project-planning

Ask and document:

1. What problem are you solving? (pain point)
2. Who has this problem? (target users)
3. How is it being solved today? (alternatives)
4. Why will your solution be better? (differentiation)

For each user type:

1. Name and demographics
2. Goals - What do they want to achieve?
3. Pain points - What frustrates them today?
4. User journey - How will they use the product?

Create 2-3 personas minimum.

Guide through requirements gathering:

1. List all features user envisions
2. Categorize by priority:

- P0 - MVP, must have for launch

- P1 - Important, soon after launch

- P2 - Nice to have, future

1. Assign requirement IDs (REQ-AREA-NNN)

Based on identified features and integrations, actively research available MCP servers:

Research steps:

1. List all external services the project will integrate with
2. Web search for each: "MCP server for [service name]"
3. Check official registry: https://github.com/modelcontextprotocol/servers
4. Check community registry: https://mcp.so/

For each discovered server:

• Does it cover needed functionality?
• What are the setup requirements?
• Is it actively maintained?

Document findings in CLAUDE.md

Identify workflows that should become skills:

Ask:

• What multi-step processes will be repeated?
• What commands are complex or error-prone?
• What workflows need consistency?

Research similar projects:

• Web search: "[project type] development workflow"
• What automation do similar teams use?

Document planned skills in CLAUDE.md

Don't assume - research what fits:

1. Web search: "[problem domain] architecture patterns 2025"
2. Web search: "[similar product] tech stack"
3. Consider constraints: budget, team skills, scale requirements

Evaluate options for:

• Frontend - Framework, styling, state management
• Backend - Language, framework, hosting model
• Database - SQL vs NoSQL, managed vs self-hosted
• Infrastructure - Cloud provider, serverless vs containers

Create ASCII architecture diagram.

Define measurable goals:

1. Business metrics - Revenue, users, engagement
2. Technical metrics - Performance, uptime, latency
3. Timeline - Milestones and deadlines

Identify:

1. Technical risks (new tech, scaling, integrations)
2. Business risks (market, competition, timing)
3. Resource risks (budget, timeline, skills)

For each: probability, impact, mitigation strategy.

Generate these documents with research findings:

1. docs/PRD.md - Complete product requirements
2. docs/ARCHITECTURE.md - Technical design with rationale
3. docs/REQUIREMENTS.md - Detailed feature specs
4. .mcp.json - Configured with researched servers
5. .claude/skills/ - Initial skills based on research
6. CLAUDE.md - Updated with full project context

• Don't skip research phases - assumptions cause rework
• Be specific - "fast" → "< 2s load time"
• Prioritize ruthlessly - MVPs should be minimal
• Document research findings - future you will thank you
• Validate MCP servers exist before planning around them