subagent-driven-development

Parallel task execution dispatches independent plan steps to multiple subagents within a single session. It partitions concurrent work across isolated files or modules, issuing clear agent briefs and consolidating outputs to reduce overall implementation time.

Automated verification enforces merge checks by running verification-before-completion and self-review steps before integrating subagent outputs. It ensures code coherence and governance by validating isolated work during the final integration check.

Subagent orchestration applies to scenarios where concurrent work can be safely partitioned across files or modules. Tasks modifying the same files without safe isolation boundaries are not suitable for this parallel dispatch approach.

The best way to maintain code coherence is to provide clear agent briefs for each subagent and enforce a final integration check. The Dojo's Sensei and self-review steps verify isolated module work before merging outputs.

You should avoid parallel task dispatch when independent plan steps cannot be safely partitioned across files or modules. If tasks require shared file access or lack clear agent briefs, isolated concurrent work is not feasible and may fail integration checks.