subagents

To prevent context pollution during multi-task implementation plans, dispatch fresh, context-free subagents for each independent task. This isolated subagent delegation eliminates cross-task contamination and keeps the main session focused purely on coordination.

A two-stage code review process enforces a mandatory spec compliance review followed by a code quality review for every task. This approach catches specification gaps and code defects early, allowing up to 5 re-review iterations per stage to ensure high-quality implementation.

You can delegate independent implementation tasks to isolated subagents by using standardized prompt templates for implementers, spec reviewers, and code quality reviewers. This ensures consistent task execution and review criteria across all delegated workstreams.

Yes, the subagents workflow is designed for multi-task feature implementation plans. It delegates each task to a dedicated subagent, runs automated spec and quality checks, and only marks tasks complete once all requirements are fully met.

Context contamination happens in a single implementation session because developers juggle multiple unrelated workstreams simultaneously. Manual execution of multi-task plans causes context pollution between unrelated tasks, leading to inconsistent quality and slow iteration.

Subagents for task delegation are limited to software development workflows with independent, well-specified implementation tasks. They rely on standardized prompts and task persistence sync, meaning poorly specified or highly interdependent tasks may not benefit from this isolated delegation approach.