dev-team

To run a broad codebase audit across multiple files, you can spawn parallel engineering agents that investigate independent streams simultaneously and synthesize findings into a single unified report.

Parallel investigation is the process of decomposing an oversized engineering task into independent streams handled by simultaneous agents. Use it for broad codebase audits, cross-subsystem performance investigations, and multi-file refactors.

You coordinate multi-agent streams by decomposing the performance investigation into tasks, assigning each stream to a specialized agent, spawning them in parallel, and synthesizing their collected results into one consolidated report.

Yes, concurrent documentation passes are supported. The process requires task decomposition, stream-to-agent assignment, and parallel spawning to analyze documentation simultaneously before merging the outputs into a unified report.

The primary limitation is orchestration overhead. Parallel agent streams require explicit task decomposition, stream-to-agent assignment, and cross-stream synthesis, making it suitable only for oversized engineering tasks rather than small, isolated code changes.