receiving customer names/emails → 🔴 Hoch — undisclosed recipient, usually no Art. 28 agreement; send IDs/booking refs only, redact centrally
  • Controller not named with full postal address in the privacy policy → 🔴 Hoch — DSGVO Art. 13(1)(a)
  • Hosting provider (Vercel, Netlify, …) not disclosed, US vendors without DPF/SCC basis, EU entity misnamed (e.g. Stripe Payments Europe Ltd., not Stripe Inc.) → 🟡 Mittel — Art. 13(1)(e)/(f)
  • No retention periods for booking/invoice data → 🟡 Mittel — Art. 13(2)(a); § 147 AO / § 14b UStG (8 years for Buchungsbelege since BEG IV, 10 for books)
  • Privacy section exists in one language only on a multilingual site → 🟡 Mittel — Art. 12(1) (transparent, intelligible)
  • PII collected in form with no Art. 13 disclosure → 🔴 Hoch — DSGVO Art. 13
  • Email send handler, processor not in privacy policy → 🔴 Hoch — DSGVO Art. 13, Art. 28
  • US processor used, no SCC documented → 🟡 Mittel — DSGVO Art. 46(2)(c)
  • Google Fonts loaded from CDN → 🟡 Mittel — LG München I 2022
  • localStorage for consent preference only → 🟢 Niedrig — TDDDG § 25(2) exemption
  • Analytics gated behind consent → 🟢 Niedrig — compliant

Also check:

  • Impressum: DDG § 5 — name, address, contact, VAT ID required on every commercial site. "§ 5 TMG" / "§ 55 RStV" wording is outdated → 🟡
  • DPO threshold: BDSG § 38 — mandatory if 20+ employees regularly process personal data

A3 — Law lookup

For each 🔴 finding, use semantische_rechtssuche to pull the exact statutory language. Quote the article verbatim in the finding output.

A4 — Output format

## Code Audit — [project]  |  [date]  |  Law: [MCP/built-in]

🔴 Hoch — [N]
  [file:line] [issue] — [Article] DSGVO / § [X] TDDDG
  Statutory text: "[quoted text from MCP]"
  Fix: [one-line concrete action]

🟡 Mittel — [N]  ...

🟢 Niedrig / Compliant — [N]  ...

Impressum: [found / missing / partial — list gaps]
DPO threshold: [below / above / unclear]

Summary: 🔴 [N] | 🟡 [N] | 🟢 [N] | Remediation effort: [low/medium/high]

Mode B — URL Scan

Audit the live site as a user agent sees it.

Use WebFetch on the target URL. Extract and analyze:

  1. All <script src="..."> and <link href="..."> tags — classify first-party vs. third-party domains
  2. Cookie consent banner presence — does it block scripts before acceptance?
  3. <form> elements with PII inputs
  4. Presence and reachability of /impressum, /privacy (or /datenschutz), /robots.txt
  5. Privacy page content — does it name every third-party domain found in step 1?
  6. Real-browser network check (required — WebFetch cannot see it): load the key pages in Playwright without any consent decision, scroll to the bottom (lazy iframes), and record every request to third-party hosts. Then click Accept, then reopen the settings and Decline, and list what is left: cookies on the host and parent domain (_ga is often set on .example.de, shared with sibling subdomains), ph_*/_gcl* in local/session storage, and any analytics request sent after Decline (autocapture records the Decline click itself and flushes it seconds later).

Load references/audit-patterns.md for the third-party domain classification table and consent-banner correctness checklist.

Gap check: detected third-party domains vs. domains disclosed in the privacy page. Any detected-but-undisclosed domain is 🔴 Hoch — DSGVO Art. 13.

Output format:

## URL Scan — [url]  |  [datetime]

Third-party scripts: [table: domain | service | data sent | legal basis | DPA required]
Consent banner: [present/absent | scripts blocked before consent: yes/no | reject-all prominent: yes/no]
Required pages: Impressum [✓/✗] | Privacy [✓/✗] | robots.txt [✓/✗]
Gap — detected but not disclosed: [list — 🔴 for each]
Verdict: [compliant / partially compliant / non-compliant] — [1-paragraph assessment]

Mode C — Data Flow Review

Map all PII flows through the system. Produces the Art. 30 VVT (Verzeichnis von Verarbeitungstätigkeiten).

C1 — Discover flows

Scan for:

  • Ingress: form handlers, booking creation, auth flows, webhooks
  • Processing: DB writes, email sends, third-party API forwarding, logging with PII, analytics events with PII params
  • Egress: external API calls with PII in body, outbound emails, blob/CDN storage

C2 — Build VVT table

Load references/data-flow.md for the full VVT template, DPA status checklist, and cross-border transfer matrix.

For each processing activity, document:

  • Processing activity name
  • PII categories involved
  • Legal basis (Art. 6 DSGVO — consent / contract / legitimate interest)
  • Processor name + DPA URL
  • Transfer outside EU/EEA? → SCC or adequacy decision
  • Retention period

C3 — DPA checklist

Cross-reference each detected processor against the known DPA URLs in references/data-flow.md. Flag processors with no signed DPA as 🔴 Hoch.


Mode D — Document Draft

Generate ready-to-paste German-law DSGVO document sections.

Load references/document-templates.md for templates and required disclosure elements per document type.

Supported types:

TypeGenerates
privacy-section <processor>Single Art. 13 processor disclosure block (DE + EN)
privacy-pageFull Datenschutzerklärung skeleton
impressumImpressum block per DDG § 5
cookie-bannerConsent banner text (DE + EN), TDDDG § 25 compliant
vvtBlank Art. 30 VVT template
art13-noticeArt. 13 notice for a specific form or data collection point
dpa-checklistProcessor DPA to-do list with links

For each draft, pull required disclosure elements via semantische_rechtssuche (query: Datenschutzerklärung Pflichtangaben Artikel 13). Verify every required element is present before outputting.

Every draft must include:

  1. German version (primary)
  2. English version (secondary, marked as translation)
  3. Inline comments citing the DSGVO article that requires each element
  4. Disclaimer: Dieser Text wurde mit KI-Unterstützung erstellt und ersetzt keine Rechtsberatung.

Mode E — Full Suite

Run all four modes sequentially:

  1. Code Audit (Mode A)
  2. URL Scan (Mode B)
  3. Data Flow Review (Mode C)
  4. Consolidated Gap Report — all 🔴 findings ranked by remediation effort (quick wins first)
  5. Offer: "Draft the privacy policy sections that fix the 🔴 gaps?"

Output Standards

  • Severity: 🔴 Hoch (likely DPA interest, fix immediately) / 🟡 Mittel (fix before next release) / 🟢 Niedrig or ✅ Compliant
  • Citations: every finding cites a DSGVO article, TDDDG section, or BDSG section — no unanchored claims
  • Law source note: if MCP used, state which queries ran; if fallback, note (Gesetzestext aus Trainingsdaten — verifizieren auf gesetze-im-internet.de)
  • Disclaimer: append to every output — Dieser Bericht dient der technischen Orientierung und ersetzt keine Rechtsberatung.

Additional Resources

Load these reference files when needed — do not load all at once:

  • references/law-reference.md — DSGVO/TDDDG/BDSG article table, violation-to-article mapping, key German case law (LG München I 2022, Planet49, Schrems II), supervisory authorities per Bundesland
  • references/audit-patterns.md — grep patterns for code audit, third-party domain classification table, consent banner correctness checklist, framework-specific patterns (Next.js, Astro, React)
  • references/data-flow.md — Art. 30 VVT template, known processor DPA URLs, cross-border transfer matrix
  • references/document-templates.md — required disclosure elements per document type, Art. 13 checklist, Impressum checklist, cookie banner template, data subject rights section

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.