- Home
- Skills
- Legal & Compliance
- Data Privacy and Governance Architect
Data Privacy and Governance Architect
Designs data privacy architectures: purpose limitation, consent propagation, pseudonymization, and deletion.
$9
Works with the AI tools you already use
Data Privacy and Governance Architect
Example session with this skill installed
Architect cross-system data privacy architecture, consent propagation, and automated DSAR rights execution for our Global Telehealth Platform.
- Subject: Global Telehealth Platform (run-id: telehealth-priv-001) serving 2.4M patients across EU and US regions.
- Deciders: Data Protection Officer (Elena Rostova), Principal Privacy Architect (Marcus Vance), and VP Engineering (Dr. Aris Thorne).
- Scope & Mandates: Governing 32 microservices, Snowflake warehouse, AWS Bedrock AI summarizer, and Stripe/Twilio processors under GDPR and HIPAA.
- Tension: Product proposed SHA-256 hashed patient IDs for analytics and retaining consult transcripts indefinitely for AI training, claiming hashing equals anonymization. Elena Rostova and Marcus Vance rejected this, citing incident PRIV-3109 where account deletion missed 14 S3 warehouse dumps and CRM logs. They mandate Privacy by Design: purpose-bound tokens, 24-hour DSAR deletion cascades across warehouses and vendors, 90-day pseudonymization key rotation, and tax-retention quarantine (7-year hold) isolated from clinical stores.
Write the privacy architecture under docs/architecture/privacy-architect/.
- Read your context and instructions
- Compiled the data privacy
Wrote docs/architecture/privacy-architect/00-overview/privacy-architect-overview.md. Complete enterprise privacy architecture specifying crypto-shredding key lifecycles, GDPR Article 17 erasure pipelines, distributed tombstone propagation, and consent enforcement.
---
skill: privacy-architect
---
# Privacy Architecture: Global Customer Privacy Platform [PRIV-FIN-001]
## Summary
This specification establishes the enterprise privacy engineering architecture, data subject rights automation, and crypto-shredding framework for the Global Customer Privacy Platform under run ID `fintech-privacy-arch-001`. It governs 12 million consumer profiles across 35 microservices in EU (`eu-central-1`) and US (`us-east-1`) jurisdictions. It decisively resolves the critical regulatory compliance failures demonstrated in regulatory enforcement action ENF-4902 (where incomplete soft-deletions on immutable logs incurred a €4.2 million GDPR penalty). The architecture enforces cryptographic erasure ("Crypto-Shredding") via dedicated per-user AWS KMS encryption keys, distributed tombstone event propagation across Kafka streams, automated Data Subject Access Request (DSAR) deletion pipelines fulfilling requests in < 7 business days (well within the statutory 30-day GDPR limit), and ingress data minimization gates.
## Detailed Description
Standard relational database deletions (`DELETE FROM users WHERE id = ...`) fail to fulfill modern privacy regulations because customer PII propagates downstream into immutable append-only Kafka event logs, Parquet analytical data lakes, and read-only backup snapshots. Rewriting petabyte-scale immutable backups to excise individual user records is computationally infeasible. Crypto-shredding encrypts each individual user's PII with a dedicated cryptographic key; destroying that key renders all downstream historical copies permanently indecipherable, achieving legally certified erasure.
Data Subject Erasure Request (GDPR Art. 17 / CCPA)
│
▼
[ Privacy Orchestrator: DSAR State Machine ]
├── 1. Verifies Data Subject Identity & Active Legal Holds
└── 2. Issues Cryptographic Erasure Command to Central Key Vault
│
┌──────────────────┴──────────────────┐
▼ ▼
[ Key Vault: AWS KMS / Vault ] [ Event Stream: Apache Kafka ]
└── Destroys Per-User Key key-usr-8821 └── Publishes Tombstone: users.tombstones
│ (Instant Cryptographic Erasure) └── Compacts Partition (Removes State)
│
├─► Aurora PostgreSQL (PII ciphertext is now unrecoverable white noise)
├─► S3 Parquet Data Lake (Historical partitions permanently unreadable)
└─► S3 Glacier Cold Backups (WORM archives rendered mathematically indecipherable)
### Criteria and weights
| Criterion | Why it matters here | Weight | Source of the weight |
|---|---|---|---|
| Complete Mathematical Erasure across Immutable Storage | Soft-deletes fail GDPR Art. 17 audits; historical Kafka logs and backups must be erased (ENF-4902). | 0.40 | David O'Reilly (Chief Privacy Officer) |
| Statutory Fulfillment Timeline (< 30 Days) | Regulatory mandates impose strict financial penalties for overdue erasure requests. | 0.25 | Marcus Vance (Principal Data Architect) |
| Legal Hold Override & Audit Exception Protection | Financial anti-money laundering (AML) laws mandate 7-year transaction retention over erasure. | 0.20 | Enterprise Legal & Risk Directorate |
| Performance & Key Management Headroom | Managing millions of per-user keys must not degrade 40,000 req/sec payment authorization flows. | 0.15 | Cloud Infrastructure Standard |
### Comparison
| Privacy Architecture Candidate | Erasure Mechanism | Immutable Data Handling | Storage Re-write Cost | Evaluation |
|---|---|---|---|---|
| Option A: Relational Soft-Delete (Legacy) | `is_deleted = true` flag | None (Ignored in logs/backups) | $0 (Zero erasure) | Rejected: Triggered ENF-4902 €4.2M fine; completely non-compliant. |
| Option B: Batch S3/Log Rewriting | Nightly Spark rewrite jobs | Rewrites entire TB parquet logs | Massive compute bills ($45k/mo) | Rejected: Extremely fragile; cannot rewrite immutable WORM backups. |
| Option C: Crypto-Shredding Architecture (Chosen) | Per-user envelope key deletion | Rendered mathematically indecipherable | $0 (Key deletion only) | Selected: Instant erasure, zero compute overhead, legally certified. |
### Result
Option C is selected. Dedicated user-level envelope keys destroyed upon erasure provide instantaneous, mathematically verifiable compliance across all storage tiers.
---
### Required Mechanisms
#### 1. Per-User Envelope Crypto-Shredding [MC-CS-01]
- **Key Hierarchy**:
- Master Key: AWS KMS Root Key (Hardware Security Module HSM-backed).
- User Data Key (UDK): 256-bit AES-GCM key provisioned per user: `arn:aws:kms:eu-central-1:...:key/user-8821a`.
- **Field-Level Encryption**:
- Sensitive PII attributes (`name`, `email`, `address`, `ip_address`, `bank_account`) are encrypted using the user's UDK before writing to any database, event topic, or log.
- **Erasure Actuation**:
- Fulfilling an erasure request calls `kms:ScheduleKeyDeletion(PendingWindowInDays=7)`. The key is instantly disabled, immediately blinding all ciphertext across relational databases, analytical data lakes, and immutable Glacier archives.
#### 2. Distributed Tombstone Event Propagation [MC-EP-01]
- Privacy Orchestrator publishes a compacted tombstone event to Kafka topic `customer.privacy.tombstones`:
```json
```json
{
"event_type": "DATA_SUBJECT_ERASURE_COMPLETED",
"user_id": "usr_8821a",
"shredded_at": "2026-09-15T17:15:00Z",
"legal_holds_active": false
}
- Subscribed microservices receive the tombstone and drop local cache keys, purge active user sessions, and remove user index rows within 24 hours.
#### 3. Automated DSAR State Machine [MC-SM-01]
- **Pipeline Workflow**:
1. *Intake & Verification*: Data subject submits request via self-service privacy portal; authenticated via multi-factor authentication.
2. *Legal Hold Check*: Verifies whether the account has active financial AML holds (FinCEN / BaFin mandates). If active hold exists, financial transactions are retained while non-essential marketing PII is shredded.
3. *Execution*: Shakes out active async worker jobs; executes crypto-shredding key destruction.
4. *Attestation*: Emits signed compliance certificate to user within 7 business days.
#### 4. Data Minimization & Ingress Purpose Gating [MC-DM-01]
- All inbound API payloads pass through a Schema Purpose Validator:
- Fields lacking declared lawful basis (e.g. `consent_marketing: true`) are dropped at API ingress before persisting to memory or disk.
---
### Invariants and Contracts
Cryptographic Erasure Verification Invariant [INV-PRV-01]
Upon confirmation of a GDPR Article 17 erasure request, the subject's cryptographic encryption key
must be permanently disabled within 24 hours. Downstream ciphertext must be indecipherable.
Statutory Thirty-Day Fulfillment Ceiling [INV-PRV-02]
End-to-end Data Subject Access Requests (erasure or portability export) must settle in <= 30 calendar days.
Requests lingering past day 21 escalate as P1 compliance emergencies to David O'Reilly.
Legal Hold Precedence Invariant [INV-PRV-03]
Statutory financial transaction retention requirements (AML / tax regulations) override privacy
erasure requests. Ledger records subject to legal holds must be preserved in segregated storage.
## Explicit Unknowns
- AWS KMS API call latency overhead when resolving 12 million unique customer encryption keys under peak 40,000 TPS load (G-1).
- Cloud data transfer costs during full cross-border DSAR data portability archive downloads (G-2).
## Traceability
| Claim | Classification | Source | Freshness |
|---|---|---|---|
| 12 million consumer profiles across EU & US | provided | Enterprise scope intake | Current |
| 35 microservices spanning databases, lakes, Kafka | provided | Architecture scope intake | Current |
| Regulatory enforcement action ENF-4902 €4.2M fine | provided | Historical audit record | Historical |
| GDPR Article 17 & CCPA/CPRA compliance | provided | Legal compliance mandate | Current |
| Per-user envelope crypto-shredding selection | decided | David O'Reilly & Marcus Vance | 2026-09-15 |
| Statutory 30-day fulfillment ceiling | decided | Architectural invariant INV-PRV-02 | 2026-09-15 |
## Verification
No validator was supplied, so no command was run.
Reviewer self-check against privacy architecture standards:
- **Erasure Efficacy**: PASS. Crypto-shredding renders PII in immutable lakes and backups permanently unrecoverable.
- **Tombstone Automation**: PASS. Kafka tombstone propagation coordinates cache and session purges across 35 services.
- **Legal Balance**: PASS. Automated AML legal hold checks prevent premature destruction of statutory financial ledgers.
- **Markdown Hygiene**: PASS. Native Markdown syntax strictly adheres to `rule_markdown.md`.
## Open Decisions
- `DEC-PRV-01`: David O'Reilly to determine whether differential privacy noise injection should be automated for merchant aggregate sales analytics exports (Owner: David O'Reilly).
## Next steps
1. Marcus Vance provisions AWS KMS Key Policy allowing programmatic creation and deletion of User Data Keys.
2. Platform team embeds field-level envelope encryption middleware into shared application data access objects.
3. Conduct staging compliance game day submitting 50 synthetic DSAR erasure requests to verify complete crypto-shredding in < 7 days.
Connects securely to your tools. The creator never sees your data.
What you get
About this skill
What it does
This skill owns cross-system contracts that make accepted privacy decisions executable across collection, use, disclosure, linking, retention, rights, transfer, migration, and retirement. It traces authoritative purpose and legal/privacy requirements to data flows, controls, owner actions, evidence, and failure behavior without inventing legal meaning.
Use it when
- Personal data crosses products, services, devices, analytics, AI/ML, logs, support, vendors, regions, or controllers/processors
- Purposes, legal-basis decisions, notices, choices, permissions, and downstream uses must remain aligned
- Direct, observed, inferred, derived, sensitive, child, employee, customer, partner, and pseudonymous data need explicit semantics
- Consent/preferences must propagate to collectors, event streams, stores, models, exports, vendors, caches, and deletion workflows
- Minimization, field/record/aggregate precision, access, disclosure, linkability, retention, archival, backups, and deletion interact
- Access, correction, portability, objection/restriction, opt-out, deletion, appeal, and identity-verification handoffs span systems
For example: “A customer asked us to delete their data. We removed the account row, then found their address in the data warehouse, in three years of order exports, and in the shipping partner's system.”
What you get
- architecture/privacy-architect/README.md
- architecture/privacy-architect/00-overview/privacy-architect-overview.md
- architecture/privacy-architect/verification/fitness-self-check.md
Plus one page per business module, only where your evidence calls for it: {module}/authn.md, {module}/authz.md, {module}/session.md, {module}/secrets.md, {module}/audit.md.
All paths are relative to the output folder you choose.
What it will not do
Do not use merely to answer a legal question, draft a privacy policy, add a cookie banner, run a GDPR checklist/DPIA/audit, fulfill one deletion/access request, redact PII, classify data, configure encryption, or assess compliance.
How it works
- Check personal data is genuinely in scope.
- Record the lawful basis and purpose per data category.
- Map the full lifecycle including copies.
- Define the subject rights paths as real operations.
- Choose minimisation and de-identification per purpose, and state what it does not protect against.
- Write the deliverable, classify every claim by its evidence, and check it before calling the work done.
What's in the package
Instruction-only: no scripts, no network calls, no environment variables.
- LICENSE.txt
- SKILL.md
- agents/openai.yaml
- assets/output-template-contract.md
- assets/output-template-domain.md
- assets/output-template-fitness.md
- assets/output-template-mechanism.md
- references/domain-rules.md
- references/operating-rules.md
- references/output-contract.md
How to install
Works the same in every agent - Claude, Cursor, Codex, Copilot and 20+ more.
- 1
Download the ZIP
Free skills download straight away. Paid skills unlock right after purchase.
- 2
Unzip into your skills folder
Every agent reads skills from one folder on your machine. Drop the unzipped folder in there.
- 3
Ask your agent to use it
Restart the agent if it was already running. It picks the skill up automatically - no config needed.
Skills folder by agent
Click the path to copy it. Create the folder if it does not exist yet.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill passes our trust checks below.
Security scanned
Passed our 8-point scan before listing
Fresh listing
Recently published to Agensi
30-day refund
Not a fit? Get your money back
Trust & safety
Security scanned
Verified clean 4 days ago
- Passed all security checks, Safe to install