Research

Research Skill - Multi-Agent Ralph v2.88

Smart Research with Zai MCP Integration - Leverages Zai MCP for comprehensive web search, article fetching, and content analysis.

Based on the principle that research should be thorough, well-sourced, and actionable.

Quick Start

# Via skill invocation
/research Latest React 19 patterns and best practices

# Via CLI
ralph research "TypeScript 5.0 performance optimizations"

# Focus on specific sources
/research "Next.js App Router" --sources github,docs

Key Changes in v2.88

• Zai MCP as primary: Uses mcp__web-search-prime__webSearchPrime instead of Minimax
• Native WebFetch: Uses Claude Code's native WebFetch for content retrieval
• Multi-source fetching: Supports GitHub, CSDN, Juejin, Linux.do article extraction
• Model-agnostic: Works with any model configured in settings.json

Available Tools

1. mcp__web-search-prime__webSearchPrime (Primary Search)

Purpose: Comprehensive web search with high-quality results

Parameters:

search_query: string      # The search query (max 70 characters recommended)
content_size: string      # "medium" (400-600 words) or "high" (2500 words)
location: string          # "cn" (Chinese) or "us" (non-Chinese)
search_recency_filter: string  # oneDay, oneWeek, oneMonth, oneYear, noLimit
search_domain_filter: string   # Optional: limit to specific domain

Optimal Patterns:

# Good: Specific, time-bounded
mcp__web-search-prime__webSearchPrime:
  search_query: "React 19 useOptimistic hook examples 2025"
  search_recency_filter: "oneMonth"
  content_size: "medium"

# Good: Error-focused debugging
mcp__web-search-prime__webSearchPrime:
  search_query: "TypeError cannot read property undefined Next.js 15"
  location: "us"

# Good: Documentation search
mcp__web-search-prime__webSearchPrime:
  search_query: "Claude Code MCP configuration"
  search_domain_filter: "docs.anthropic.com"

# Bad: Too vague
mcp__web-search-prime__webSearchPrime:
  search_query: "javascript"  # Too broad, be specific

2. Native WebFetch (Content Retrieval)

Purpose: Fetch and analyze full web page content

When to Use:

• Deep-dive into specific articles found via search
• Documentation reading
• GitHub repository exploration
• API documentation analysis

3. mcp__web-search__fetchGithubReadme

Purpose: Extract README content from GitHub repositories

Parameters:

url: string  # GitHub repository URL

Example:

mcp__web-search__fetchGithubReadme:
  url: "https://github.com/vercel/next.js"

4. mcp__web-reader__webReader

Purpose: Convert URL content to LLM-friendly format

Parameters:

url: string              # URL to fetch and read
return_format: string    # "markdown" (default) or "text"
retain_images: boolean   # Keep images (default: true)
with_links_summary: boolean  # Include links summary
with_images_summary: boolean # Include images summary

Example:

mcp__web-reader__webReader:
  url: "https://docs.anthropic.com/claude/docs"
  return_format: "markdown"
  with_links_summary: true

5. Chinese Article Fetchers

mcp__web-search__fetchCsdnArticle

url: string  # CSDN article URL

mcp__web-search__fetchJuejinArticle

url: string  # Juejin article URL

mcp__web-search__fetchLinuxDoArticle

url: string  # Linux.do post URL

Research Workflow (5 Steps)

Step 1: INITIAL SEARCH

# Broad search first
mcp__web-search-prime__webSearchPrime:
  search_query: "${TOPIC} overview guide 2025"
  content_size: "high"
  search_recency_filter: "oneMonth"

Step 2: REFINE & DEEPEN

# Targeted follow-up searches based on initial results
mcp__web-search-prime__webSearchPrime:
  search_query: "${SPECIFIC_ASPECT} implementation ${TOPIC}"
  search_recency_filter: "oneWeek"

Step 3: FETCH CONTENT

# For documentation sites
mcp__web-reader__webReader:
  url: "${DOC_URL}"
  return_format: "markdown"

# For GitHub repos
mcp__web-search__fetchGithubReadme:
  url: "${REPO_URL}"

Step 4: SYNTHESIZE

Compile findings into structured report:

• Summary: Key findings in 2-3 sentences
• Sources: All URLs with brief descriptions
• Details: Relevant code snippets and explanations
• Recommendations: Suggested approach based on research
• Related Topics: Areas for further exploration

Step 5: PERSIST

Save research to memory for future reference:

ralph memvid save "Research on ${TOPIC}: [key findings]"

Research Templates

Technology Research

# 1. Official docs
mcp__web-search-prime__webSearchPrime:
  search_query: "${TECH} official documentation"
  search_domain_filter: "official-site.com"

# 2. Best practices
mcp__web-search-prime__webSearchPrime:
  search_query: "${TECH} best practices 2025"

# 3. GitHub examples
mcp__web-search-prime__webSearchPrime:
  search_query: "${TECH} examples github"

Error Research

# 1. Exact error message
mcp__web-search-prime__webSearchPrime:
  search_query: "${ERROR_MESSAGE} ${FRAMEWORK}"

# 2. Stack Overflow solutions
mcp__web-search-prime__webSearchPrime:
  search_query: "site:stackoverflow.com ${ERROR_MESSAGE}"

# 3. GitHub issues
mcp__web-search-prime__webSearchPrime:
  search_query: "site:github.com ${ERROR_MESSAGE}"

Security Research

# 1. CVE lookup
mcp__web-search-prime__webSearchPrime:
  search_query: "CVE ${VERSION} vulnerability"

# 2. Security advisories
mcp__web-search-prime__webSearchPrime:
  search_query: "${PACKAGE} security advisory 2025"

# 3. Exploit database
mcp__web-search-prime__webSearchPrime:
  search_query: "${CVE_ID} exploit"

Integration with Ralph Loop

# Research phase in orchestrator
Task:
  prompt: |
    Research latest patterns for $TOPIC using mcp__web-search-prime__webSearchPrime.
    Use mcp__web-reader__webReader for detailed content.
    Compile findings into structured report with sources.

# Code research
Task:
  prompt: |
    Search for $TOPIC implementation examples on GitHub.
    Use mcp__web-search__fetchGithubReadme to analyze repositories.
    Identify best patterns and anti-patterns.

Comparison: Zai MCP vs Minimax

When to Use Each Tool

Anti-Patterns

• Too broad queries: "javascript" - always be specific
• Skipping sources: Always cite URLs
• Ignoring recency: Use recency filter for fast-moving topics
• Single source: Cross-reference multiple sources
• No synthesis: Don't just list results, analyze them
• Missing memory: Save learnings for future sessions

Output Format

Structure research reports as:

# Research: [TOPIC]

**Date**: YYYY-MM-DD
**Sources**: X articles analyzed

## Summary
[2-3 sentence key findings]

## Key Findings
1. [Finding 1]
   - Source: [URL]
   - Details: [Explanation]

2. [Finding 2]
   - Source: [URL]
   - Details: [Explanation]

## Code Examples
```language
// Relevant code snippets

Recommendations

1. [Recommendation 1]
2. [Recommendation 2]

Related Topics

• [Topic for further research]

Sources

1. Title - [Brief description]
2. Title - [Brief description]

## CLI Commands

```bash
# Standard research
ralph research "topic description"

# With source focus
ralph research "topic" --sources github,docs

# Chinese content
ralph research "topic" --location cn

Related Skills

• /orchestrator - Full orchestration with research phase
• /smart-fork - Pattern extraction from external repos
• /clarify - Requirement clarification with research

Agent Teams Integration (v2.88)

Optimal Scenario: B (Pure Custom Subagents)

Why Scenario B for Research

• Independent execution: Research is mostly self-contained
• Specialization > Coordination: Tool expertise matters more than inter-agent coordination
• Simpler setup: No team overhead for single-purpose research tasks
• Tool restrictions: ralph-researcher has specialized tools (WebSearch, WebFetch)

Scenario Analysis

Workflow

# Scenario B: Direct spawn without TeamCreate
Task(subagent_type="ralph-researcher", prompt="Research ${TOPIC}")
→ Execute with specialized tools
→ Compile structured report
→ Return findings

Usage

Direct Spawn (Recommended):

Task:
  subagent_type: "ralph-researcher"
  prompt: |
    Research ${TOPIC} using:
    1. mcp__web-search-prime__webSearchPrime for initial search
    2. mcp__web-reader__webReader for content extraction
    3. mcp__web-search__fetchGithubReadme for GitHub repos
    Compile into structured report with all sources.

Parallel Research (Multiple Topics):

# Spawn multiple researchers for different topics
Task(subagent_type="ralph-researcher", prompt="Research React 19 features")
Task(subagent_type="ralph-researcher", prompt="Research TypeScript 5.5")
Task(subagent_type="ralph-researcher", prompt="Research Node.js performance")
# Results aggregated independently

References

• Zai MCP Documentation
• Claude Code Skills Documentation
• Web Search Tool Documentation