requirements-gatherer

name	requirements-gatherer
description	Use when gathering project requirements - asks structured questions about project type, scope, integrations, and constraints to populate project_state.md
version	1.3.1
allowed-tools	Read, Write, Glob
user-invocable	false

Requirements Gatherer

Ask structured questions to define project requirements.

Activation

Activate when:

• Invoked by project-initializer after project setup
• User says "Gather requirements" or "Define requirements"
• Starting Phase 1 on a new project

Workflow

1. Load Project Context

Use Read to get {project_path}/project_state.md. Extract:

• Project name
• Current scope (if any)
• Existing requirements (if resuming)

2. Ask Questions by Category

Ask one category at a time. Wait for response before proceeding.

Category 1: Project Type & Scope

What type of project is this?
- [ ] New module or package (starting from scratch)
- [ ] Feature for an existing module or package
- [ ] Framework core patch or bug fix
- [ ] Theme, template, or UI layer
- [ ] Configuration or site/app building
- [ ] Other (Claude Code plugin, documentation, etc.)

What does this project include?
- [ ] A single module or package
- [ ] Multiple related modules or packages
- [ ] A theme or template layer
- [ ] Configuration or build artifacts
- [ ] Mix of the above

Please describe:

Category 2: Core Functionality

What is the main purpose?
1. What problem does it solve?
2. What are the must-have features?
3. What are nice-to-have features?

Category 3: User Roles

Who will use this?
1. Which user roles need access?
2. What permissions are needed?
3. Any role-specific features?

Category 4: Data Requirements

What data is involved?
1. New data types or entities needed?
2. New fields on existing types?
3. Configuration data to store?
4. Relationships to existing data?

Category 5: Integrations

What existing systems does this connect to?
1. Existing packages or modules to integrate with?
2. External APIs or services?
3. Event or workflow hooks needed?
4. Third-party libraries?

Category 6: UI Requirements

What interfaces are needed?
1. Admin forms or pages?
2. Frontend display requirements?
3. Specific UX patterns to follow?
4. Mobile considerations?

Category 7: Constraints

What limitations should we know about?
1. Framework or runtime version?
2. Performance requirements?
3. Accessibility requirements (WCAG level)?
4. Browser support needs?

3. Handle Responses

For each category:

• Accept the answer
• Allow "skip" or "not sure yet" - mark as open question
• Summarize understanding back to user
• Proceed to next category

4. Update project_state.md

After all categories, use Edit tool to update the Requirements section:

## Requirements

### Project Type
- {new module/package, existing feature, framework patch or fix, theme/template, config, other}

### Scope
- {what project includes}

### Core Functionality
- {requirement 1}
- {requirement 2}

### User Roles
| Role | Permissions | Features |
|------|-------------|----------|
| {role} | {perms} | {features} |

### Data Requirements
- {data type}: {storage approach}

### Integrations
- {module/service}: {integration type}

### UI Requirements
- {element}: {requirement}

### Constraints
- {constraint}: {details}

### Open Questions
- {questions marked as "not sure yet"}

5. Confirm Requirements

Show summary to user:

Requirements gathered for {project_name}:

Core: {1-2 sentence summary}
Roles: {list roles}
Integrations: {list integrations}
Open questions: {count}

Does this look complete? (yes/no/add more)

If "add more", ask what category to expand.

6. Transition to Task Definition

After user confirms requirements are complete:

Requirements complete!

Now let's define your first task. What feature or component do you want to work on first?

Examples:
- "Add settings form for API configuration"
- "Define data model for storing records"
- "Build admin dashboard"

Each task will go through:
1. Research (understand & find patterns)
2. Architecture (design the approach)
3. Implementation (build with TDD)

Enter your first task:

IMPORTANT: Do NOT suggest /research or /design at the project level. Tasks are what go through phases, not the project.

7. Create First Task

When user provides a task name:

1. Validate task name - convert to lowercase with underscores (e.g., "Add settings form" → settings_form)

2. Confirm task creation:

   Creating task: {task_name}
   
   Brief description (one sentence - what does this task accomplish?):
   

3. Wait for user to provide description

4. Invoke research phase:

   Starting research phase for: {task_name}
   

   Then invoke /ai-dev-assistant:research {task_name}

CRITICAL: Do NOT start research until:

• Task name is confirmed
• User provides a description
• Task creation is acknowledged

Stop Points

STOP and wait for user after:

• Each category question
• Showing summary for confirmation
• Asking for first task
• After asking for task description (Step 7.2)
• User says "add more"