event-storming（事件风暴）

Event Storming

Interactive Workshop Configuration

Use AskUserQuestion to configure the event storming session:

# Question 1: Workshop Mode (MCP: Event Storming methodology)
question: "Which event storming mode do you need?"
header: "Mode"
options:
  - label: "Full Simulation (Recommended)"
    description: "All 5 personas debate in parallel, 6 phases (~15K tokens)"
  - label: "Quick"
    description: "Single analysis pass, no personas (~3K tokens)"
  - label: "Guided"
    description: "Interactive with user, personas on-demand"
  - label: "Codebase Analysis"
    description: "Discover events from existing code structure"

# Question 2: Domain Scope (MCP: DDD bounded context patterns)
question: "How complex is the domain you're exploring?"
header: "Scope"
options:
  - label: "Single Bounded Context"
    description: "One domain area, focused discovery"
  - label: "Multiple Contexts"
    description: "Identify 2-5 bounded context boundaries"
  - label: "Enterprise Domain"
    description: "Comprehensive cross-domain mapping"
  - label: "Unknown"
    description: "Help me determine scope first"

Use these responses to select the appropriate workshop mode and calibrate depth of analysis.

When to Use This Skill

Use this skill when you need to:

• Discover domain events for a business process
• Identify commands and actors in a system
• Find bounded context boundaries
• Simulate a multi-stakeholder event storming workshop
• Prepare for actual event storming facilitation
• Analyze an existing codebase for domain events

Keywords: event storming, domain events, commands, actors, bounded contexts, aggregates, domain discovery, workshop simulation, multi-persona, DDD

What is Event Storming?

Event Storming is a workshop-based method for collaborative domain discovery. Participants use sticky notes to map out:

• Events (Orange) - Things that happen in the domain
• Commands (Blue) - Actions that trigger events
• Actors (Yellow) - Who issues commands
• Aggregates (Yellow) - Business entities that handle commands
• Read Models (Green) - Information needed for decisions
• Policies (Purple) - Business rules and reactions
• External Systems (Pink) - Outside integrations
• Hot Spots (Red/Pink) - Areas of confusion or conflict

Workshop Modes

This skill supports three modes of operation:

Mode Selection Guide

Use full-simulation when:

• Starting a new project or major feature
• You need comprehensive domain discovery
• Multiple perspectives are valuable
• Time is less critical than thoroughness

Use quick when:

• You need a fast domain overview
• Token budget is constrained
• Domain is relatively simple
• You'll refine later

Use guided when:

• You want to drive the process
• You have specific questions
• You want to invoke specific personas
• Interactive exploration is preferred

Multi-Persona Simulation

The full-simulation mode uses 5 specialized agents to simulate different stakeholder perspectives:

6 Workshop Phases (Full Simulation)

Phase 1: Chaotic Exploration

All personas brainstorm events independently. No constraints, no ordering.

Orchestration:

Launch 5 parallel Task agents:
- Task(event-storming-persona domain-expert, "Identify all domain events for: {domain}")
- Task(event-storming-persona developer, "Identify technical events for: {domain}")
- Task(event-storming-persona business-analyst, "Identify commands and actors for: {domain}")
- Task(event-storming-persona product-owner, "Prioritize and identify MVP scope for: {domain}")
- Task(event-storming-persona devils-advocate, "Challenge and identify gaps for: {domain}")

Phase 2: Timeline Ordering

Synthesize and order events chronologically. Create the timeline.

Phase 3: Command Discovery

Identify what triggers each event. Map commands to events.

Phase 4: Actor Identification

Map who issues each command. Define roles and systems.

Phase 5: Bounded Context Discovery

Group related events/commands. Identify natural boundaries.

Phase 6: Hot Spot Resolution

Devil's advocate challenges. Resolve conflicts and gaps.

Detailed phase guidance: See references/workshop-facilitation.md

Sticky Note Color Convention

Detailed conventions: See references/sticky-note-types.md

Orchestration Pattern

Since Claude Code subagents cannot spawn other subagents, the main conversation orchestrates:

Main Conversation
    ↓
Invokes event-storming skill
    ↓
Skill guides parallel Task tool calls:
    ├── Task(event-storming-persona domain-expert, prompt)
    ├── Task(event-storming-persona developer, prompt)
    ├── Task(event-storming-persona business-analyst, prompt)
    ├── Task(event-storming-persona product-owner, prompt)
    └── Task(event-storming-persona devils-advocate, prompt)
    ↓
Skill synthesizes results with provenance tracking
    ↓
Outputs event catalog with [persona] attribution

Quick Start

Full Simulation Mode

I want to run a full event storming simulation for an e-commerce order management system.

Please:
1. Launch all 5 persona agents in parallel
2. Have them analyze the domain
3. Synthesize their findings
4. Identify bounded contexts
5. Resolve any hot spots

Quick Mode

Give me a quick event storm overview for a subscription billing system.
Focus on the core happy path events.

Guided Mode

Let's do a guided event storming session for a hospital appointment system.
Start with the patient booking journey and I'll guide from there.

Output Format

The event storming session produces a structured event catalog:

# Event Storm: [Domain Name]

## Event Catalog

### [Bounded Context Name]

**Events:**
- [Event Name] [Domain Expert] - [Description]
- [Event Name] [Developer] - [Description]

**Commands:**
- [Command Name] → [Event Name] [Business Analyst]

**Actors:**
- [Actor Name]: [Commands they can issue]

**Aggregates:**
- [Aggregate Name]: [Events it produces]

**Policies:**
- [Policy Name]: [Trigger] → [Action]

## Bounded Contexts Identified

1. [Context Name]
   - Core Domain / Supporting / Generic
   - [Events in this context]

## Hot Spots

- [Issue] - [Resolution or TODO]

Full template: See references/templates/event-storm-output.md

Integration with Other Skills

Event storming connects with:

• domain-storytelling - Stories feed into event discovery
• modular-architecture - Bounded contexts become modules
• fitness-functions - Module isolation tests
• adr-management - Document bounded context decisions

Workflow:

Domain Storytelling → Event Storming → Modular Architecture
(understand "what")   (design "how")   (implement "where")

Best Practices

1. Start with events - Events are facts, commands are debatable
2. Use past tense - "Order Placed" not "Place Order"
3. Explore boundaries - Events help find module boundaries
4. Embrace chaos - Initial brainstorming should be messy
5. Track provenance - Know which perspective each insight came from
6. Resolve hot spots - Don't leave conflicts unaddressed

References

• references/workshop-facilitation.md - Phase orchestration and timing
• references/persona-prompts.md - Prompt templates for each persona
• references/sticky-note-types.md - Color conventions and usage
• references/bounded-context-discovery.md - Context identification patterns
• references/templates/event-storm-output.md - Output format with provenance

User-Facing Interface

When invoked directly by the user, this skill runs an AI-simulated event storming workshop.

Execution Workflow

1. Parse Arguments - Extract domain description and mode. If no description provided, ask the user what domain to event storm. Default mode is guided.
2. Execute Based on Mode:

- Full: Launch 5 persona agents in parallel (domain-expert, developer, business-analyst, product-owner, devils-advocate). Synthesize outputs with provenance tracking, merge event catalogs, identify hot spots, and group into bounded context candidates. - Quick: Single-pass analysis identifying domain events, commands, actors, and potential bounded context boundaries. - Guided: Interactive 6-phase workshop (Chaotic Exploration, Timeline Ordering, Command Discovery, Actor Identification, Bounded Context Discovery, Hot Spot Resolution).

1. Generate Output - Produce event catalog with domain events, commands, bounded context candidates, and hot spots.
2. Save Results - Save to docs/event-storming/[domain-name]-[date].md (or custom --dir).
3. Suggest Follow-Ups - Recommend domain storytelling for unclear processes, context mapping for integration strategies, and ADRs for key decisions.

Version History

• v1.0.0 (2025-12-22): Initial release

- Multi-persona simulation with 5 agents - Three workshop modes (full, quick, guided) - 6-phase workshop structure - Provenance tracking for insights - Integration with domain-storytelling and modular-architecture

Last Updated

Date: 2025-12-22 Model: claude-opus-4-5-20251101