subagent-driven-development

Subagent-driven development solves this by distributing work to fresh subagents, each owning a task with isolated context. This approach accelerates feedback loops and prevents context leakage while executing complex, multi-task workflows within a single session.

Two-stage review first checks spec compliance to verify requirements are met, then evaluates code quality. This process runs after each individual task completes, ensuring both functional correctness and maintainability across the orchestrated workflow.

Task orchestration is designed for complex, multi-task workflows where tasks are mostly independent. It supports per-task context isolation and template-driven prompts, allowing fast iteration without cross-task interference.

Context isolation means each subagent receives a fresh context for its specific task, preventing leakage from previous tasks. This ensures deterministic progress tracking and maintains clean execution boundaries throughout the orchestration workflow.

Avoid fresh subagent orchestration when tasks are highly dependent on each other or require shared context. This approach works best for mostly independent tasks where isolated execution and fast iteration are prioritized over cross-task communication.