Agent Skills: Feature Research Skill

Feature Research Skill

Purpose

Guide interactive research and understanding of features through conversation, documentation consultation, and selective proof-of-concept (POC) creation. This skill transforms feature ideas into well-understood, documented technical specifications ready for implementation planning.

Note: This skill focuses on TECHNICAL research (HOW to implement). For requirement clarification (WHAT to implement), use feature-specification first.

Input

This skill can receive:

• CDC.md (from feature-specification) - Provides clear, validated requirements to guide research
• Direct feature request - If requirements are already clear

When CDC.md exists, use it as the source of truth for requirements and focus research on technical implementation.

IMPORTANT: User Interaction

ALWAYS use the AskUserQuestion tool when asking clarifying questions.

AskUserQuestion:
  questions:
    - question: "Should we create a POC to validate this approach?"
      header: "POC"
      options:
        - label: "Yes, create POC"
          description: "Validate technical feasibility before proceeding"
        - label: "No, skip POC"
          description: "Documentation is sufficient, proceed to planning"
      multiSelect: false

This ensures:

• Clear, structured questions
• User can choose from options or provide custom response
• Conversation stays focused on technical decisions

When to Use This Skill

Use this skill when:

• CDC.md is ready and technical research is needed
• Starting a new feature implementation with clear requirements
• Researching design patterns or architectural approaches for a feature
• Needing to understand how a feature integrates with existing codebase
• Consulting up-to-date documentation for frameworks, libraries, or APIs
• Validating technical feasibility through minimal POCs
• Gathering information before creating an implementation plan

Research Workflow

Phase 1: Understand the Feature Request

1. Initial Analysis

   • Read the feature request carefully
   • Identify key requirements and constraints
   • List unknowns and areas requiring research

2. Interactive Questioning

   • Ask clarifying questions to the user
   • Understand the "why" behind the feature
   • Identify expected behavior and edge cases
   • Determine acceptance criteria

Example Questions:

• What problem does this feature solve?
• Who are the end users of this feature?
• What are the expected inputs and outputs?
• Are there performance or scalability requirements?
• Are there security or compliance considerations?

Phase 2: Codebase Integration Research

1. Analyze Existing Architecture

   • Use Glob to find relevant files and patterns
   • Use Grep to search for similar implementations
   • Use Read to understand integration points
   • Identify existing patterns and conventions

2. Identify Integration Points

   • Where does this feature fit in the architecture?
   • What existing components will be affected?
   • What new components need to be created?
   • How will data flow through the system?

Search Strategy:

# Example searches (adapt to your project)
# Find similar features
grep -r "pattern:similar-feature" src/

# Find integration points
glob "**/*Controller.cs"  # For .NET APIs
glob "**/*Service.ts"     # For TypeScript services

# Understand patterns
read important-file.ts

Phase 3: Documentation and Design Pattern Research

1. Use MCP Deep Wiki for Accurate Documentation
   • Consult MCP Deep Wiki for framework/library documentation
   • Search for design patterns applicable to your use case
   • Get up-to-date best practices and examples

When to Use Deep Wiki:

• Need framework-specific implementation details
• Want to understand API capabilities
• Looking for recommended design patterns
• Need examples of proper usage

Deep Wiki Usage Tips:

• Be specific in queries: "How to implement authentication in FastAPI"
• Ask for code examples: "Show example of React useEffect with cleanup"
• Request best practices: "Best practices for error handling in .NET Web APIs"

2. Research Design Patterns
   • Identify applicable design patterns (Repository, Factory, Observer, etc.)
   • Understand trade-offs of different approaches
   • Consider patterns used elsewhere in codebase

See references/deep-wiki-usage.md for detailed Deep Wiki guidance.

Phase 4: Proof of Concept (POC) Creation

When to Create a POC:

• Technical feasibility is uncertain
• Need to validate a new library or framework
• Testing integration between components
• Exploring performance characteristics
• User wants to see a working prototype

When NOT to Create a POC:

• Feature is straightforward with clear implementation path
• Similar features already exist in codebase
• Documentation provides sufficient clarity
• User explicitly doesn't need a POC

POC Creation Guidelines:

1. Keep It Minimal

   • Focus on the core concept being validated
   • Don't build production-ready code
   • Use hardcoded data when possible
   • Skip error handling and edge cases

2. Adaptive to Project Context

   • If working within existing project: Create POC in a temporary branch or file
   • If starting from scratch: Create minimal standalone project
   • Use tools available in current environment (React, .NET, Python, etc.)

3. POC Structure

   poc-feature-name/
   ├── README.md           # What is being tested and why
   ├── main.* (or app.*)   # Entry point
   └── [minimal files]     # Only what's needed
   

4. Document POC Results

   • What was validated?
   • What was learned?
   • What issues were discovered?
   • Recommendations for production implementation

See references/poc-guidelines.md for detailed POC creation guidance.

Phase 5: Document Findings

Create a findings document that captures all research results.

Findings Document Structure:

# Feature Research: [Feature Name]

## Overview
Brief description of the feature

## Requirements
- Key requirement 1
- Key requirement 2
- ...

## Design Decisions
### Chosen Approach
Description and rationale

### Alternatives Considered
- Alternative 1: Pros/Cons
- Alternative 2: Pros/Cons

## Integration Points
- Component A: How it integrates
- Component B: How it integrates

## Technical Considerations
- Performance implications
- Security considerations
- Scalability concerns

## Dependencies
- External libraries needed
- Internal components required
- API endpoints to create/modify

## POC Results (if applicable)
- What was tested
- Findings
- Recommendations

## References
- Links to Deep Wiki queries
- Relevant documentation
- Similar implementations in codebase

## Next Steps
- Ready for implementation planning
- Additional research needed (if any)

See references/research-template.md for a complete template.

Iterative Research Process

Research is iterative. The workflow above may repeat multiple times:

1. Ask Questions → Get Answers → Refine Understanding
2. Research Docs → Find Patterns → Adjust Approach
3. Build POC → Learn from Results → Revise Design

Don't Rush: Take time to understand the feature thoroughly. A well-researched feature is faster to implement and has fewer issues.

Output

The output of this skill is a comprehensive findings document (findings.md or research-notes.md) that contains:

• Clear understanding of the feature requirements
• Documented design decisions with rationale
• Identified integration points in codebase
• Technical considerations and dependencies
• POC results (if created)
• References to documentation and research
• Clear next steps for implementation planning

This document becomes the input for the implementation-planner skill.

Tips for Effective Research

1. Ask Before Assuming - Clarify ambiguities with the user early
2. Search Before Creating - Check if similar features exist
3. Document As You Go - Capture findings in real-time
4. Be Thorough - Better to over-research than under-research
5. Validate Uncertainty - Use POCs for technical validation
6. Think About Integration - Consider how the feature fits into the system
7. Use Deep Wiki - Leverage up-to-date documentation
8. Communicate - Keep the user informed throughout the research process

Example Usage

User: "I want to add email notifications when a form is submitted"

Claude (using feature-research skill):
1. "Let me research this feature. A few questions first:
   - Should emails be sent immediately or queued?
   - Are there specific email templates to use?
   - Should we use an external service (SendGrid, etc.) or SMTP?"

2. [Searches codebase for existing email functionality]

3. [Consults Deep Wiki: "How to implement background job queues in .NET"]

4. "I found that we're using Hangfire for background jobs.
   I recommend queuing emails using Hangfire for reliability.
   Should I create a small POC to validate the Microsoft Graph integration?"

5. [Creates minimal POC if user agrees]

6. [Documents findings in findings.md with design decisions,
   integration points, and recommendations]

7. "Research complete! I've documented the approach in findings.md.
   Ready to create the implementation plan?"

Bundled Resources

• references/research-template.md - Template for findings document
• references/deep-wiki-usage.md - Guide for using MCP Deep Wiki effectively
• references/poc-guidelines.md - Detailed POC creation guidelines