agentcmd: Skill for Claude Code

Planning Skill

Overview

This skill provides comprehensive planning workflows for features and system changes. It focuses on the research and clarification phase - gathering context, exploring architecture, and asking good questions before spec generation.

When to Use This Skill

Use this skill when:

• In planning/plan mode (auto-trigger if possible)
• User says: "plan this", "research this feature", "help me understand"
• User asks: "what do I need to know before implementing X?"
• Before generating specs with /cmd:generate-spec
• Need to explore codebase patterns and architecture

Core Workflow

1. Research Phase

• Explore codebase for similar existing features
• Identify which domain/area this belongs in
• Locate related types, schemas, services, routes
• Check test patterns in relevant areas
• Review database models (if applicable)
• Use references/research-checklist.md for systematic exploration

2. Context Gathering

• Internal docs: Check .agent/docs/ for relevant patterns
• CLAUDE.md: Review project conventions and rules
• context7 MCP (if available): Fetch external library docs
  • Check for mcp__context7__* tools
  • Use for React, Fastify, Prisma, etc. documentation
  • Fall back to WebFetch/WebSearch if not available
• See references/context-sources.md for detailed guidance

3. Clarification (If Needed)

Ask clarifying questions ONE AT A TIME if implementation approach is unclear:

• Don't use the Question tool
• Use this template:

  **Question**: [Your question]
  **Suggestions**:
  
  1. [Option 1] (recommended - why)
  2. [Option 2]
  3. Other - user specifies
  

• See references/clarification-questions.md for examples and templates

4. Complexity Assessment

• Estimate file count and scope
• Identify cross-cutting concerns
• Assess if new patterns needed or extending existing
• Determine context requirements for implementation
• Flag if complexity suggests breaking into smaller features
• See references/complexity-assessment.md for detailed rubric