Parallel Research

Research $ARGUMENTS across multiple dimensions simultaneously.

Target

Parse the topic from $ARGUMENTS. If no topic specified, ask the user what they want to research.

Dimensions

Spawn parallel subagents for these research dimensions:

1. •Technical - What's technically possible? Constraints? Requirements?
2. •Ecosystem - What tools/libraries exist? What's standard? What's popular?
3. •Patterns - How do experts approach this? Best practices? Architecture?
4. •Pitfalls - Common mistakes? Gotchas? What to avoid?
5. •Path Forward - Recommended approach? First steps? Priorities?

Process

Phase 1: Parallel Investigation

For each dimension, spawn a subagent to:

• •Investigate the dimension thoroughly
• •Save structured findings to .research/<dimension>.md
• •Return a 2-3 sentence summary

Run all 5 dimensions in parallel.

Subagent instructions: Each subagent should focus ONLY on its assigned dimension. Do not overlap with other dimensions. Save findings in this format:

# [Dimension] Research: [Topic]

## Findings
- [Key finding 1]
- [Key finding 2]
- [Key finding 3]

## Details
[Detailed explanation with sources]

## Confidence
[HIGH/MEDIUM/LOW with rationale]

Phase 2: Synthesis

After all subagents complete:

1. •Read all .research/*.md files
2. •Identify cross-dimensional themes (multiple dimensions agree)
3. •Resolve contradictions (dimensions disagree)
4. •Prioritize findings:
   • •HIGH: Multiple dimensions confirm
   • •MEDIUM: Single dimension, well-supported
   • •LOW: Speculative or uncertain
5. •Synthesize into unified recommendations
6. •Clean up .research/ directory (remove temporary files)

Output Format

After synthesis, provide output in this exact structure:

Research Complete: [Topic]

Summary: [2-3 sentence executive summary synthesizing all dimensions]

Key Findings

Recommended Approach

[Synthesized recommendation with rationale, drawing from all dimensions]

Watch Out For

• •[Key pitfall 1 from Pitfalls dimension]
• •[Key pitfall 2]
• •[Key pitfall 3]

Next Steps

1. • [Specific action item from Path Forward]
2. • [Specific action item]
3. • [Specific action item]

Sources

• •[URLs or references used during research]

Important

• •This skill uses parallel subagent execution for faster research
• •Each subagent has its own isolated context window
• •Findings are saved to files for reliable synthesis
• •The .research/ directory is temporary and cleaned up after synthesis
• •Output is a structured summary, not raw exploration logs