What problem does it solve? Executing a multi-task implementation plan in a single AI session leads to context pollution, lost progress after context compaction, and unreviewed code. This Skill orchestrates plan execution by delegating each task to an isolated subagent with its own context, enforcing per-task spec and quality reviews, and tracking progress in a ledger file that survives compaction. ## Core Features & Use Cases - Isolated Task Execution: Dispatches a fresh implementer subagent per task using extracted task briefs, so no subagent inherits session history or unrelated context. - Two-Stage Review Gates: Runs a task-level review (spec compliance plus code quality) after every task and a final whole-branch review with the strongest available model before finishing. - Bounded Fix Loop with Circuit Breaker: Cycles up to five fix rounds per task—re-summoning the original implementer for rounds 1-3, escalating to a stronger model for rounds 4-5—then adjudicates or escalates unresolved findings instead of looping forever. - Compaction-Safe Progress Ledger: Records task completion, fix rounds, and parked findings in a per-plan workspace under .superpowers/sdd/, enabling reliable resume after context loss. - Use Case: You have a written implementation plan with eight mostly independent tasks. Invoke this Skill to execute all tasks end-to-end: each is implemented, tested, reviewed, and committed by dedicated subagents while you retain a clean coordination context. ## Quick Start Execute the implementation plan at docs/plans/feature-plan.md using subagent-driven development, dispatching a fresh implementer subagent for each task and reviewing each one before moving on.
