Works with the AI Agents you already use
analytical dataset privacy assessor
Releasing analytical datasets without a formal privacy review risks re-identification and regulatory breaches.
$2.99
analytical dataset privacy assessor
Example session with this skill installed
Review the attached 'customer_churn_dictionary.csv' and 50 sample records. Purpose is internal churn analysis by the marketing team. Jurisdiction is EU. Approval ref: AR-992. k_target=5.
- Read your context and instructions
- Compiled the analytical dataset privacy
I have processed the customer churn dataset (E1) and access approval (E2) for the marketing team's internal analysis.
Analytical Dataset Privacy Assessment — customer_churn_dictionary, 2023-11-24
1. Scope, sources and access
- Sources: E1 (Data Dictionary), E2 (Access Approval), E3 (50 Sample Records).
- Purpose: Internal churn analysis.
- Audience: Restricted (Marketing Team).
- Jurisdiction: EU.
- Access Status: Approved (AR-992).
- Fields: 12 fields in scope, 0 excluded.
2. Field classification
| Field | Type | Declared | Effective role | Basis | Conflict |
|---|---|---|---|---|---|
cust_id | String | ID | Direct Identifier | UUID Pattern | None |
postcode | String | QI | Quasi-identifier | Geospatial | None |
birth_year | Integer | QI | Quasi-identifier | Date/Age | None |
churn_status | Boolean | S | Sensitive | Outcome | None |
3. Re-identification screen
Analytical estimate — validation required
- k-min: 2
- Classes below target: 8
- Unique records: 4
- Max re-id probability: 0.50
- L-diversity (min): 1
Note: Measured on 50 records. Anonymization is not claimed.
4. Disclosure control
No output tables were supplied.
5. Minimization plan
| Field | Role | Draft action | Rationale |
|---|---|---|---|
cust_id | Direct ID | Suppress | Not needed for aggregate churn trends. |
postcode | Quasi-ID | Generalize | Reduce to outward code (first 3 chars). |
birth_year | Quasi-ID | Bin | Convert to 5-year age brackets. |
6. Release checklist
| ID | Check | Status | Basis |
|---|---|---|---|
| C01 | Purpose defined | Pass | Internal analysis |
| C05 | k-anonymity check | Fail | k=2 is below target 5 |
| C12 | Minimization draft | Pass | Included in Section 5 |
7. Limitations and items for human review
- Information gap — organizational validation required: Verify if
marketing_segmentcontains free-text notes. - Management decision required: k-anonymity is below the internal threshold of 5.
- Management decision required: Approval of the proposed generalization for
postcode.
This assessment supports, and does not replace, professional and legal judgment.
Next steps
- Apply the generalization rules for
postcodeandbirth_yearto the full dataset. - Re-run the assessment with a larger record sample to verify if k-anonymity improves.
- Obtain management sign-off on the small cell risks identified in the quasi-identifier combinations.
Connects securely to your tools. The creator never sees your data.
About this skill
Releasing analytical datasets without a formal privacy review risks re-identification and regulatory breaches. This skill provides a structured framework to screen data dictionaries and sample records for disclosure risks before they leave your environment.
What it does
- Classifies fields by identifying direct identifiers, quasi-identifiers, and sensitive attributes.
- Measures k-anonymity to quantify the risk of re-identification within a specific record set.
- Assesses l-diversity to ensure sensitive attributes are sufficiently distributed across groups.
- Detects small cells in proposed aggregate output tables to suggest necessary suppressions.
- Drafts minimization plans recommending specific actions for each field based on the intended purpose.
- Generates release checklists covering thirteen critical readiness checkpoints for human review.
How it works
- Register the dataset sources including data dictionaries, access approvals, and optional sample records.
- Provide context on the intended audience, jurisdiction (e.g., GDPR, CCPA), and the output format.
- The skill executes a deterministic assessment script to calculate risk metrics and classification heuristics.
- Review the generated assessment report, which includes field roles, re-identification estimates, and suppression drafts.
Frameworks & tools
Built for data privacy workflows, it utilizes a deterministic Python-based assessment engine for metric calculation. It supports jurisdictions like EU (GDPR) and US-CA (CCPA) and integrates with JSON-schema based data dictionaries.
Why this beats prompting it yourself
Generic LLM prompts often hallucinate privacy compliance or miss subtle quasi-identifier combinations. This skill uses a deterministic script to ensure k-anonymity and l-diversity values are calculated, not estimated, while maintaining a strict evidence-based discipline that never falsely claims anonymization.
Use cases
- Screening a new research dataset for PII before granting internal team access.
- Assessing the re-identification risk of a sample dataset for a restricted third-party release.
- Checking aggregate report tables for small cell counts that could leak individual identities.
- Generating a data minimization draft to justify why specific columns are retained for a project.
Known limitations
Does not provide legal advice or official certification of anonymization. Requires explicit access approval metadata to process record values. Calculations are limited to the supplied sample records.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse All Claude Skills.
Requires explicit access approval metadata. Metrics only reflect the supplied sample. Does not provide legal certification or guarantee compliance with privacy laws.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 3 days ago