Parallel Research Patterns

Patterns for delegating work to multiple sub-agents for maximum parallel efficiency.

Core Principle

NEVER do work that can be parallelized - delegate to multiple sub-agents simultaneously.

Research Phase (Always First)

Before any implementation, launch parallel research:

Technology Detection

Scan project files to identify ALL frameworks, libraries, and tools:

•package.json - Node.js dependencies
•requirements.txt - Python dependencies
•Cargo.toml - Rust dependencies
•go.mod - Go dependencies
•Existing code patterns and conventions

Parallel Research Launch

Deploy 5-8 sub-agents simultaneously:

Agent	Focus Area	Example Prompt
1	Framework docs	"Research Express.js v4.18 official docs for middleware patterns"
2	Codebase patterns	"Analyze `backend/src/routes/` for established routing patterns"
3	Testing patterns	"Research Jest patterns for API testing with Supertest"
4	Language idioms	"Research TypeScript best practices for error handling"
5	Security	"Research OWASP security practices for Express.js APIs"
6+	Specialized	Additional research as needed

Research Prompt Templates

Framework Documentation

code

Research [framework] v[version] official documentation for [feature] implementation patterns

Codebase Analysis

code

Analyze codebase in [directory] to identify established patterns for [functionality]

Testing Approach

code

Research [testing framework] best practices and patterns for [test type] with [libraries]

Integration Research

code

Find best practices for [library] integration with [other tools] in [language]

Security Research

code

Research security best practices for [framework] when implementing [feature type]

Delegation Categories

Research & Analysis (Always First)

•Framework documentation research
•Codebase pattern analysis
•Testing approach research
•Security and performance research
•Language idiom research

Implementation (After Research)

•Component implementation following researched patterns
•Test suite creation using discovered patterns
•Configuration following framework best practices
•Utility functions following language idioms

Quality & Review (Continuous)

•Implementation review against best practices
•Code validation against project standards
•Security review for feature type
•Test coverage verification

Documentation (Parallel)

•Technical documentation updates
•Pattern documentation
•Reference documentation

Parallel Execution Phases

Phase 1: Simultaneous Research (ALL in parallel)

code

Sub-agent 1: Framework/library documentation research
Sub-agent 2: Codebase pattern analysis
Sub-agent 3: Testing approach research
Sub-agent 4: Security and performance research
Sub-agent 5: Language idiom research

Phase 2: Parallel Implementation (after research)

code

Sub-agent 1: Test implementation following discovered patterns
Sub-agent 2: Core functionality implementation
Sub-agent 3: Configuration and setup
Sub-agent 4: Documentation updates
Sub-agent 5: Quality validation

Phase 3: Parallel Validation

code

Sub-agent 1: Integration testing
Sub-agent 2: Performance and security review
Sub-agent 3: Documentation completeness
Sub-agent 4: Code quality review

Research Synthesis

After parallel research completes:

•Compile findings from all sub-agents
•Identify patterns that apply to current task
•Resolve any conflicting recommendations
•Document key decisions for implementation

Benefits

•Maximum parallelization: 5-8 agents working simultaneously
•Comprehensive research: All tools/frameworks thoroughly researched
•Specialized expertise: Each agent focused on specific domain
•Reduced time: Parallel work dramatically speeds development
•Higher quality: Independent research from multiple perspectives
•Framework mastery: Deep understanding of actual tools being used

Anti-Patterns

•Sequential research: Doing one research task at a time
•Skipping research: Implementing without understanding framework patterns
•Incomplete coverage: Only researching some tools/libraries
•Ignoring findings: Not applying researched patterns to implementation
•Duplicate work: Multiple agents researching the same thing

When to Use Parallel Research

•Starting new feature implementation
•Working with unfamiliar frameworks
•Implementing complex integrations
•Establishing new patterns
•Major refactoring efforts

When NOT to Use

•Simple bug fixes with known cause
•Minor changes to existing patterns
•Tasks in well-understood codebase areas