Research

Additional focus (optional): $ARGUMENTS

Context

• Current tasks: !ls -la .tasks/ 2>/dev/null || echo "No .tasks directory"

Your Task

1. Identify the current task and read:

   • .tasks/{task-name}/task.md (requirements)
   • .tasks/{task-name}/explore.md (codebase context)

2. Identify what needs to be researched based on requirements and stack

3. IMPORTANT: Delegate research to the research sub-agent with GRANULAR, SPECIFIC questions. Do NOT send one big research request. Instead, spawn MULTIPLE research agents for specific topics:

   Example - if implementing auth:

   • Agent 1: "How to configure Better Auth 1.2.3 with session strategy"
   • Agent 2: "Drizzle ORM schema patterns for auth tables"
   • Agent 3: "Next.js 15 middleware for route protection"
   • Agent 4: "Better Auth + Drizzle adapter configuration"

   Each agent should focus on ONE specific thing and go deep.

4. Aggregate all research findings into .tasks/{task-name}/research.md:

   # Research: {task-name}
   
   ## Stack Reference
   [Installed versions for reference]
   
   ## Topic 1: [Name]
   ### Summary
   [What we learned]
   ### Code Snippets
   [Copy-paste ready examples]
   ### Source
   [Links to official docs]
   
   ## Topic 2: [Name]
   ...
   
   ## Key Considerations
   [Important patterns, pitfalls, best practices]

5. Update .tasks/{task-name}/status.md activity log

6. Use plan mode to:

   • Summarize research findings
   • Ask if any areas need deeper research
   • Confirm we have enough info to create the spec

Notes

• Research must match INSTALLED versions, not latest
• Better to over-research than miss something
• Code snippets should be actionable and copy-paste ready
• Always cite official documentation sources