subagent-driven-development

Deploy fresh subagents to handle focused subtasks independently. This isolation prevents context pollution by keeping each subagent's knowledge scoped to its specific deliverable, ensuring clean, uncontaminated outputs across code, research, and documentation work.

Use subagents when you need fresh perspectives, require parallel work on distinct tasks, or want to prevent information from one subtask contaminating another. Define explicit scope, provide context, specify deliverables, and set measurable success criteria for each subagent handoff.

A well-scoped subagent task includes clear scope definition, provided context specific to that task, explicit deliverable specification, and measurable success criteria. This structure ensures the subagent focuses on functional and technical requirements without scope creep or context bleed.

Structure handoffs by defining the scope boundary, providing all necessary context for that subtask, specifying the exact deliverable format and content expected, and establishing measurable success criteria. This prevents context pollution and ensures focused, scoped outputs.

Yes. Subagent-driven development enables parallel work by isolating subtasks so each fresh agent operates independently with its own scoped context. This approach avoids context pollution while allowing simultaneous progress on multiple focused deliverables.

Context pollution occurs when information from unrelated tasks contaminates reasoning and outputs. Subagent-driven development prevents this by enforcing fresh agent deployment per subtask, maintaining isolation so each deliverable reflects only its intended scope and requirements.