---

name: research description: When and how to research the codebase via the librarian agent

Research - Codebase Discovery via Librarian

This skill covers when and how to use the librarian agent for codebase research and discovery.

When to Research

Research the codebase when you need to:

• Understand existing patterns before proposing new ones
• Find integration points for new functionality
• Locate relevant files for a change
• Understand how something works before modifying it
• Verify assumptions about codebase structure
• Find similar implementations to follow established patterns

The Librarian

The librarian agent is the universal research agent. It:

• Clarifies what you're looking for
• Performs fast reconnaissance and deep codebase analysis exhaustively
• Returns structured, comprehensive findings
• Determines when research is complete

When doing SDD work, prefer using the librarian agent for codebase research so you follow existing patterns and avoid missed context.

How to Consult Librarian

Consult the librarian agent with a clear research prompt (the runtime will route this appropriately):

Copy<your research question>

Good Research Prompts

Be specific about what you need:

Copy# Finding files/locations
"Where are the authentication handlers in this codebase?"

# Understanding patterns
"How does this codebase handle error responses? Show me the pattern."

# Finding integration points
"I need to add a new API endpoint for user preferences. Where should it go and what patterns should it follow?"

# Understanding flows
"Trace how a request flows from the API route to the database for user creation."

# Finding similar implementations
"Find examples of form validation in this codebase so I can follow the same pattern."

What Librarian Returns

Librarian provides:

1. Summary - Direct answer to your question
2. Key Sources - Important file paths with annotations
3. Explanation - How pieces connect
4. Evidence - Specific citations (path/file:line)
5. Next Steps - What to investigate further (if needed)

Research Patterns by Phase

During Proposal

Research to understand:

• Does similar functionality exist?
• What would this change interact with?
• Are there patterns to follow or constraints to respect?

During Specs

Research to understand:

• Where do existing specs live?
• What's the current spec taxonomy?
• Are there related capabilities already specified?

During Discovery

Research to understand:

• How does the proposed change fit existing architecture?
• What code will be affected?
• Are there conflicts with existing patterns?

During Planning

Research to understand:

• Exact file paths for changes
• Current implementation details
• Test patterns to follow
• Integration points

During Implementation

Research when you encounter:

• Unexpected code structure
• Need to understand a dependency
• Unclear how to integrate with existing code

Iterating with Librarian

If librarian's first response doesn't fully answer your question:

1. Ask a follow-up - Be more specific about what's missing
2. Narrow the scope - Focus on one aspect at a time
3. Provide context - Share what you learned so far

Example iteration:

Copy# First query
"How does authentication work in this codebase?"

# Librarian returns high-level overview

# Follow-up query
"Thanks. Now specifically, how are JWT tokens validated? I need to add a new claim check."

When NOT to Research

Skip librarian when:

• You already know the file paths and patterns
• The change is isolated and well-understood
• You're following an existing plan that already did the research
• The task is purely about SDD artifacts (state.md, proposal.md, etc.)

Research First, Then Act

For any non-trivial codebase change:

1. Research first - Understand before modifying
2. Document findings - Capture relevant paths and patterns in your plan
3. Then implement - With confidence about where and how