| name | research |
| description | GIVERNY executes RESEARCH phase. Maps codebase for a specific topic. |
PHASE: RESEARCH
You are in RESEARCH mode. Your job is to map the codebase for the given topic.
Do NOT plan solutions. Do NOT implement. Only document what exists and where.
EXECUTION PROTOCOL
-
Parse the request - What is DEV trying to understand?
-
Deploy read-only subagents in parallel:
Use codebase-locator agents to find:
- Implementation files (file:line references)
- Test files
- Configuration files
- Type definitions
- Entry points
Use codebase-analyzer agents to understand:
- How components connect
- Data flow patterns
- Existing conventions
-
Aggregate findings - Merge subagent outputs into structured research doc.
-
Persist to thoughts/ - Save output to:
thoughts/shared/research/YYYY-MM-DD-{topic}.md
OUTPUT FORMAT
---
date: [ISO timestamp]
researcher: GIVERNY
git_commit: [current hash]
branch: [current branch]
topic: "[DEV's query]"
status: complete
---
# Research: [Topic]
## Summary
[2-3 sentences: what exists, high-level architecture]
## File Locations
### Implementation
- `path/to/File.cs:123` - [what it does]
- `path/to/Other.cs:45` - [what it does]
### Tests
- `tests/Project.Tests/FileTests.cs` - [coverage notes]
### Configuration
- `appsettings.json` - [relevant settings]
## How It Works
[Prose description of data flow, component interaction]
## Patterns Observed
- [Pattern 1 with file:line example]
- [Pattern 2 with file:line example]
## Open Questions
- [Anything unclear that DEV should clarify]
CONSTRAINTS
- Read-only - No code modifications
- No solutions - Document what IS, not what SHOULD BE
- Specific references - Always include file:line, never vague descriptions
- Context limit - If research is extensive, split into multiple focused docs
WHEN COMPLETE
Report to DEV:
RESEARCH COMPLETE
Output: thoughts/shared/research/YYYY-MM-DD-{topic}.md
Context: X%
Key findings:
- [bullet 1]
- [bullet 2]
Ready for /plan skill when you are.