privacy-dfd
Create or update Microsoft privacy compliance Data Flow Diagrams (DFDs) and data inventory tables for services. Use when: privacy review, data flow diagram, DFD, data inventory, GDPR, data classification, privacy compliance, 1CS privacy artifact, data handling attestation.
- License 许可
- MIT
- 所属仓库
- microsoft/rayfin
- 最近更新
- 2026-11-12
- Stars
- ★ 662
- Forks
- 77
01 前言 Front Matter
- name
- privacy-dfd
- description
- Create or update Microsoft privacy compliance Data Flow Diagrams (DFDs) and data inventory tables for services. Use when: privacy review, data flow diagram, DFD, data inventory, GDPR, data classification, privacy compliance, 1CS privacy artifact, data handling attestation.
- argument-hint
- Describe the service or component to diagram, or ask to update an existing DFD/inventory.
02 资源文件 6
- 参考文档references/classifications.md3.4 KB
- 参考文档references/data-subjects.md2.1 KB
- 参考文档references/dfd-conventions.md2.2 KB
- 参考文档references/inventory-columns.md4.3 KB
- 参考文档references/key-questions.md2.0 KB
- 参考文档references/personal-data.md1.5 KB
03 SKILL.md 正文
Privacy DFD and Data Inventory
Produce Microsoft privacy compliance artifacts — a Mermaid Data Flow Diagram (DFD) and a Markdown data inventory table — for a service or system under privacy review.
When to Use
- Creating a new privacy DFD for a service
- Updating an existing DFD after architecture changes
- Building a data inventory table for privacy review
- Preparing 1CS privacy review artifacts
- Answering privacy review key questions (Who, What, Where, When, Why, How, AI/ML)
- Classifying data elements with Microsoft taxonomy
Outputs
- Mermaid DFD (
.mmdfile) — a singleflowchart LRshowing all privacy-relevant data flows - Data Inventory Table (
.mdfile) — a Markdown table listing every data element the service collects, processes, stores, or transmits
Procedure
Step 1: Gather Context
Before diagramming, collect evidence from code and docs. Identify:
- Personas: Who interacts with the system? (Builders, App Users, Admins, platform services)
- Data stores: Databases, caches, blob storage, telemetry sinks
- External dependencies: Platform shared services, identity providers, CDNs, gateways
- Data flows: What data moves between components, in what direction, with what classification?
- Auth flows: Token issuance, refresh, validation, session management
- Telemetry: What events are emitted, where do they go, what properties do they carry?
Evidence sources to check:
- Architecture docs and RFCs
- Controller and API endpoint code
- Auth/identity modules
- Database models and schema definitions
- Telemetry/logging configuration
- CLI and extension code (for developer tool telemetry)
- UX extension code (for portal interactions)
Step 2: Classify Data Elements
Apply Microsoft data classifications to every data element. Use the official taxonomy — do not invent classifications.
Key rules:
- If classification is uncertain, include the element but mark as
Requires confirmationin Comments - A single data element can have multiple classifications (e.g., a JWT may be EUPI + Access Control Data)
- Distinguish between data at rest (>= 48 hrs) and data in transit/processing (< 48 hrs)
- Identify the data subject for each personal data element
Step 3: Build the DFD
Create a single flowchart LR Mermaid diagram. Follow the DFD conventions.
Key requirements:
- Use subgraphs to group related components (e.g., "Platform Shared Services", "Workload Boundary", "Runtime Integrations")
- Label every edge with: data classification(s) and sample reference numbers cross-referencing the inventory table (e.g.,
Sample #1, #3) - Show external clients, platform shared services, the service boundary, dependent Azure services, and data stores
- Include telemetry flows (where telemetry data goes, what classification)
- Mark direction-uncertain flows with a comment
%% direction uncertain - Include a title comment at the top with service name and date
Step 4: Build the Data Inventory
Create a Markdown table with the required columns.
Key requirements:
- Number each row with a
#column that matches the DFD edge sample references - Provide concrete sample data (realistic but fake values)
- Fill in all columns — use
N/Avariants from the legend when a column does not apply - For GDPR DSR processing, specify the mechanism (agents, API call, manual, etc.)
- For retention, specify exact periods or policies
- For geo-boundary questions, state whether within or outside
Step 5: Validate Completeness
Check the artifacts against the privacy review key questions:
- Who — Is every data subject identified? Are collection sources clear?
- What — Is every data element classified? Are controls and mitigations noted?
- Where — Are all storage locations identified with geo-boundary status?
- When — Are retention periods specified for all persisted data?
- Why — Is the purpose of each data collection clear?
- How — Are notice/consent mechanisms documented? Is data usage clear?
- AI/ML — Is AI usage explicitly stated (including "No AI" if applicable)?
Also verify:
- Every processing component in the DFD has both inbound and outbound arrows
- Every external data transfer has labeled data type plus classification
- The DFD sample reference numbers match the inventory row numbers
- No data flows are invented — all are evidenced from code or docs
- Uncertain flows are marked as such
Step 6: Document Confidence
For each data flow or element, assess confidence:
- High: Directly evidenced in code with clear implementation
- Medium: Evidenced in architecture docs or partially implemented
- Low: Inferred from naming, comments, or design docs only
Mark anything uncertain with "Requires confirmation" in Comments.
Constraints
- Do not invent data flows not evidenced in code or architecture docs
- Treat platform internals (Fabric, Azure services) as opaque external systems
- Do not create threat models — this is a privacy review, not a security review
- If the service has no AI/ML features, state that explicitly
- Use the prior analysis as a starting point but verify claims against source code
- Prefer completeness over minimalism — the output should be ready for export as a privacy review artifact
References
- Data Classifications — Microsoft Enterprise Data Taxonomy
- Data Subjects — Data subject definitions
- Inventory Columns — Required columns and value options
- DFD Conventions — Mermaid diagram shape and labeling rules
- Key Questions — Privacy review key questions and data lifecycle
- Personal Data — Which classifications constitute personal data
04 同仓库更多 Skills
rayfin-getting-started
Use when starting or creating a NEW Rayfin app, or when a Rayfin task comes up and you are not yet inside a Rayfin project. Gets you into a project with the Rayfin CLI, then hands off to the authoritative, version-locked in-project rayfin skill/docs that own all in-project work. Triggers: build a Rayfin app, start a Rayfin project, create a new Rayfin app, create-rayfin, npm create @microsoft/rayfin, rayfin init, scaffold rayfin, rayfin CLI, rayfin template, universal app, awesome-rayfin gallery, get started with Rayfin
visuals
Use when user wants to incorporate charts, graphs, data grid, or other visual representations of data into their project. Use VegaVisual and DataGrid components to create these visuals, utilizing the shared DataTable input, formatting, theme, and interactivity. Covers the onInteraction selection/click events host apps consume, named multi-table data input for layered overlays and reference lines, and Vega-Lite native selections. Triggers: chart, graph, plot, dashboard, KPI, metric, trend, scorecard, data grid, visualize.
schema-discovery
Explore Power BI semantic model schemas using DAX INFO functions. Progressive metadata discovery strategy, scope estimation, and narrowing techniques for tables, columns, measures, and relationships.
query-design
Separates DAX data-fetching from TypeScript presentation. Guides when to use DAX vs. TypeScript vs. Vega-Lite for aggregation, total rows, FORMAT(), SELECTCOLUMNS, BLANK handling, filtering, multi-grain queries, cross-filtering and cross-highlight subset/overlay queries, and format strings.
functions-capability
Apply and build the Universal App functions capability. Use for trusted server-side logic, secrets, external APIs, custom model endpoints, published Fabric data agents, webhooks, or privileged operations.
fabric-sdk
How to query Fabric data from a running app through the Rayfin connectors client. Use getRayfinClient().connectors.<name>.executeQuery to run DAX against a semantic model, and toQueryResult to normalize the response.