dev-research

name: dev-research description: Development-focused research for codebase understanding, library internals, implementation examples, and architecture decisions. Uses Librarian for code research with GitHub permalinks, Oracle for strategic technical advice, and Explore for fast codebase navigation. USE WHEN user asks about code implementation, library internals, GitHub examples, architecture decisions, or "how does X work in code".

Dev-Research Skill

Development-focused research system for codebase understanding, library analysis, implementation examples, and strategic technical decisions.

When to Use This Skill

USE this skill when:

• User asks "how does library X work?"
• User wants to find code implementation examples
• User asks about library internals or source code
• User needs architecture decisions or code review
• User wants GitHub code examples
• User asks "why was this changed?" (code history)
• User needs strategic technical advice

DO NOT use this skill when:

• User wants general web research → Use research skill
• User asks about current events or news → Use research skill
• User wants to "research topic X" broadly → Use research skill

Available Agents

Librarian (@librarian)

Purpose: Codebase understanding and multi-repository analysis

Use for:

• "How does library X implement Y?"
• "Show me the source code for Z"
• "Find examples of pattern X in open source"
• "What's the history of this feature?"
• Looking up official documentation
• Finding implementation examples with GitHub permalinks

Key Features:

• Returns evidence with GitHub permalinks
• Uses Context7 for official documentation
• Uses grep.app for GitHub code search
• Uses gh CLI for repository analysis

Oracle (@oracle)

Purpose: Strategic technical advisor with deep reasoning

Use for:

• Architecture decisions
• Code review and analysis
• Engineering guidance
• Complex technical questions
• Refactoring roadmaps

Key Features:

• Pragmatic minimalism approach
• Effort estimates (Quick/Short/Medium/Large)
• Single clear recommendation
• Deep reasoning capabilities

Explore (@explore)

Purpose: Fast codebase exploration

Use for:

• "Where is X implemented?"
• "Which files contain Y?"
• "Find the code that does Z"
• Quick file and pattern searches

Key Features:

• Blazing fast searches
• Multiple parallel tool calls
• Returns absolute file paths
• Contextual grep for codebases

Workflow Routing

Implementation Questions

Trigger: "How does X implement Y?", "Show me the source of Z", "Internal logic of..." → Use: @librarian → Workflow: workflows/librarian.md

Strategic/Architecture Questions

Trigger: "Should I use X or Y?", "Review this design", "Best approach for..." → Use: @oracle → Workflow: workflows/oracle.md

Code Location Questions

Trigger: "Where is X?", "Find files with Y", "Which file has Z?" → Use: @explore → Workflow: Fast parallel search

Comprehensive Research

Trigger: Complex questions needing multiple perspectives → Use: @librarian + @oracle in parallel → Process: Librarian gathers evidence, Oracle provides strategic analysis

MCP Servers Used

This skill leverages these MCP servers (configured in OpenCode):

Request Classification

Before taking action, classify the request:

Key Principles

1. Evidence-based answers - Every claim needs a GitHub permalink or source
2. Parallel execution - Launch multiple searches simultaneously
3. Pragmatic advice - Simple solutions over complex ones
4. Fast responses - Speed matters, don't over-research
5. Clear recommendations - One path forward, not endless options

Example Usage

Example 1: Implementation Question

User: "How does React Query implement cache invalidation?"

Process:

1. Route to @librarian
2. Librarian clones tanstack/query repo
3. Searches for cache invalidation logic
4. Returns findings with GitHub permalinks

Example 2: Architecture Decision

User: "Should I use Redux or Zustand for my app?"

Process:

1. Route to @oracle
2. Oracle analyzes requirements
3. Returns single recommendation with effort estimate
4. Includes trade-offs and when to reconsider

Example 3: Code Location

User: "Where is the authentication logic in this codebase?"

Process:

1. Route to @explore
2. Explore launches parallel searches (grep, glob, LSP)
3. Returns file paths with explanations
4. Includes next steps

Example 4: Comprehensive Research

User: "I need to implement real-time updates. What's the best approach and how do others do it?"

Process:

1. Launch @librarian and @oracle in parallel
2. Librarian finds implementation examples (Socket.io, SSE, WebSockets)
3. Oracle provides strategic recommendation for user's context
4. Combine into comprehensive answer