/quick-web-research

allowed-tools: Task, WebSearch, WebFetch, Write, Bash(gdate:*)
name: "Quick Web Research"
description: "Ultra-fast parallel web research using sub-agents for 5-10x speedup"
author: "wcygan"
tags: ["analyze","research"]
version: "1.0.0"
created_at: "2025-07-14T00:00:00Z"
updated_at: "2025-07-14T00:00:00Z"

Context

• Session ID: !`gdate +%s%N`
• Research query: $ARGUMENTS
• Results directory: /tmp/research-results-$SESSION_ID/
• Current timestamp: !`gdate '+%Y-%m-%d %H:%M:%S'`
• Working directory: !`pwd`

Your Task

**IMMEDIATELY DEPLOY 8 PARALLEL SUB-AGENTS** for lightning-fast research on: "$ARGUMENTS"

STEP 1: Initialize Research Session

• Create session state file: /tmp/research-state-$SESSION_ID.json
• Initialize results directory: /tmp/research-results-$SESSION_ID/
• Log research parameters and timestamp

STEP 2: Parallel Research Execution

**LAUNCH ALL 8 AGENTS SIMULTANEOUSLY:**

1. **Primary Search Agent**: Execute main WebSearch for "$ARGUMENTS" 2. **Alternative Search Agent**: Search with refined/alternative query terms 3. **Technical Docs Agent**: Focus on official documentation and APIs 4. **Best Practices Agent**: Find current industry best practices 5. **Tutorial Agent**: Locate practical tutorials and examples 6. **Community Agent**: Search Stack Overflow, forums, discussions 7. **Security Agent**: Research security considerations and vulnerabilities 8. **Performance Agent**: Find performance benchmarks and optimizations

Each agent should:

• Perform independent searches with domain-specific focus
• Extract and analyze 2-3 top sources
• Rate content quality and relevance
• Save findings to /tmp/research-results-$SESSION_ID/agent-N.md

STEP 3: Parallel Content Analysis

**NO SEQUENTIAL PROCESSING** - All agents work concurrently:

• Each agent uses WebFetch on their discovered sources
• Focus on their specific domain expertise
• Extract actionable insights and patterns
• Identify current vs outdated information

STEP 4: Synthesis and Analysis

After all agents complete:

• Aggregate findings from all 8 parallel research streams
• Cross-reference information for validation
• Identify consensus recommendations
• Flag conflicting or outdated information
• Create unified knowledge base

STEP 5: Structured Output

Generate comprehensive report with:

• Executive summary from multi-agent findings
• Key insights by category (technical, security, performance, etc.)
• Confidence scores based on source agreement
• Recommended next steps with priority ranking
• Source quality matrix

STEP 6: Session Cleanup

• Save final synthesized report
• Archive individual agent findings
• Update session state with completion metrics
• Report performance gain (expected 5-10x faster)

Error Handling

TRY:

• Execute primary research workflow

CATCH (WebSearch failures):

• Log error and attempt alternative search terms
• Provide partial results if any sources were successfully fetched

CATCH (WebFetch failures):

• Skip failed sources and continue with available content
• Note limitations in final report

FINALLY:

• Ensure session state is updated
• Clean up temporary files if requested

Quality Assurance

• Verify all sources are accessible and current
• Cross-reference information across multiple sources
• Flag potentially outdated or unreliable information
• Provide confidence indicators for each finding

Research Examples

**Technology Research:**

• "Next.js 15 new features" → Focus on official docs, release notes, migration guides
• "Python asyncio best practices 2024" → Prioritize recent articles, official Python docs
• "Docker security vulnerabilities" → Emphasize security advisories, CVE databases

**Development Focus:**

• Prioritize official documentation over blog posts
• Favor recent content (last 12 months) for rapidly evolving technologies
• Include practical examples and implementation guidance
• Verify information against multiple authoritative sources