What problem does it solve? Spec Kit's stock task generation marks tasks as parallelizable based on LLM guesses, which can schedule conflicting edits to the same files in parallel and produce broken implementations. This Skill grounds parallel markers in a real dependency graph of the codebase and emits a machine-readable wave plan that a parallel implementer can execute directly. ## Core Features & Use Cases - Graph-verified [P] markers: Uses graphify dependency edges and shared/mutable file lists to confirm or strip parallel markers instead of relying on heuristics. - Execution Waves DAG: Appends a topologically-ordered, machine-readable ## Execution Waves section to tasks.md that /speckit-implement-parallel consumes, packed by greedy earliest-admission to minimize wave count. - Session-sized task merging: Merges structurally dependent tasks (depth capped at 2, never across phase barriers) so each dispatched agent session gets a right-sized unit of work. - Use Case: After running spec-kit plan on a multi-story feature, invoke this Skill to produce a tasks.md where parallel waves are provably conflict-free, then hand it to /speckit-implement-parallel for concurrent agent execution. ## Quick Start Ask the assistant to generate the task list for the current feature using graph-aware parallel markers and an execution waves plan instead of the stock speckit-tasks output.
