What problem does it solve? SparkEngine's runtime core has strict, easy-to-violate rules: one live ECS tick path, one physics stepper per process, main-thread-only registry access, and deferred mutation during entity iteration. This Skill gives an AI assistant the verified ground truth so it can add ECS systems, parallelize work, or debug lifecycle bugs without causing double-ticks, data races, or silent fault-isolated failures. ## Core Features & Use Cases - ECS phase wiring guidance: Explains the PhaseSystemManager phase order, the single production registration site in EngineSetup.h, and the CI-enforced harden tests that prove a system actually ticks. - Threading and synchronization rules: Provides the thread-affinity table for the EnTT registry, PhysicsSystem, JobSystem, and allocators, plus the gather-on-main/process-on-workers pattern. - Lifecycle and memory contracts: Covers DeferredQueue iteration safety, physics stepping ownership, world-origin rebasing hazards, and EngineRuntime ownership doctrine. - Use Case: When asked to add a new ECS system or diagnose why a system stopped ticking, the assistant follows the decision rules and failure-mode checklist instead of guessing from stale docs. ## Quick Start Ask the assistant to add a new ECS system to SparkEngine and have it follow this runbook to register it in the correct phase and prove it ticks with a harden test.