subagent-driven-development

Dispatch a fresh implementer subagent per task with a file-based task brief, then dispatch a task reviewer subagent against the generated diff package. Fix Critical and Important findings, re-review, and finish with a whole-branch review after all tasks complete.

Subagent-driven development runs in the same session with fresh subagents per task and continuous execution without human check-ins. Executing plans uses a parallel session with a handoff, which suits plans needing 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 tasks and the final whole-branch review.

Assess the blocker before re-dispatching: provide missing context, switch to a more capable model, break the task into smaller pieces, or escalate to the human if the plan itself is wrong. Never force the same model to retry unchanged.

A progress ledger file at .superpowers/sdd/progress.md records each completed task with its commit range. After compaction, trust the ledger and git log over conversation memory, and never re-dispatch tasks the ledger marks complete.

Avoid it when tasks are tightly coupled and cannot be delegated independently, when no implementation plan exists yet, or when you cannot dispatch subagents in your environment. Brainstorm or write the plan first in those cases.