research-codebase

To document an unfamiliar codebase architecture, you can run a systematic research pass using parallel sub-agents to locate files, analyze structure, and identify design patterns. This generates reference-rich documentation outlining components and data flows.

Understanding cross-module interactions is achieved by applying multi-agent workflows that analyze feature end-to-end flows and synthesize conclusions with precise file paths and line-level evidence, ensuring accurate mapping of components and data flows.

Parallel codebase research works by orchestrating specialized sub-agents, such as codebase-locator, codebase-analyzer, and codebase-pattern-finder, to concurrently scan the repository and identify structural design patterns and architecture decisions.

Yes, you can map a repository's architecture and generate a structured research plan by applying these multi-agent workflows to unfamiliar projects, producing concise blueprints of modules and their interactions for feature development.

Yes, this codebase research approach provides specific file paths and line-level evidence. The multi-agent workflows ensure precise references are included in the synthesized conclusions and architecture documentation.

You should avoid parallel agents for codebase analysis when evaluating extremely small or single-file scripts, as the overhead of orchestrating multiple sub-agents outweighs the benefits of concurrent file location and pattern finding.