researching-codebases

Codebase research across multiple files is handled by decomposing the question into a task plan, running parallel sub-agents to trace dependencies concurrently, and synthesizing findings with file:line references into a coherent explanation.

Tracing data flow and architecture across components is achieved by spawning parallel investigations that map interactions independently, then aggregating the results into a contextual synthesis highlighting patterns and references.

Coordinate parallel agents by providing a multi-file codebase question and specifying target modules; the skill generates a decomposed task plan, executes investigations concurrently, and delivers a synthesized summary with open questions.

Parallel codebase research applies to questions spanning multiple files and components, requiring the tracing of interactions and data flow, making it suitable for mapping dependencies between disconnected modules.

Parallel sub-agent codebase synthesis provides a decomposed plan and aggregated summary but may leave open questions, indicating that not all complex architectural queries are fully resolved by the initial parallel execution.