subagent-driven-development

Read the plan once, create todos for all tasks, then dispatch a fresh implementer subagent per task using the implementer prompt template. After each task, generate a review package and dispatch a task reviewer before marking the task complete.

Subagent-driven development runs in the same session with a fresh subagent per task and continuous execution without human check-ins. Executing plans uses a parallel session with a handoff, which suits tightly coupled tasks or when you want a separate context.

Always specify the model explicitly when dispatching. Use cheap models for mechanical single-file tasks with complete specs, standard models for multi-file integration work, and the most capable model for architecture decisions and the final whole-branch review.

Assess the blocker before re-dispatching: provide more context for context problems, use a more capable model for reasoning-heavy tasks, break oversized tasks into smaller pieces, or escalate to the human if the plan itself is wrong. Never retry the same model unchanged.

A progress ledger file at .superpowers/sdd/progress.md records each completed task with its commit range. After compaction, check the ledger and git log before dispatching anything, and resume at the first task not marked complete.

Avoid it when tasks are tightly coupled and cannot be implemented independently, when no written implementation plan exists yet, or when you cannot run git commands and bash scripts in the working environment.