subagent-driven-development

Subagent-driven development isolates context by dispatching a fresh subagent per task, ensuring rapid iteration without cross-task contamination. Each task runs independently within the same session to maintain a clean execution environment.

The two-stage review process first verifies spec compliance, then checks code quality. A fresh subagent executes the task, followed by a spec-compliance reviewer and a code-quality reviewer running in sequence until all tasks pass.

You can execute complex plans by delegating work to fresh subagents for each task. This orchestrates plan execution by preventing context leakage and enabling rapid iteration across independently runnable tasks.

Yes, you need a SKILL.md frontmatter file containing a name and description. You can also include optional directories such as scripts, references, and assets to support the workflow execution.

Automating spec compliance and code quality checks is best handled by dispatching a fresh subagent per task, then running a spec-compliance reviewer and code-quality reviewer in sequence for each task until all pass.

Per-task subagent execution targets plans with independently runnable tasks within the same session. If your tasks are highly interdependent or require shared context across steps, this isolated subagent approach may not fit.