SOW (Statement of Work) Generator

A Claude Code workspace configured with specialized AI agents to help create comprehensive Statement of Work documents from project requirements.

What is This?

This is a Claude Code workspace setup that orchestrates multiple AI agents to collaborate on creating professional SOW documents. The system uses:

• Custom AI Agents: Specialized personas for business analysis and system architecture
• Structured Workflows: Step-by-step processes for requirement gathering, specification writing, and work planning
• Document Generation: Automated creation of specs, work items, risks, and assumptions

Components

AI Agents

Located in .claude/agents/:

• business-analyst.md: Expert Business Analyst agent with 15+ years experience in requirements engineering. Handles:

  • Requirements discovery and clarification
  • Asking probing questions to uncover true needs
  • Converting requirements into testable specifications
  • Documentation in structured scenario format

• system-architect.md: Experienced Software Architect with 15+ years designing production systems. Handles:

  • Analyzing codebases to identify implementation gaps
  • Breaking down specs into work items
  • Estimating development effort
  • Identifying risks and assumptions
  • Prioritizing battle-tested technologies and patterns

Commands

Located in .claude/commands/:

• sow.md: Main orchestration command that guides the entire SOW creation process

How to Use

Creating a Statement of Work

1. Start the SOW Command:

   Use the /sow command in Claude Code
   

2. Provide Requirements:

   • Describe your project, feature, or system requirements
   • The business-analyst agent will ask clarifying questions

3. Answer Clarifying Questions:

   • The BA will probe for unclear requirements
   • Provide answers to help refine the specifications

4. Review Specifications:

   • The BA will generate a specs document with format:
     • Requirement statements (SHALL format)
     • Scenarios with WHEN/THEN conditions
   • File created: YYYY-MM-DD-[project-title]-specs.md

5. Add Codebase Context (if applicable):

   • Use /add-dir to add relevant codebase directories
   • The system-architect will analyze the code

6. Review Work Items:

   • The architect will identify gaps and create work items
   • Development effort estimates are provided (assumes senior dev + AI)
   • Files created:
     • YYYY-MM-DD-[project-title]-work-items.md
     • YYYY-MM-DD-[project-title]-risks-and-assumptions.md

7. Get CSV Outputs:

   • Automatically generates CSV files for project management:
     • YYYY-MM-DD-[project-title]-work-items.csv
     • YYYY-MM-DD-[project-title]-risks.csv
     • YYYY-MM-DD-[project-title]-assumptions.csv

Workflow Steps

The SOW command orchestrates these steps automatically:

1. BA analyzes requirements and asks questions → STOP
2. BA converts requirements to specs → STOP
3. User adds codebase directories via /add-dir → STOP
4. Architect analyzes codebase and identifies gaps
5. Architect creates work items with effort estimates
6. Architect documents risks and assumptions
7. Auto-generate CSV files for all outputs

Output Documents

All generated documents are stored in history/[project-name]/:

• Specs: YYYY-MM-DD-[project-title]-specs.md - Testable requirements with scenarios
• Work Items: YYYY-MM-DD-[project-title]-work-items.md - Implementation tasks with estimates
• Risks & Assumptions: YYYY-MM-DD-[project-title]-risks-and-assumptions.md - Project risks and assumptions
• CSVs: Spreadsheet-ready formats for project management tools

Specification Format

Requirements follow this structure:

### Requirement: [Feature Name]
The system SHALL [clear, testable requirement statement]

#### Scenario: [Scenario Name]
- **WHEN** [preconditions and actions]
- **THEN** [expected outcomes and system behavior]

#### Scenario: [Edge Case Name]
- **WHEN** [corner case condition]
- **THEN** [expected system response]

Configuration

Permissions

Located in .claude/settings.local.json:

{
  "permissions": {
    "allow": [
      "Bash(/tmp/parse_markdown.py:*)",
      "Bash(python3:*)",
      "Bash(wc:*)"
    ]
  }
}

Best Practices

1. Be Thorough with Initial Requirements: The more context you provide upfront, the fewer clarification rounds needed
2. Answer BA Questions Carefully: The quality of specs depends on clear answers
3. Add Relevant Codebase: Only include directories relevant to the project for accurate gap analysis
4. Review Each Stop Point: The workflow has built-in stops for your review and approval
5. Keep Documents Organized: All outputs go into history/[project-name]/ for version control

Agent Capabilities

Business Analyst

• 5 Whys technique for root cause analysis
• User story mapping
• MoSCoW prioritization
• SMART requirements validation
• Non-functional requirements elicitation

System Architect

• Proven technology patterns (3+ years production use)
• Operational complexity analysis
• Scalability and performance planning
• Security and compliance considerations
• Risk identification and mitigation strategies

Contributing

When enhancing this workspace:

1. Update agent prompts in .claude/agents/ for improved behavior
2. Modify workflow in .claude/commands/sow.md for process changes
3. Update permissions in .claude/settings.local.json as needed
4. Document changes in this README