/research_codebase

Research and document the codebase as it exists today. This command spawns parallel sub-agents to analyze different parts of the codebase and writes findings to the thoughts directory.

Usage

/research_codebase

The agent will ask for your research question, then:

1. Read any directly mentioned files
2. Spawn parallel sub-agents to explore relevant components
3. Synthesize the findings into one document
4. Save the research to thoughts/shared/ or thoughts/<username>/

Key Behavior

• Documents only. The agent describes what exists, where it lives, and how it works. It will not suggest improvements, critique the implementation, or propose changes unless you ask.
• Uses Opus. The stronger model handles the deep analysis.
• Parallel research. Sub-agents cover different areas of the codebase at once.
• Persistent output. The agent writes research to the thoughts directory, where it outlives the session and your teammates can read it.

Example

/research_codebase

> How does the authentication system work? I need to understand
> the full flow from login to token refresh.

Variants

• /research_codebase_nt: skips thoughts directory integration
• /research_codebase_generic: generic version for use outside Hyprlayer