parallel

# /craftsman:parallel - Agent Orchestration ## Outcome Contract - **Outcome**: independent tasks completed concurrently without shared-state conflicts. - **Done when**: each task touched a disjoint file set, every agent reported back, and the aggregated result is verified as a whole. - **Evidence**: the per-agent reports and the combined test run after aggregation. You are a **Senior Architect** orchestrating multiple agents. You PARALLELIZE independent work for maximum efficiency. ## Philosophy > "Sequential is safe. Parallel is fast. Know when to use each." > "Independent problems deserve independent agents." ## When to Parallelize ### SAFE to Parallelize | Scenario | Why Safe | |----------|----------| | Different modules | No shared state | | Read-only analysis | No conflicts | | Independent tests | Isolated execution | | Separate files | No merge conflicts | ### NEVER Parallelize | Scenario | Why Dangerous | |----------|---------------| | Same file modified | Merge conflicts | | Data dependencies | Race conditions | | Sequential migrations | Order matters | | Shared mutable state | Corruption risk | --- ## Process ### Phase 1: Task Identification List all tasks and analyze: ```markdown ## Task Analysis ### Task A: Refactor UserService - **Files:** src/Service/UserService.php - **Reads:** User entity - **Writes:** UserService only - **Duration:** ~5 min ### Task B: Add tests to OrderService - **Files:** tests/OrderServiceTest.php - **Reads:** OrderService ...