What problem does it solve?
NexusOS workflows with many batch tasks require reliable dispatch, progress reconciliation, and recovery from context or session limits; manual orchestration is slow, error-prone, and hard to resume. This Skill centralizes scanning DISPATCH.md files, dispatching agents, tracking PROGRESS.md, and preserving checkpoint state so batches can be resumed or paused safely.
Core Features & Use Cases
- Checkpointed Dispatch Loop: Initializes and updates a JSON checkpoint to survive context exhaustion and allow --resume behavior across restarts.
- Targeted and Bulk Processing: Accepts --task-id for single-task handling or --all-active to process all tasks in the active workspace, with configurable batching and concurrency limits.
- Automated Reconciliation and Safety: Periodically reads PROGRESS.md to move tasks between in_progress, completed, and failed, applies a circuit-breaker for repeated failures, and gracefully pauses when limits are reached or context degrades.
Quick Start
Run the auto-run skill with --all-active to dispatch and reconcile all DISPATCH.md tasks in the active workspace.