research

Research and gather knowledge about: $ARGUMENTS

ultrathink — use extended reasoning for question formulation and knowledge synthesis.

Context

Read CLAUDE.md for project architecture and conventions. Read claude/architecture.md for the tech stack.

Instructions

Phase 1: Analyze Information Gaps

Based on $ARGUMENTS and the project context:

1. Identify what you already know from the codebase
2. Identify knowledge gaps where external information would be valuable
3. Categorize gaps: technical implementation, best practices, security, performance, alternatives

Phase 2: Generate Targeted Research Questions

Generate 5-7 focused research questions that are:

• Specific — target a distinct piece of information, not general knowledge
• Actionable — the answer should directly inform a design or implementation decision
• Prioritized — most critical questions first
• Diverse — cover different aspects (implementation patterns, pitfalls, performance, security)

Example questions:

• "What are the current best practices for [specific pattern] in [framework version]?"
• "What are common pitfalls when integrating [library A] with [library B]?"
• "How does [approach A] compare to [approach B] for [specific use case] in terms of performance and maintainability?"

Phase 3: Research

For each question:

1. Check the codebase first — grep for existing patterns, read relevant files
2. Check project docs — claude/*.md may already have the answer
3. If MCP web search is available (check .mcp.json): use mcp__SERVER_KEY__search to find current information
4. If no web search: use your training knowledge, clearly marking confidence levels

Phase 4: Synthesize Findings

Present findings as a structured report:

## Research Report: [Topic]

### Question 1: [Question]
**Confidence:** [High/Medium/Low]
**Finding:** [Concise answer]
**Source:** [Codebase pattern / MCP search / Training knowledge]
**Recommendation:** [Actionable next step]

### Question 2: ...