Skuld is (or aims to become) a hybrid AP/CP distributed task queue, targeting linear scaling with nodes, robustness to N/2-1 failures, extremely high availability for enqueues, guaranteed at-least-once delivery, approximate priority+FIFO ordering, and reasonable bounds on task execution mutexing. Each run of a task can log status updates to Skuld, checkpointing their progress and allowing users to check how far along their tasks have gone. Skuld combines techniques from many distributed systems: Dynamo-style consistent hashing, quorums over vnodes, and anti-entropy provide the highly-available basis for Skuld's immutable dataset, including enqueues, updates, and completions. All AP operations are represented by Convergent Replicated Data Types (CRDTs), ensuring convergence in the absence of strong consistency. CP operations (e.g. claims) are supported by a leader election/quorum protocol similar to Viewstamped Replication or Raft, supported by additional constraints on handoff transitions between disparate cohorts.