prd-diagram

User Input

$ARGUMENTS

You MUST consider the user input before proceeding (if not empty).

Goal

Extract structured data from specific PRD sections and generate Mermaid diagram syntax. Each diagram type maps to a specific PRD section. Output is saved as a .mmd file that can be rendered by any Mermaid-compatible viewer.

Execution Steps

1. Resolve Input

Parse $ARGUMENTS for:

• •PRD path (required) — path to the PRD file
• •Diagram type (optional) — one of: state-machine, user-flow, data-model, all

If no diagram type is specified, default to all (generate all applicable diagrams).

If $ARGUMENTS is empty, abort with: "Please provide a PRD path and optional diagram type. Example: /prd-diagram prd/my-feature/prd.md state-machine"

Read the PRD file. If it does not exist, abort with a clear error message.

2. Determine Output Path

• •Feature name from parent directory: prd/my-feature/prd.md → my-feature
• •Output directory: prd/<feature-name>/diagrams/
• •Output files: <type>.mmd (e.g., state-machine.mmd, user-flow.mmd, data-model.mmd)

3. Generate Diagrams

A. State Machine Diagram (state-machine)

Source: Section 10 — State Machines & Lifecycle

Extract state transition tables and generate Mermaid stateDiagram-v2:

stateDiagram-v2
    [*] --> Draft
    Draft --> Active : Organizer starts competition
    Active --> Completed : All fixtures completed
    Active --> Cancelled : Organizer cancels
    Draft --> Cancelled : Organizer cancels
    Completed --> [*]
    Cancelled --> [*]

Extraction rules:

• •Each row in a state transition table becomes an arrow
• •"From" column → source state
• •"To" column → target state
• •"Trigger" column → arrow label
• •Add [*] for initial and terminal states
• •If multiple entities have state machines, generate one diagram per entity (separate files: state-machine-<entity>.mmd)

B. User Flow Diagram (user-flow)

Source: Section 6 — User Stories & Flows

Extract numbered user flow steps and generate Mermaid flowchart TD:

flowchart TD
    A[Landing Page] --> B[Click 'Create Competition']
    B --> C{Authenticated?}
    C -->|Yes| D[Competition Form]
    C -->|No| E[Login Page]
    E --> D
    D --> F[Add Teams]
    F --> G[Generate Fixtures]
    G --> H[Competition Active]

Extraction rules:

• •Each numbered step becomes a node
• •Sequential steps become arrows
• •Conditional steps (if/else, error paths) become diamond decision nodes
• •Use descriptive labels on arrows for conditions
• •Generate one diagram per major user flow (separate files: user-flow-<flow-name>.mmd)

C. Data Model Diagram (data-model)

Source: Section 8 — Data Model & Validation Rules

Extract entity definitions and relationships, generate Mermaid erDiagram:

erDiagram
    Competition ||--o{ Fixture : has
    Competition ||--o{ Standing : tracks
    Fixture }o--|| Team : home
    Fixture }o--|| Team : away
    Fixture ||--o{ MatchEvent : records

    Competition {
        string id PK
        string name
        enum status
        date startDate
    }

    Team {
        string id PK
        string name
        int playerCount
    }

Extraction rules:

• •Each named entity/model becomes an ER entity
• •Fields become attributes with types
• •Required fields are marked
• •Relationships inferred from foreign key references, "belongs to", "has many" language
• •Use standard cardinality notation: ||--o{ (one-to-many), ||--|| (one-to-one), }o--o{ (many-to-many)

4. Save & Report

1. •Create the diagrams/ directory if it doesn't exist.
2. •Write each diagram to its output file.
3. •Report:
   • •Files created with paths
   • •Diagram type and source section for each
   • •Any sections that were missing or insufficient for diagram generation
   • •Rendering hint: "Preview these files in any Mermaid-compatible viewer (VS Code, GitHub, Mermaid Live Editor)"

Operating Principles

• •Do not ask questions — this is a generator skill
• •If a source section is missing or empty, skip that diagram type and note it in the report
• •Generate clean, readable Mermaid syntax with proper indentation
• •Use descriptive node labels (not cryptic IDs)
• •Keep diagrams under 50 nodes — split into multiple diagrams if larger
• •Never modify the source PRD
• •If no diagram-worthy content is found, report: "No diagrammable content found. Ensure the PRD has State Machines (§10), User Flows (§6), or Data Model (§8) sections."