BMAD-METHOD/bmad-core/tasks/request-research.md

8.4 KiB

Request Research Task

Purpose

This task provides a unified interface for any agent to request specialized research from the Research Coordinator, which can spawn up to 3 domain-specific researcher agents to attack problems from different angles.

Key Features

  • Multi-Perspective Analysis: Coordinator spawns specialized researchers with different domain expertise
  • Web Search Capabilities: Researchers have access to current information and data
  • Adaptive Specialization: Research agents adapt to specific domains as needed by the requesting context
  • Research Logging: All synthesis results stored in indexed research log to avoid duplicate work
  • Configurable Team Size: Default 3 researchers, configurable based on complexity

Usage Scenarios

From Any Agent

Any agent can call this task to get specialized research assistance:

*task request-research

Common Use Cases

  • Analyst: Competitive analysis, market research, industry trends
  • Architect: Technology assessment, scalability analysis, security research
  • PM: Market validation, user research, feasibility studies
  • Dev: Technical implementation research, library comparisons, best practices
  • QA: Testing methodologies, quality standards, compliance requirements

Task Process

1. Research Request Specification

The task will elicit a structured research request with these components:

Research Context

  • Requesting Agent: Which agent is making the request
  • Project Context: Current project phase and relevant background
  • Previous Research: Check research log for related prior work
  • Urgency Level: Timeline constraints and priority

Research Objective

  • Primary Goal: What specific question or problem needs researching
  • Success Criteria: How to measure if research achieved its objective
  • Scope Boundaries: What to include/exclude from research
  • Decision Impact: How results will be used

Domain Specialization Requirements

  • Primary Domain: Main area of expertise needed (technical, market, user, etc.)
  • Secondary Domains: Additional perspectives required
  • Specific Expertise: Particular skills or knowledge areas
  • Research Depth: High-level overview vs deep technical analysis

Output Requirements

  • Format: Executive summary, detailed report, comparison matrix, etc.
  • Audience: Who will consume the research results
  • Integration: How results feed into next steps
  • Documentation: Level of source citation needed

2. Research Coordination

The Research Coordinator will:

  1. Check Research Log: Review docs/research/research-index.md for prior related work
  2. Design Research Strategy: Plan multi-perspective approach
  3. Spawn Researcher Agents: Deploy 1-3 specialized researchers with distinct angles
  4. Monitor Progress: Coordinate between researchers to avoid overlap
  5. Synthesize Results: Combine findings into coherent analysis

3. Research Execution

Each Researcher Agent will:

  1. Adapt Domain Expertise: Configure specialization based on assigned perspective
  2. Conduct Web Research: Use search capabilities to gather current information
  3. Analyze and Synthesize: Process information through domain-specific lens
  4. Generate Findings: Create structured report for their perspective
  5. Cite Sources: Document credible sources and evidence

4. Result Delivery

To Requesting Agent

  • Executive Summary: Key findings and recommendations
  • Detailed Analysis: Comprehensive research results
  • Source Documentation: Links and citations for verification
  • Next Steps: Recommended actions or follow-up research

To Research Log

  • Research Entry: Concise summary stored in docs/research/YYYY-MM-DD-research-topic.md
  • Index Update: Add entry to docs/research/research-index.md
  • Tag Classification: Add searchable tags for future reference

5. Quality Assurance

  • Source Credibility: Verify information from reputable sources
  • Cross-Perspective Validation: Ensure consistency across researcher findings
  • Bias Detection: Identify and flag potential biases or limitations
  • Completeness Check: Confirm all research objectives addressed

Research Request Template

When executing this task, use this structure for research requests:

research_request:
  metadata:
    requesting_agent: '[agent-id]'
    request_date: '[YYYY-MM-DD]'
    priority: '[high|medium|low]'
    timeline: '[timeframe needed]'

  context:
    project_phase: '[planning|development|validation|etc]'
    background: '[relevant project context]'
    related_docs: '[PRD, architecture, stories, etc]'
    previous_research: '[check research log references]'

  objective:
    primary_goal: '[specific research question]'
    success_criteria: '[how to measure success]'
    scope: '[boundaries and limitations]'
    decision_impact: '[how results will be used]'

  specialization:
    primary_domain: '[technical|market|user|competitive|regulatory|etc]'
    secondary_domains: '[additional perspectives needed]'
    specific_expertise: '[particular skills required]'
    research_depth: '[overview|detailed|comprehensive]'

  team_config:
    researcher_count: '[1-3, default 3]'
    perspective_1: '[domain and focus area]'
    perspective_2: '[domain and focus area]'
    perspective_3: '[domain and focus area]'

  output:
    format: '[executive_summary|detailed_report|comparison_matrix|etc]'
    audience: '[who will use results]'
    integration: '[how results feed into workflow]'
    citation_level: '[minimal|standard|comprehensive]'

Integration with Existing Agents

Adding Research Capability to Agents

To add research capabilities to existing agents, add this dependency:

dependencies:
  tasks:
    - request-research.md

Then add a research command:

commands:
  - research {topic}: Request specialized research analysis using task request-research

Research Command Examples

  • *research "competitor API pricing models" (from PM)
  • *research "microservices vs monolith for our scale" (from Architect)
  • *research "React vs Vue for dashboard components" (from Dev)
  • *research "automated testing strategies for ML models" (from QA)

Research Log Structure

Research Index (docs/research/research-index.md)

# Research Index

## Recent Research

- [2024-01-15: AI Model Comparison](2024-01-15-ai-model-comparison.md) - Technical analysis of LLM options
- [2024-01-12: Payment Gateway Analysis](2024-01-12-payment-gateway-analysis.md) - Market comparison of payment solutions

## Research by Category

### Technical Research

- AI/ML Models
- Architecture Decisions
- Technology Stacks

### Market Research

- Competitive Analysis
- User Behavior
- Industry Trends

Individual Research Files (docs/research/YYYY-MM-DD-topic.md)

# Research: [Topic]

**Date**: YYYY-MM-DD
**Requested by**: [agent-name]
**Research Team**: [perspectives used]

## Executive Summary

[Key findings and recommendations]

## Research Objective

[What was being researched and why]

## Key Findings

[Main insights from all perspectives]

## Recommendations

[Actionable next steps]

## Research Team Perspectives

### Perspective 1: [Domain]

[Key insights from this angle]

### Perspective 2: [Domain]

[Key insights from this angle]

### Perspective 3: [Domain]

[Key insights from this angle]

## Sources and References

[Credible sources cited by research team]

## Tags

[Searchable tags for future reference]

Important Notes

  • Research Log Maintenance: Research Coordinator automatically maintains the research index
  • Duplicate Prevention: Always check existing research before launching new requests
  • Source Quality: Prioritize credible, recent sources with proper attribution
  • Perspective Diversity: Ensure research angles provide genuinely different viewpoints
  • Synthesis Quality: Coordinator must reconcile conflicting findings and highlight uncertainties
  • Integration Focus: All research should provide actionable insights for decision-making

Error Handling

  • Web Search Failures: Graceful degradation to available information
  • Conflicting Research: Document disagreements and uncertainty levels
  • Incomplete Coverage: Flag areas needing additional research
  • Source Quality Issues: Clearly mark uncertain or low-confidence findings