subagent-driven-development

To execute implementation plans using subagents, the system dispatches a fresh subagent per task to maintain clean context, followed by a two-stage review process for spec compliance and code quality. This ensures structured plan execution without cross-task contamination.

The two-stage review process first checks spec compliance to verify the implementation meets requirements, then evaluates code quality. It uses independent reviewer roles and scalable gating to adjust review rigor based on task complexity.

Preventing cross-task context contamination involves dispatching fresh subagents per task. This isolates task-focused context, ensuring each implementation step runs independently before triggering structured reporting and escalation paths.

Yes, you can adjust the strictness of automated code reviews using scalable gating. This feature allows you to modify review rigor, and you can optionally elide reviews entirely with user permission depending on task needs.

Subagent-driven development requires three specific roles: an implementer to execute tasks, a spec reviewer to check requirement compliance, and a code-quality reviewer to assess the final output. These roles enforce structured reporting and final checks.

Use subagent-driven development when executing complex implementation plans that require strict spec compliance and code quality enforcement. It provides structured statuses, escalation paths, and model-selection guidance that single continuous workflows lack for scalable task delegation.