What problem does it solve? When implementing a spec-driven change with many tasks, running them in parallel risks file conflicts and broken dependencies. This Skill inspects the codebase and tasks.md to produce a conservative dependency graph (task-dag.md) that groups tasks into sequential waves of parallel-safe agent bundles. ## Core Features & Use Cases - Dependency Graph Generation: Reads tasks.md, design.md, and proposal.md, then inspects the codebase to identify which files each task touches and which tasks depend on others. - Wave and Bundle Planning: Assigns tasks to sequential waves and bundles them into agent slots (W1-A, W1-B, ...) where bundles in the same wave share no files and can run in parallel. - Use Case: Before launching the sdd-dev-orchestrator for parallel implementation, run this Skill on a change like "add-auth" to get a task-dag.md showing three waves, max parallelism of two agents, and all file conflicts resolved by serialization. ## Quick Start Ask the AI to build the task DAG for your current change, for example: "Generate the task-dag for the add-auth change so I can plan parallel execution."